Search results for: pixel swapping

Authors: René A. Sørensen, Søren Skovsen, Peter Christiansen, Henrik Karstoft

Abstract:

Autonomous agricultural machines act in stochastic surroundings and therefore, must be able to perceive the surroundings in real time. This perception can be achieved using image sensors combined with advanced machine learning, in particular Deep Learning. Deep convolutional neural networks excel in labeling and perceiving color images and since the cost of high-quality RGB-cameras is low, the hardware cost of good perception depends heavily on memory and computation power. This paper investigates the possibility of designing lightweight convolutional neural networks for semantic segmentation (pixel wise classification) with reduced hardware requirements, to allow for embedded usage in autonomous agricultural machines. Using compression techniques, a lightweight convolutional neural network is designed to perform real-time semantic segmentation on an embedded platform. The network is trained on two large datasets, ImageNet and Pascal Context, to recognize up to 400 individual classes. The 400 classes are remapped into agricultural superclasses (e.g. human, animal, sky, road, field, shelterbelt and obstacle) and the ability to provide accurate real-time perception of agricultural surroundings is studied. The network is applied to the case of autonomous grass mowing using the NVIDIA Tegra X1 embedded platform. Feeding case-specific images to the network results in a fully segmented map of the superclasses in the image. As the network is still being designed and optimized, only a qualitative analysis of the method is complete at the abstract submission deadline. Proceeding this deadline, the finalized design is quantitatively evaluated on 20 annotated grass mowing images. Lightweight convolutional neural networks for semantic segmentation can be implemented on an embedded platform and show competitive performance with regards to accuracy and speed. It is feasible to provide cost-efficient perceptive capabilities related to semantic segmentation for autonomous agricultural machines.

Keywords: autonomous agricultural machines, deep learning, safety, visual perception

Procedia PDF Downloads 396

Authors: Saad Bin Ismail, Muhammad Ateeq Qureshi, Rao Muhammad Zahid Khalil

Abstract:

In the last few decades, urbanization worldwide has been sprawled manifold, which is denunciated in the growth of urban infrastructure and transportation. Urban Heat Island (UHI) can induce deterioration of the living environment, disabilities, and rises in energy usages. In this study, the prevalence/presence of Surface Urban Heat Island (SUHI) effect in major cities of Pakistan, including Islamabad, Rawalpindi, Lahore, Karachi, Quetta, and Peshawar has been investigated. Landsat and SPOT satellite images were acquired for the assessment of urban sprawl. MODIS Land Surface Temperature product MOD11A2 was acquired between 1000-1200 hours (local time) for assessment of urban heat island. The results of urban sprawl informed that the extent of Islamabad and Rawalpindi urban area increased from 240 km2 to 624 km2 between 2000 and 2016, accounted 24 km2 per year, Lahore 29 km2, accounted 1.6 km2 per year, Karachi 261 km2, accounted for 16 km2/ per year, Peshawar 63 km2, accounted 4 km2/per year, and Quetta 76 km2/per year, accounted 5 km2/per year approximately. The average Surface Urban Heat Island (SUHI) magnitude is observed at a scale of 0.63 ᵒC for Islamabad and Rawalpindi, 1.25 ᵒC for Lahore, and 1.16 ᵒC for Karachi, which is 0.89 ᵒC for Quetta, and 1.08 ᵒC for Peshawar from 2000 to 2016. The pixel-based maximum SUHI intensity reaches up to about 11.40 ᵒC for Islamabad and Rawalpindi, 15.66 ᵒC for Lahore, 11.20 ᵒC for Karachi, 14.61 ᵒC for Quetta, and 15.22 ᵒC for Peshawar from the baseline of zero degrees Centigrade (ᵒC). The overall trend of SUHI in planned cities (e.g., Islamabad) is not found to increase significantly. Spatial and temporal patterns of SUHI for selected cities reveal heterogeneity and a unique pattern for each city. It is well recognized that SUHI intensity is modulated by land use/land cover patterns (due to their different surface properties and cooling rates), meteorological conditions, and anthropogenic activities. The study concluded that the selected cities (Islamabad, Rawalpindi, Lahore, Karachi, Quetta, and Peshawar) are examples where dense urban pockets observed about 15 ᵒC warmer than a nearby rural area.

Keywords: urban heat island , surface urban heat island , urbanization, anthropogenic source

Procedia PDF Downloads 322

Authors: Parimalah Velo, Ahmad Zakaria

Abstract:

Introduction: Monte Carlo simulations of preclinical imaging systems allow opportunity to enable new research that could range from designing hardware up to discovery of new imaging application. The simulation system which could accurately model an imaging modality provides a platform for imaging developments that might be inconvenient in physical experiment systems due to the expense, unnecessary radiation exposures and technological difficulties. The aim of present study is to validate the Monte Carlo simulation of thyroid phantom imaging using Geant4-GATE for Siemen’s e-cam single head gamma camera. Upon the validation of the gamma camera simulation model by comparing physical characteristic such as energy resolution, spatial resolution, sensitivity, and dead time, the GATE simulation of thyroid phantom imaging is carried out. Methods: A thyroid phantom is defined geometrically which comprises of 2 lobes with 80mm in diameter, 1 hot spot, and 3 cold spots. This geometry accurately resembling the actual dimensions of thyroid phantom. A planar image of 500k counts with 128x128 matrix size was acquired using simulation model and in actual experimental setup. Upon image acquisition, quantitative image analysis was performed by investigating the total number of counts in image, the contrast of the image, radioactivity distributions on image and the dimension of hot spot. Algorithm for each quantification is described in detail. The difference in estimated and actual values for both simulation and experimental setup is analyzed for radioactivity distribution and dimension of hot spot. Results: The results show that the difference between contrast level of simulation image and experimental image is within 2%. The difference in the total count between simulation and actual study is 0.4%. The results of activity estimation show that the relative difference between estimated and actual activity for experimental and simulation is 4.62% and 3.03% respectively. The deviation in estimated diameter of hot spot for both simulation and experimental study are similar which is 0.5 pixel. In conclusion, the comparisons show good agreement between the simulation and experimental data.

Keywords: gamma camera, Geant4 application of tomographic emission (GATE), Monte Carlo, thyroid imaging

Procedia PDF Downloads 271

Authors: Loonibha Manandhar, Rajendra Bhandari, Kumud Raj Kafle

Abstract:

Nepal’s Terai region is a part of the Ganges river basin which is one of the most disaster-prone areas of the world, with recurrent monsoon flooding causing millions in damage and the death and displacement of hundreds of people and households every year. The vulnerability of human settlements to natural disasters such as floods is increasing, and mapping changes in land use practices and hydro-geological parameters is essential in developing resilient communities and strong disaster management policies. The objective of this study was to develop a flood hazard zonation map of Orai Khola watershed and map the decadal land use/land cover dynamics of the watershed. The watershed area was delineated using SRTM DEM, and LANDSAT images were classified into five land use classes (forest, grassland, sediment and bare land, settlement area and cropland, and water body) using pixel-based semi-automated supervised maximum likelihood classification. Decadal changes in each class were then quantified using spatial modelling. Flood hazard mapping was performed by assigning weights to factors slope, rainfall distribution, distance from the river and land use/land cover on the basis of their estimated influence in causing flood hazard and performing weighed overlay analysis to identify areas that are highly vulnerable. The forest and grassland coverage increased by 11.53 km² (3.8%) and 1.43 km² (0.47%) from 1996 to 2016. The sediment and bare land areas decreased by 12.45 km² (4.12%) from 1996 to 2016 whereas settlement and cropland areas showed a consistent increase to 14.22 km² (4.7%). Waterbody coverage also increased to 0.3 km² (0.09%) from 1996-2016. 1.27% (3.65 km²) of total watershed area was categorized into very low hazard zone, 20.94% (60.31 km²) area into low hazard zone, 37.59% (108.3 km²) area into moderate hazard zone, 29.25% (84.27 km²) area into high hazard zone and 31 villages which comprised 10.95% (31.55 km²) were categorized into high hazard zone area.

Keywords: flood hazard, land use/land cover, Orai river, supervised maximum likelihood classification, weighed overlay analysis

Procedia PDF Downloads 353

Authors: Evan Lowhorn, Rocio Alba-Flores

Abstract:

The purpose of this work is to develop a method for classifying hand signals and using the output in a drone control algorithm. To achieve this, methods based on Convolutional Neural Networks (CNN) were applied. CNN's are a subset of deep learning, which allows grid-like inputs to be processed and passed through a neural network to be trained for classification. This type of neural network allows for classification via imaging, which is less intrusive than previous methods using biosensors, such as EMG sensors. Classification CNN's operate purely from the pixel values in an image; therefore they can be used without additional exteroceptive sensors. A development bench was constructed using a desktop computer connected to a high-definition webcam mounted on a scissor arm. This allowed the camera to be pointed downwards at the desk to provide a constant solid background for the dataset and a clear detection area for the user. A MATLAB script was created to automate dataset image capture at the development bench and save the images to the desktop. This allowed the user to create their own dataset of 12,000 images within three hours. These images were evenly distributed among seven classes. The defined classes include forward, backward, left, right, idle, and land. The drone has a popular flip function which was also included as an additional class. To simplify control, the corresponding hand signals chosen were the numerical hand signs for one through five for movements, a fist for land, and the universal “ok” sign for the flip command. Transfer learning with PyTorch (Python) was performed using a pre-trained 18-layer residual learning network (ResNet-18) to retrain the network for custom classification. An algorithm was created to interpret the classification and send encoded messages to a Ryze Tello drone over its 2.4 GHz Wi-Fi connection. The drone’s movements were performed in half-meter distance increments at a constant speed. When combined with the drone control algorithm, the classification performed as desired with negligible latency when compared to the delay in the drone’s movement commands.

Keywords: classification, computer vision, convolutional neural networks, drone control

Procedia PDF Downloads 210

Authors: Zhengming Wu

Abstract:

NanoFrazor lithography systems were developed as a first true alternative or extension to standard mask-less nanolithography methods like electron beam lithography (EBL). In contrast to EBL they are based on thermal scanning probe lithography (t-SPL). Here a heatable ultra-sharp probe tip with an apex of a few nm is used for patterning and simultaneously inspecting complex nanostructures. The heat impact from the probe on a thermal responsive resist generates those high-resolution nanostructures. The patterning depth of each individual pixel can be controlled with better than 1 nm precision using an integrated in-situ metrology method. Furthermore, the inherent imaging capability of the Nanofrazor technology allows for markerless overlay, which has been achieved with sub-5 nm accuracy as well as it supports stitching layout sections together with < 10 nm error. Pattern transfer from such resist features below 10 nm resolution were demonstrated. The technology has proven its value as an enabler of new kinds of ultra-high resolution nanodevices as well as for improving the performance of existing device concepts. The application range for this new nanolithography technique is very broad spanning from ultra-high resolution 2D and 3D patterning to chemical and physical modification of matter at the nanoscale. Nanometer-precise markerless overlay and non-invasiveness to sensitive materials are among the key strengths of the technology. However, while patterning at below 10 nm resolution is achieved, significantly increasing the patterning speed at the expense of resolution is not feasible by using the heated tip alone. Towards this end, an integrated laser write head for direct laser sublimation (DLS) of the thermal resist has been introduced for significantly faster patterning of micrometer to millimeter-scale features. Remarkably, the areas patterned by the tip and the laser are seamlessly stitched together and both processes work on the very same resist material enabling a true mix-and-match process with no developing or any other processing steps in between. The presentation will include examples for (i) high-quality metal contacting of 2D materials, (ii) tuning photonic molecules, (iii) generating nanofluidic devices and (iv) generating spintronic circuits. Some of these applications have been enabled only due to the various unique capabilities of NanoFrazor lithography like the absence of damage from a charged particle beam.

Keywords: nanofabrication, grayscale lithography, 2D materials device, nano-optics, photonics, spintronic circuits

Procedia PDF Downloads 72

Authors: Victor Bloch, Eerikki Kaila, Reetta Palva

Abstract:

Field robots are an important tool for enhancing efficiency and decreasing the climatic impact of food production. There exists a number of commercial field robots; however, since this technology is still new, the robot advantages and limitations, as well as methods for optimal using of robots, are still unclear. In this study, the performance of a commercial field robot for seeding and weeding was assessed. A research 2-ha sugar beet field with 0.5m row width was used for testing, which included robotic sowing of sugar beet and weeding five times during the first two months of the growing. About three and five percent of the field were used as untreated and chemically weeded control areas, respectively. The plant detection was based on the exact plant location without image processing. The robot was equipped with six seeding and weeding tools, including passive between-rows harrow hoes and active hoes cutting inside rows between the plants, and it moved with a maximal speed of 0.9 km/h. The robot's performance was assessed by image processing. The field images were collected by an action camera with a height of 2 m and a resolution 27M pixels installed on the robot and by a drone with a 16M pixel camera flying at 4 m height. To detect plants and weeds, the YOLO model was trained with transfer learning from two available datasets. A preliminary analysis of the entire field showed that in the areas treated by the robot, the weed average density varied across the field from 6.8 to 9.1 weeds/m² (compared with 0.8 in the chemically treated area and 24.3 in the untreated area), the weed average density inside rows was 2.0-2.9 weeds / m (compared with 0 on the chemically treated area), and the emergence rate was 90-95%. The information about the robot's performance has high importance for the application of robotics for field tasks. With the help of the developed method, the performance can be assessed several times during the growth according to the robotic weeding frequency. When it’s used by farmers, they can know the field condition and efficiency of the robotic treatment all over the field. Farmers and researchers could develop optimal strategies for using the robot, such as seeding and weeding timing, robot settings, and plant and field parameters and geometry. The robot producers can have quantitative information from an actual working environment and improve the robots accordingly.

Keywords: agricultural robot, field robot, plant detection, robot performance

Procedia PDF Downloads 87

Authors: Molakala Mourya Vardhan Reddy, Kandimala Revanth, Koduru Sumanth, Beena B. M.

Abstract:

Emergency response and recovery operations are severely hampered by natural catastrophes, especially floods. Understanding post-flood scenarios is essential to disaster management because it facilitates quick evaluation and decision-making. To this end, we introduce FloodNet, a brand-new high-resolution aerial picture collection created especially for comprehending post-flood scenes. A varied collection of excellent aerial photos taken during and after flood occurrences make up FloodNet, which offers comprehensive representations of flooded landscapes, damaged infrastructure, and changed topographies. The dataset provides a thorough resource for training and assessing computer vision models designed to handle the complexity of post-flood scenarios, including a variety of environmental conditions and geographic regions. Pixel-level semantic segmentation masks are used to label the pictures in FloodNet, allowing for a more detailed examination of flood-related characteristics, including debris, water bodies, and damaged structures. Furthermore, temporal and positional metadata improve the dataset's usefulness for longitudinal research and spatiotemporal analysis. For activities like flood extent mapping, damage assessment, and infrastructure recovery projection, we provide baseline standards and evaluation metrics to promote research and development in the field of post-flood scene comprehension. By integrating FloodNet into machine learning pipelines, it will be easier to create reliable algorithms that will help politicians, urban planners, and first responders make choices both before and after floods. The goal of the FloodNet dataset is to support advances in computer vision, remote sensing, and disaster response technologies by providing a useful resource for researchers. FloodNet helps to create creative solutions for boosting communities' resilience in the face of natural catastrophes by tackling the particular problems presented by post-flood situations.

Keywords: image classification, segmentation, computer vision, nature disaster, unmanned arial vehicle(UAV), machine learning.

Procedia PDF Downloads 78

Authors: E. M. D. Akuoko, Eliana Vasquez Osorio, Marcel Van Herk, Marianne Aznar

Abstract:

Radiotherapy is an integral part of treatment for many patients with breast cancer. However, side effects can occur, e.g., fibrosis or erythema. If patients at higher risks of radiation-induced side effects could be identified before treatment, they could be given more individual information about the risks and benefits of radiotherapy. We hypothesize that breast density is correlated with the risk of side effects and present a novel method for automatic evaluation based on radiotherapy planning CT scans. Methods: 799 supine CT scans of breast radiotherapy patients were available from the REQUITE dataset. The methodology was first established in a subset of 114 patients (cohort 1) before being applied to the whole dataset (cohort 2). All patients were scanned in the supine position, with arms up, and the treated breast (ipsilateral) was identified. Manual experts contour available in 96 patients for both the ipsilateral and contralateral breast in cohort 1. Breast tissue was segmented using atlas-based automatic contouring software, ADMIRE® v3.4 (Elekta AB, Sweden). Once validated, the automatic segmentation method was applied to cohort 2. Breast density was then investigated by thresholding voxels within the contours, using Otsu threshold and pixel intensity ranges based on Hounsfield units (-200 to -100 for fatty tissue, and -99 to +100 for fibro-glandular tissue). Volumetric breast density (VBD) was defined as the volume of fibro-glandular tissue / (volume of fibro-glandular tissue + volume of fatty tissue). A sensitivity analysis was performed to verify whether calculated VBD was affected by the choice of breast contour. In addition, we investigated the correlation between volumetric breast density (VBD) and patient age and breast size. VBD values were compared between ipsilateral and contralateral breast contours. Results: Estimated VBD values were 0.40 (range 0.17-0.91) in cohort 1, and 0.43 (0.096-0.99) in cohort 2. We observed ipsilateral breasts to be denser than contralateral breasts. Breast density was negatively associated with breast volume (Spearman: R=-0.5, p-value < 2.2e-16) and age (Spearman: R=-0.24, p-value = 4.6e-10). Conclusion: VBD estimates could be obtained automatically on a large CT dataset. Patients’ age or breast volume may not be the only variables that explain breast density. Future work will focus on assessing the usefulness of VBD as a predictive variable for radiation-induced side effects.

Keywords: breast cancer, automatic image segmentation, radiotherapy, big data, breast density, medical imaging

Procedia PDF Downloads 132

Authors: Nassima Noufail, Sara Bouhali

Abstract:

In this work, we develop a semi-supervised solution for the purpose of action detection in videos and propose an efficient algorithm for video segmentation. The approach is divided into video segmentation, feature extraction, and classification. In the first part, a video is segmented into clips, and we used the K-means algorithm for this segmentation; our goal is to find groups based on similarity in the video. The application of k-means clustering into all the frames is time-consuming; therefore, we started by the identification of transition frames where the scene in the video changes significantly, and then we applied K-means clustering into these transition frames. We used two image filters, the gaussian filter and the Laplacian of Gaussian. Each filter extracts a set of features from the frames. The Gaussian filter blurs the image and omits the higher frequencies, and the Laplacian of gaussian detects regions of rapid intensity changes; we then used this vector of filter responses as an input to our k-means algorithm. The output is a set of cluster centers. Each video frame pixel is then mapped to the nearest cluster center and painted with a corresponding color to form a visual map. The resulting visual map had similar pixels grouped. We then computed a cluster score indicating how clusters are near each other and plotted a signal representing frame number vs. clustering score. Our hypothesis was that the evolution of the signal would not change if semantically related events were happening in the scene. We marked the breakpoints at which the root mean square level of the signal changes significantly, and each breakpoint is an indication of the beginning of a new video segment. In the second part, for each segment from part 1, we randomly selected a 16-frame clip, then we extracted spatiotemporal features using convolutional 3D network C3D for every 16 frames using a pre-trained model. The C3D final output is a 512-feature vector dimension; hence we used principal component analysis (PCA) for dimensionality reduction. The final part is the classification. The C3D feature vectors are used as input to a multi-class linear support vector machine (SVM) for the training model, and we used a multi-classifier to detect the action. We evaluated our experiment on the UCF101 dataset, which consists of 101 human action categories, and we achieved an accuracy that outperforms the state of art by 1.2%.

Keywords: video segmentation, action detection, classification, Kmeans, C3D

Procedia PDF Downloads 77

Authors: Inigo Beckett

Abstract:

In forensic investigations, it is often the case that the most potentially useful recorded evidence derives from coincidental imagery, recorded immediately before or during an incident, and that during the incident (e.g. a ‘failure’ or fire event), the evidence is changed or destroyed. To an image analysis expert involved in photogrammetric analysis for Civil or Criminal Proceedings, traditional computer vision methods involving calibrated cameras is often not appropriate because image metadata cannot be relied upon. This paper presents an approach for resolving this problem, considering in particular and by way of a case study, the angle of a simple ladder shown in a photograph. The UK Health and Safety Executive (HSE) guidance document published in 2014 (INDG455) advises that a leaning ladder should be erected at 75 degrees to the horizontal axis. Personal injury cases can arise in the construction industry because a ladder is too steep or too shallow. Ad-hoc photographs of such ladders in their incident position provide a basis for analysis of their angle. This paper presents a direct approach for ascertaining the position of the camera and the angle of the ladder simultaneously from the photograph(s) by way of a workflow that encompasses a novel application of projective geometry and matrix analysis. Mathematical analysis shows that for a given pixel ratio of directly measured collinear points (i.e. features that lie on the same line segment) from the 2D digital photograph with respect to a given viewing point, we can constrain the 3D camera position to a surface of a sphere in the scene. Depending on what we know about the ladder, we can enforce another independent constraint on the possible camera positions which enables us to constrain the possible positions even further. Experiments were conducted using synthetic and real-world data. The synthetic data modeled a vertical plane with a ladder on a horizontally flat plane resting against a vertical wall. The real-world data was captured using an Apple iPhone 13 Pro and 3D laser scan survey data whereby a ladder was placed in a known location and angle to the vertical axis. For each case, we calculated camera positions and the ladder angles using this method and cross-compared them against their respective ‘true’ values.

Keywords: image analysis, projective geometry, homography, photogrammetry, ladders, Forensics, Mathematical modeling, planar geometry, matrix analysis, collinear, cameras, photographs

Procedia PDF Downloads 52

Authors: Ali Hamza

Abstract:

Accurate segmentation of breast ultrasound images is of paramount importance in enhancing the diagnostic capabilities of breast cancer detection. This study presents an approach utilizing the U-Net architecture for segmenting breast ultrasound images aimed at improving the accuracy and reliability of mass identification within the breast tissue. The proposed method encompasses a multi-stage process. Initially, preprocessing techniques are employed to refine image quality and diminish noise interference. Subsequently, the U-Net architecture, a deep learning convolutional neural network (CNN), is employed for pixel-wise segmentation of regions of interest corresponding to potential breast masses. The U-Net's distinctive architecture, characterized by a contracting and expansive pathway, enables accurate boundary delineation and detailed feature extraction. To evaluate the effectiveness of the proposed approach, an extensive dataset of breast ultrasound images is employed, encompassing diverse cases. Quantitative performance metrics such as the Dice coefficient, Jaccard index, sensitivity, specificity, and Hausdorff distance are employed to comprehensively assess the segmentation accuracy. Comparative analyses against traditional segmentation methods showcase the superiority of the U-Net architecture in capturing intricate details and accurately segmenting breast masses. The outcomes of this study emphasize the potential of the U-Net-based segmentation approach in bolstering breast ultrasound image analysis. The method's ability to reliably pinpoint mass boundaries holds promise for aiding radiologists in precise diagnosis and treatment planning. However, further validation and integration within clinical workflows are necessary to ascertain their practical clinical utility and facilitate seamless adoption by healthcare professionals. In conclusion, leveraging the U-Net architecture for breast ultrasound image segmentation showcases a robust framework that can significantly enhance diagnostic accuracy and advance the field of breast cancer detection. This approach represents a pivotal step towards empowering medical professionals with a more potent tool for early and accurate breast cancer diagnosis.

Keywords: mage segmentation, U-Net, deep learning, breast cancer detection, diagnostic accuracy, mass identification, convolutional neural network

Procedia PDF Downloads 84

Authors: Kirolos Gerges Yakoub Gerges

Abstract:

Self-sustaining agricultural machines act in stochastic surroundings and therefore, should be capable of perceive the surroundings in real time. This notion can be done using image sensors blended with superior device learning, mainly Deep mastering. Deep convolutional neural networks excel in labeling and perceiving colour pix and since the fee of RGB-cameras is low, the hardware cost of accurate notion relies upon heavily on memory and computation power. This paper investigates the opportunity of designing lightweight convolutional neural networks for semantic segmentation (pixel clever class) with reduced hardware requirements, to allow for embedded usage in self-reliant agricultural machines. The usage of compression techniques, a lightweight convolutional neural community is designed to carry out actual-time semantic segmentation on an embedded platform. The community is skilled on two big datasets, ImageNet and Pascal Context, to apprehend as much as four hundred man or woman instructions. The 400 training are remapped into agricultural superclasses (e.g. human, animal, sky, road, area, shelterbelt and impediment) and the capacity to provide correct actual-time perception of agricultural environment is studied. The network is carried out to the case of self-sufficient grass mowing the usage of the NVIDIA Tegra X1 embedded platform. Feeding case-unique pics to the community consequences in a fully segmented map of the superclasses within the picture. As the network remains being designed and optimized, handiest a qualitative analysis of the technique is entire on the abstract submission deadline. intending this cut-off date, the finalized layout is quantitatively evaluated on 20 annotated grass mowing pictures. Light-weight convolutional neural networks for semantic segmentation can be implemented on an embedded platform and show aggressive performance on the subject of accuracy and speed. It’s miles viable to offer value-efficient perceptive capabilities related to semantic segmentation for autonomous agricultural machines.

Keywords: centrifuge pump, hydraulic energy, agricultural applications, irrigationaxial flux machines, axial flux applications, coreless machines, PM machinesautonomous agricultural machines, deep learning, safety, visual perception

Procedia PDF Downloads 26

Authors: Lalit Kumar, Rashid Al Shidi

Abstract:

Phoenix dactylifera (date palm) is a major crop in many middle-eastern countries, including Oman. The Dubas bug Ommatissus lybicus is the main pest that affects date palm crops. However not all plantations are infested. It is still uncertain why some plantations get infested while others are not. This research investigated whether tree density and the system of planting (random versus systematic) had any relationship with infestation and levels of infestation. Remote Sensing and Geographic Information Systems were used to determine the density of trees (number of trees per unit area) while infestation levels were determined by manual counting of insects on 40 leaflets from two fronds on each tree, with a total of 20-60 trees in each village. The infestation was recorded as the average number of insects per leaflet. For tree density estimation, WorldView-3 scenes, with eight bands and 2m spatial resolution, were used. The Local maxima method, which depends on locating of the pixel of highest brightness inside a certain exploration window, was used to identify the trees in the image and delineating individual trees. This information was then used to determine whether the plantation was random or systematic. The ordinary least square regression (OLS) was used to test the global correlation between tree density and infestation level and the Geographic Weight Regression (GWR) was used to find the local spatial relationship. The accuracy of detecting trees varied from 83–99% in agricultural lands with systematic planting patterns to 50–70% in natural forest areas. Results revealed that the density of the trees in most of the villages was higher than the recommended planting number (120–125 trees/hectare). For infestation correlations, the GWR model showed a good positive significant relationship between infestation and tree density in the spring season with R² = 0.60 and medium positive significant relationship in the autumn season, with R² = 0.30. In contrast, the OLS model results showed a weaker positive significant relationship in the spring season with R² = 0.02, p < 0.05 and insignificant relationship in the autumn season with R² = 0.01, p > 0.05. The results showed a positive correlation between infestation and tree density, which suggests the infestation severity increased as the density of date palm trees increased. The correlation result showed that the density alone was responsible for about 60% of the increase in the infestation. This information can be used by the relevant authorities to better control infestations as well as to manage their pesticide spraying programs.

Keywords: dubas bug, date palm, tree density, infestation levels

Procedia PDF Downloads 193

Authors: Marek Dosbaba

Abstract:

Within the mining sector, SEM-based Automated Mineralogy (AM) has been the standard application for quickly and efficiently handling mineral processing tasks. Over the last decade, the trend has been to analyze larger numbers of samples, often with a higher level of detail. This has necessitated a shift from interactive sample analysis performed by an operator using a SEM, to an increased reliance on offline processing to analyze and report the data. In response to this trend, TESCAN TIMA Mineral Analyzer is designed to quickly create a virtual copy of the studied samples, thereby preserving all the necessary information. Depending on the selected data acquisition mode, TESCAN TIMA can perform hyperspectral mapping and save an X-ray spectrum for each pixel or segment, respectively. This approach allows the user to browse through elemental distribution maps of all elements detectable by means of energy dispersive spectroscopy. Re-evaluation of the existing data for the presence of previously unconsidered elements is possible without the need to repeat the analysis. Additional tiers of data such as a secondary electron or cathodoluminescence images can also be recorded. To take full advantage of these information-rich datasets, TIMA utilizes a new archiving tool introduced by TESCAN. The dataset size can be reduced for long-term storage and all information can be recovered on-demand in case of renewed interest. TESCAN TIMA is optimized for network storage of its datasets because of the larger data storage capacity of servers compared to local drives, which also allows multiple users to access the data remotely. This goes hand in hand with the support of remote control for the entire data acquisition process. TESCAN also brings a newly extended open-source data format that allows other applications to extract, process and report AM data. This offers the ability to link TIMA data to large databases feeding plant performance dashboards or geometallurgical models. The traditional tabular particle-by-particle or grain-by-grain export process is preserved and can be customized with scripts to include user-defined particle/grain properties.

Keywords: Tescan, electron microscopy, mineralogy, SEM, automated mineralogy, database, TESCAN TIMA, open format, archiving, big data

Procedia PDF Downloads 110

Authors: Walid Wahballah, Taher Bazan, Fawzy Eltohamy

Abstract:

One of the important functions of the satellite attitude control system is to provide the required pointing accuracy and attitude stability for optical remote sensing satellites to achieve good image quality. Although offering noise reduction and increased sensitivity, time delay and integration (TDI) charge coupled devices (CCDs) utilized in high-resolution satellites (HRS) are prone to introduce large amounts of pixel smear due to the instability of the line of sight. During on-orbit imaging, as a result of the Earth’s rotation and the satellite platform instability, the moving direction of the TDI-CCD linear array and the imaging direction of the camera become different. The speed of the image moving on the image plane (focal plane) represents the image motion velocity whereas the angle between the two directions is known as the drift angle (β). The drift angle occurs due to the rotation of the earth around its axis during satellite imaging; affecting the geometric accuracy and, consequently, causing image quality degradation. Therefore, the image motion velocity vector and the drift angle are two important factors used in the assessment of the image quality of TDI-CCD based optical remote sensing satellites. A model for estimating the image motion velocity and the drift angle in HRS is derived. The six satellite attitude control parameters represented in the derived model are the (roll angle φ, pitch angle θ, yaw angle ψ, roll angular velocity φ֗, pitch angular velocity θ֗ and yaw angular velocity ψ֗ ). The influence of these attitude parameters on the image quality is analyzed by establishing a relationship between the image motion velocity vector, drift angle and the six satellite attitude parameters. The influence of the satellite attitude parameters on the image quality is assessed by the presented model in terms of modulation transfer function (MTF) in both cross- and along-track directions. Three different cases representing the effect of pointing accuracy (φ, θ, ψ) bias are considered using four different sets of pointing accuracy typical values, while the satellite attitude stability parameters are ideal. In the same manner, the influence of satellite attitude stability (φ֗, θ֗, ψ֗) on image quality is also analysed for ideal pointing accuracy parameters. The results reveal that cross-track image quality is influenced seriously by the yaw angle bias and the roll angular velocity bias, while along-track image quality is influenced only by the pitch angular velocity bias.

Keywords: high-resolution satellites, pointing accuracy, attitude stability, TDI-CCD, smear, MTF

Procedia PDF Downloads 402

Authors: Bilel Chalghaf, Mathieu Varin

Abstract:

Forest characterization in Quebec, Canada, is usually assessed based on photo-interpretation at the stand level. For species identification, this often results in a lack of precision. Very high spatial resolution imagery, such as DigitalGlobe, and Light Detection and Ranging (LiDAR), have the potential to overcome the limitations of aerial imagery. To date, few studies have used that data to map a large number of species at the tree level using machine learning techniques. The main objective of this study is to map 11 individual high tree species ( > 17m) at the tree level using an object-based approach in the broadleaf forest of Kenauk Nature, Quebec. For the individual tree crown segmentation, three canopy-height models (CHMs) from LiDAR data were assessed: 1) the original, 2) a filtered, and 3) a corrected model. The corrected CHM gave the best accuracy and was then coupled with imagery to refine tree species crown identification. When compared with photo-interpretation, 90% of the objects represented a single species. For modeling, 313 variables were derived from 16-band WorldView-3 imagery and LiDAR data, using radiance, reflectance, pixel, and object-based calculation techniques. Variable selection procedures were employed to reduce their number from 313 to 16, using only 11 bands to aid reproducibility. For classification, a global approach using all 11 species was compared to a semi-hierarchical hybrid classification approach at two levels: (1) tree type (broadleaf/conifer) and (2) individual broadleaf (five) and conifer (six) species. Five different model techniques were used: (1) support vector machine (SVM), (2) classification and regression tree (CART), (3) random forest (RF), (4) k-nearest neighbors (k-NN), and (5) linear discriminant analysis (LDA). Each model was tuned separately for all approaches and levels. For the global approach, the best model was the SVM using eight variables (overall accuracy (OA): 80%, Kappa: 0.77). With the semi-hierarchical hybrid approach, at the tree type level, the best model was the k-NN using six variables (OA: 100% and Kappa: 1.00). At the level of identifying broadleaf and conifer species, the best model was the SVM, with OA of 80% and 97% and Kappa values of 0.74 and 0.97, respectively, using seven variables for both models. This paper demonstrates that a hybrid classification approach gives better results and that using 16-band WorldView-3 with LiDAR data leads to more precise predictions for tree segmentation and classification, especially when the number of tree species is large.

Keywords: tree species, object-based, classification, multispectral, machine learning, WorldView-3, LiDAR

Procedia PDF Downloads 134

Authors: D. Signore, M. Ferraiuolo, P. Caramuta, O. Petrella, C. Toscano

Abstract:

The Digital Image Correlation (DIC) is a newly developed optical technique that is spreading in all engineering sectors because it allows the non-destructive estimation of the entire surface deformation without any contact with the component under analysis. These characteristics make the DIC very appealing in all the cases the global deformation state is to be known without using strain gages, which are the most used measuring device. The DIC is applicable to any material subjected to distortion caused by either thermal or mechanical load, allowing to obtain high-definition mapping of displacements and deformations. That is why in the civil and the transportation industry, DIC is very useful for studying the behavior of metallic materials as well as of composite materials. DIC is also used in the medical field for the characterization of the local strain field of the vascular tissues surface subjected to uniaxial tensile loading. DIC can be carried out in the two dimension mode (2D DIC) if a single camera is used or in a three dimension mode (3D DIC) if two cameras are involved. Each point of the test surface framed by the cameras can be associated with a specific pixel of the image, and the coordinates of each point are calculated knowing the relative distance between the two cameras together with their orientation. In both arrangements, when a component is subjected to a load, several images related to different deformation states can be are acquired through the cameras. A specific software analyzes the images via the mutual correlation between the reference image (obtained without any applied load) and those acquired during the deformation giving the relative displacements. In this paper, a metrological characterization of the digital image correlation is performed on aluminum and composite targets both in static and dynamic loading conditions by comparison between DIC and strain gauges measures. In the static test, interesting results have been obtained thanks to an excellent agreement between the two measuring techniques. In addition, the deformation detected by the DIC is compliant with the result of a FEM simulation. In the dynamic test, the DIC was able to follow with a good accuracy the periodic deformation of the specimen giving results coherent with the ones given by FEM simulation. In both situations, it was seen that the DIC measurement accuracy depends on several parameters such as the optical focusing, the parameters chosen to perform the mutual correlation between the images and, finally, the reference points on image to be analyzed. In the future, the influence of these parameters will be studied, and a method to increase the accuracy of the measurements will be developed in accordance with the requirements of the industries especially of the aerospace one.

Keywords: accuracy, deformation, image correlation, mechanical analysis

Procedia PDF Downloads 311

Authors: Oskar Sekowski

Abstract:

Current global warming causes changes in the traditional zones of viticulture worldwide. During 20th century, the average global air temperature increased by 0.89˚C. The models of climate change indicate that viticulture, currently concentrating in narrow geographic niches, may move towards the poles, to higher geographic latitudes. Global warming may cause changes in traditional viticulture regions. Therefore, there is a need to estimate the climatic conditions and climate change in areas that are not traditionally associated with viticulture, e.g., Poland. The primary objective of this paper is to prepare methodology to evaluate the climatic and meteorological potential for viticulture in Poland based on a case study. Moreover, the additional aim is to evaluate the climatic potential of a mesoregion where a university vineyard is located. The daily data of temperature, precipitation, insolation, and wind speed (1988-2018) from the meteorological station located in Łazy, southern Poland, was used to evaluate 15 climatological parameters and indices connected with viticulture. The next steps of the methodology are based on Geographic Information System methods. The topographical factors such as a slope gradient and slope exposure were created using Digital Elevation Models. The spatial distribution of climatological elements was interpolated by ordinary kriging. The values of each factor and indices were also ranked and classified. The viticultural potential was determined by integrating two suitability maps, i.e., the topographical and climatic ones, and by calculating the average for each pixel. Data analysis shows significant changes in heat accumulation indices that are driven by increases in maximum temperature, mostly increasing number of days with Tmax > 30˚C. The climatic conditions of this mesoregion are sufficient for vitis vinifera viticulture. The values of indicators and insolation are similar to those in the known wine regions located on similar geographical latitudes in Europe. The smallest threat to viticulture in study area is the occurrence of hail and the highest occurrence of frost in the winter. This research provides the basis for evaluating general suitability and climatologic potential for viticulture in Poland. To characterize the climatic potential for viticulture, it is necessary to assess the suitability of all climatological and topographical factors that can influence viticulture. The methodology used in this case study shows places where there is a possibility to create vineyards. It may also be helpful for wine-makers to select grape varieties.

Keywords: climatologic potential, climatic classification, Poland, viticulture

Procedia PDF Downloads 106

Authors: A. G. Cunha, M. Inácio, M. C. Freitas, C. Antunes, T. Silva, C. Andrade, V. Lopes

Abstract:

Saltmarshes are essential ecosystems both from an ecological and biological point of view. Furthermore, they constitute an important social niche, providing valuable economic and protection functions. Thus, understanding their rates and patterns of sedimentation is critical for functional management and rehabilitation, especially in an SLR scenario. The Sado estuary is located 40 km south of Lisbon. It is a bar built estuary, separated from the sea by a large sand spit: the Tróia barrier. Caldeira de Tróia is located on the free edge of this barrier, and encompasses a salt marsh with ca. 21,000 m². Sediment cores were collected in the high and low marshes and in the mudflat area of the North bank of Caldeira de Tróia. From the low marsh core, fifteen samples were chosen for ²¹⁰Pb and ¹³⁷Cs determination at University of Geneva. The cores from the high marsh and the mudflat are still being analyzed. A sedimentation rate of 2.96 mm/year was derived from ²¹⁰Pb using the Constant Flux Constant Sedimentation model. The ¹³⁷Cs profile shows a peak in activity (1963) between 15.50 and 18.50 cm, giving a 3.1 mm/year sedimentation rate for the past 53 years. The adopted sea level rise scenario was based on a model built with the initial rate of SLR of 2.1 mm/year in 2000 and an acceleration of 0.08 mm/year². Based on the harmonic analysis of Setubal-Tróia tide gauge of 2005 data, the tide model was estimated and used to build the tidal tables to the period 2000-2016. With these tables, the average mean water levels were determined for the same time span. A digital terrain model was created from LIDAR scanning with 2m horizontal resolution (APA-DGT, 2011) and validated with altimetric data obtained with a DGPS-RTK. The response model calculates a new elevation for each pixel of the DTM for 2050 and 2100 based on the sedimentation rates specific of each environment. At this stage, theoretical values were chosen for the high marsh and the mudflat (respectively, equal and double the low marsh rate – 2.92 mm/year). These values will be rectified once sedimentation rates are determined for the other environments. For both projections, the total surface of the marsh decreases: 2% in 2050 and 61% in 2100. Additionally, the high marsh coverage diminishes significantly, indicating a regression in terms of maturity.

Keywords: ¹³⁷Cs, ²¹⁰Pb, saltmarsh, sea level rise, response model

Procedia PDF Downloads 250

Authors: Shubham Gogri, Lucia Florescu

Abstract:

Medical diagnosis and radiotherapy treatment planning using Computed Tomography (CT) rely on the quantitative accuracy and quality of the CT images. At the same time, requirements for CT imaging include reducing the radiation dose exposure to patients and minimizing scanning time. A solution to this is the sparse-view CT technique, based on a reduced number of projection views. This, however, introduces a new problem— the incomplete projection data results in lower quality of the reconstructed images. To tackle this issue, deep learning methods have been applied to enhance the quality of the sparse-view CT images. A first approach involved employing Mir-Net, a dedicated deep neural network designed for image enhancement. This showed promise, utilizing an intricate architecture comprising encoder and decoder networks, along with the incorporation of the Charbonnier Loss. However, this approach was computationally demanding. Subsequently, a specialized Generative Adversarial Network (GAN) architecture, rooted in the Pix2Pix framework, was implemented. This GAN framework involves a U-Net-based Generator and a Discriminator based on Convolutional Neural Networks. To bolster the GAN's performance, both Charbonnier and Wasserstein loss functions were introduced, collectively focusing on capturing minute details while ensuring training stability. The integration of the perceptual loss, calculated based on feature vectors extracted from the VGG16 network pretrained on the ImageNet dataset, further enhanced the network's ability to synthesize relevant images. A series of comprehensive experiments with clinical CT data were conducted, exploring various GAN loss functions, including Wasserstein, Charbonnier, and perceptual loss. The outcomes demonstrated significant image quality improvements, confirmed through pertinent metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) between the corrected images and the ground truth. Furthermore, learning curves and qualitative comparisons added evidence of the enhanced image quality and the network's increased stability, while preserving pixel value intensity. The experiments underscored the potential of deep learning frameworks in enhancing the visual interpretation of CT scans, achieving outcomes with SSIM values close to one and PSNR values reaching up to 76.

Keywords: generative adversarial networks, sparse view computed tomography, CT image correction, Mir-Net

Procedia PDF Downloads 162

Authors: Farina Riaz, Shahab Abdulla, Srinjoy Ganguly, Hajime Suzuki, Ravinesh C. Deo, Susan Hopkins

Abstract:

Recently, quantum machine learning has received significant attention. For various types of data, including text and images, numerous quantum machine learning (QML) models have been created and are being tested. Images are exceedingly complex data components that demand more processing power. Despite being mature, classical machine learning still has difficulties with big data applications. Furthermore, quantum technology has revolutionized how machine learning is thought of, by employing quantum features to address optimization issues. Since quantum hardware is currently extremely noisy, it is not practicable to run machine learning algorithms on it without risking the production of inaccurate results. To discover the advantages of quantum versus classical approaches, this research has concentrated on colored image data. Deep learning classification models are currently being created on Quantum platforms, but they are still in a very early stage. Black and white benchmark image datasets like MNIST and Fashion MINIST have been used in recent research. MNIST and CIFAR-10 were compared for binary classification, but the comparison showed that MNIST performed more accurately than colored CIFAR-10. This research will evaluate the performance of the QML algorithm on the colored benchmark dataset CIFAR-10 to advance QML's real-time applicability. However, deep learning classification models have not been developed to compare colored images like Quantum Convolutional Neural Network (QCNN) to determine how much it is better to classical. Only a few models, such as quantum variational circuits, take colored images. The methodology adopted in this research is a hybrid approach by using penny lane as a simulator. To process the 10 classes of CIFAR-10, the image data has been translated into grey scale and the 28 × 28-pixel image containing 10,000 test and 50,000 training images were used. The objective of this work is to determine how much the quantum approach can outperform a classical approach for a comprehensive dataset of color images. After pre-processing 50,000 images from a classical computer, the QCNN model adopted a hybrid method and encoded the images into a quantum simulator for feature extraction using quantum gate rotations. The measurements were carried out on the classical computer after the rotations were applied. According to the results, we note that the QCNN approach is ~12% more effective than the traditional classical CNN approaches and it is possible that applying data augmentation may increase the accuracy. This study has demonstrated that quantum machine and deep learning models can be relatively superior to the classical machine learning approaches in terms of their processing speed and accuracy when used to perform classification on colored classes.

Keywords: CIFAR-10, quantum convolutional neural networks, quantum deep learning, quantum machine learning

Procedia PDF Downloads 129

Authors: Bencherif Kada

Abstract:

In the forest management practice, landscape and Mediterranean forest are never posed as linked objects. But sustainable forestry requires the valorization of the forest landscape, and this aim involves assessing the spatial distribution of biodiversity by mapping forest landscaped units and subunits and by monitoring the environmental trends. This contribution aims to highlight, through object-oriented classifications, the landscaped biodiversity of the National Park of Tlemcen (Algeria). The methodology used is based on ground data and on the basic processing units of object-oriented classification, that are segments, so-called image-objects, representing a relatively homogenous units on the ground. The classification of Landsat Enhanced Thematic Mapper plus (ETM+) imagery is performed on image objects and not on pixels. Advantages of object-oriented classification are to make full use of meaningful statistic and texture calculation, uncorrelated shape information (e.g., length-to-width ratio, direction, and area of an object, etc.), and topological features (neighbor, super-object, etc.), and the close relation between real-world objects and image objects. The results show that per object classification using the k-nearest neighbor’s method is more efficient than per pixel one. It permits to simplify of the content of the image while preserving spectrally and spatially homogeneous types of land covers such as Aleppo pine stands, cork oak groves, mixed groves of cork oak, holm oak, and zen oak, mixed groves of holm oak and thuja, water plan, dense and open shrub-lands of oaks, vegetable crops or orchard, herbaceous plants, and bare soils. Texture attributes seem to provide no useful information, while spatial attributes of shape and compactness seem to be performant for all the dominant features, such as pure stands of Aleppo pine and/or cork oak and bare soils. Landscaped sub-units are individualized while conserving the spatial information. Continuously dominant dense stands over a large area were formed into a single class, such as dense, fragmented stands with clear stands. Low shrublands formations and high wooded shrublands are well individualized but with some confusion with enclaves for the former. Overall, a visual evaluation of the classification shows that the classification reflects the actual spatial state of the study area at the landscape level.

Keywords: forest, oaks, remote sensing, diversity, shrublands

Procedia PDF Downloads 124

Authors: Aitor Bilbao, Dragos Axinte, John Billingham

Abstract:

The inverse problem in Energy Beam (EB) Processes consists of defining the control parameters, in particular the 2D beam path (position and orientation of the beam as a function of time), to arrive at a prescribed solution (freeform surface). This inverse problem is well understood for conventional machining, because the cutting tool geometry is well defined and the material removal is a time independent process. In contrast, EB machining is achieved through the local interaction of a beam of particular characteristics (e.g. energy distribution), which leads to a surface-dependent removal rate. Furthermore, EB machining is a time-dependent process in which not only the beam varies with the dwell time, but any acceleration/deceleration of the machine/beam delivery system, when performing raster paths will influence the actual geometry of the surface to be generated. Two different EB processes, Abrasive Water Machining (AWJM) and Pulsed Laser Ablation (PLA), are studied. Even though they are considered as independent different technologies, both can be described as time-dependent processes. AWJM can be considered as a continuous process and the etched material depends on the feed speed of the jet at each instant during the process. On the other hand, PLA processes are usually defined as discrete systems and the total removed material is calculated by the summation of the different pulses shot during the process. The overlapping of these shots depends on the feed speed and the frequency between two consecutive shots. However, if the feed speed is sufficiently slow compared with the frequency, then consecutive shots are close enough and the behaviour can be similar to a continuous process. Using this approximation a generic continuous model can be described for both processes. The inverse problem is usually solved for this kind of process by simply controlling dwell time in proportion to the required depth of milling at each single pixel on the surface using a linear model of the process. However, this approach does not always lead to the good solution since linear models are only valid when shallow surfaces are etched. The solution of the inverse problem is improved by using a discrete adjoint optimization algorithm. Moreover, the calculation of the Jacobian matrix consumes less computation time than finite difference approaches. The influence of the dynamics of the machine on the actual movement of the jet is also important and should be taken into account. When the parameters of the controller are not known or cannot be changed, a simple approximation is used for the choice of the slope of a step profile. Several experimental tests are performed for both technologies to show the usefulness of this approach.

Keywords: abrasive waterjet machining, energy beam processes, inverse problem, pulsed laser ablation

Procedia PDF Downloads 275

Authors: Bencherif Kada

Abstract:

In forest management practice, landscape and Mediterranean forest are never posed as linked objects. But sustainable forestry requires the valorization of the forest landscape and this aim involves assessing the spatial distribution of biodiversity by mapping forest landscaped units and subunits and by monitoring the environmental trends. This contribution aims to highlight, through object-oriented classifications, the landscaped biodiversity of the National Park of Tlemcen (Algeria). The methodology used is based on ground data and on the basic processing units of object-oriented classification that are segments, so-called image-objects, representing a relatively homogenous units on the ground. The classification of Landsat Enhanced Thematic Mapper plus (ETM+) imagery is performed on image objects, and not on pixels. Advantages of object-oriented classification are to make full use of meaningful statistic and texture calculation, uncorrelated shape information (e.g., length-to-width ratio, direction and area of an object, etc.) and topological features (neighbor, super-object, etc.), and the close relation between real-world objects and image objects. The results show that per object classification using the k-nearest neighbor’s method is more efficient than per pixel one. It permits to simplify the content of the image while preserving spectrally and spatially homogeneous types of land covers such as Aleppo pine stands, cork oak groves, mixed groves of cork oak, holm oak and zen oak, mixed groves of holm oak and thuja, water plan, dense and open shrub-lands of oaks, vegetable crops or orchard, herbaceous plants and bare soils. Texture attributes seem to provide no useful information while spatial attributes of shape, compactness seem to be performant for all the dominant features, such as pure stands of Aleppo pine and/or cork oak and bare soils. Landscaped sub-units are individualized while conserving the spatial information. Continuously dominant dense stands over a large area were formed into a single class, such as dense, fragmented stands with clear stands. Low shrublands formations and high wooded shrublands are well individualized but with some confusion with enclaves for the former. Overall, a visual evaluation of the classification shows that the classification reflects the actual spatial state of the study area at the landscape level.

Keywords: forest, oaks, remote sensing, biodiversity, shrublands

Procedia PDF Downloads 30

Authors: Fatemeh Deldarabdolmaleki, Mohamad Fakri Zaky Bin Ja'afar

Abstract:

Having a daylit space together with view results in a pleasant and productive environment for office employees. A daylit space is a space which utilizes daylight as a basic source of illumination to fulfill user’s visual demands and minimizes the electric energy consumption. Malaysian weather is hot and humid all over the year because of its location in the equatorial belt. however, because most of the commercial buildings in Malaysia are air-conditioned, huge glass windows are normally installed in order to keep the physical and visual relation between inside and outside. As a result of climatic situation and mentioned new trend, an ordinary office has huge heat gain, glare, and discomfort for occupants. Balancing occupant’s comfort and energy conservation in a tropical climate is a real challenge. This study concentrates on evaluating a venetian blind system using per pixel analyzing tools based on the suggested cut-out metrics by the literature. Workplace area in a private office room has been selected as a case study. Eight-day measurement experiment was conducted to investigate the effect of different venetian blind angles in an office area under daylight conditions in Serdang, Malaysia. The study goal was to explore daylight comfort of a commercially available venetian blind system, its’ daylight sufficiency and excess (8:00 AM to 5 PM) as well as Glare examination. Recently developed software, analyzing High Dynamic Range Images (HDRI captured by CCD camera), such as radiance based Evalglare and hdrscope help to investigate luminance-based metrics. The main key factors are illuminance and luminance levels, mean and maximum luminance, daylight glare probability (DGP) and luminance ratio of the selected mask regions. The findings show that in most cases, morning session needs artificial lighting in order to achieve daylight comfort. However, in some conditions (e.g. 10° and 40° slat angles) in the second half of day the workplane illuminance level exceeds the maximum of 2000 lx. Generally, a rising trend is discovered toward mean window luminance and the most unpleasant cases occur after 2 P.M. Considering the luminance criteria rating, the uncomfortable conditions occur in the afternoon session. Surprisingly in no blind condition, extreme case of window/task ratio is not common. Studying the daylight glare probability, there is not any DGP value higher than 0.35 in this experiment.

Keywords: daylighting, energy simulation, office environment, Venetian blind

Procedia PDF Downloads 259

Authors: Mehrnaz Mostafavi

Abstract:

Recent advances in single-molecule protein identification (ID) and quantification techniques are poised to revolutionize proteomics, enabling researchers to delve into single-cell proteomics and identify low-abundance proteins crucial for biomedical and clinical research. This paper introduces a different approach to single-molecule protein ID and quantification using tri-color amino acid tags and a plasmonic nanopore device. A comprehensive simulator incorporating various physical phenomena was designed to predict and model the device's behavior under diverse experimental conditions, providing insights into its feasibility and limitations. The study employs a whole-proteome single-molecule identification algorithm based on convolutional neural networks, achieving high accuracies (>90%), particularly in challenging conditions (95–97%). To address potential challenges in clinical samples, where post-translational modifications affecting labeling efficiency, the paper evaluates protein identification accuracy under partial labeling conditions. Solid-state nanopores, capable of processing tens of individual proteins per second, are explored as a platform for this method. Unlike techniques relying solely on ion-current measurements, this approach enables parallel readout using high-density nanopore arrays and multi-pixel single-photon sensors. Convolutional neural networks contribute to the method's versatility and robustness, simplifying calibration procedures and potentially allowing protein ID based on partial reads. The study also discusses the efficacy of the approach in real experimental conditions, resolving functionally similar proteins. The theoretical analysis, protein labeler program, finite difference time domain calculation of plasmonic fields, and simulation of nanopore-based optical sensing are detailed in the methods section. The study anticipates further exploration of temporal distributions of protein translocation dwell-times and the impact on convolutional neural network identification accuracy. Overall, the research presents a promising avenue for advancing single-molecule protein identification and quantification with broad applications in proteomics research. The contributions made in methodology, accuracy, robustness, and technological exploration collectively position this work at the forefront of transformative developments in the field.

Keywords: nano proteomics, nanopore-based optical sensing, deep learning, artificial intelligence

Procedia PDF Downloads 96

Authors: Yuta Nakano, Kozo Kajiwara, Atsushi Hori, Takeshi Fujita

Abstract:

In association with the wave of electric vehicles (EV), low energy consumption systems have become more and more important. One of the key technologies to realize low energy consumption is a dynamic vision sensor (DVS), or we can call it an event sensor, neuromorphic vision sensor and so on. This sensor has several features, such as high temporal resolution, which can achieve 1 Mframe/s, and a high dynamic range (120 DB). However, the point that can contribute to low energy consumption the most is its sparsity; to be more specific, this sensor only captures the pixels that have intensity change. In other words, there is no signal in the area that does not have any intensity change. That is to say, this sensor is more energy efficient than conventional sensors such as RGB cameras because we can remove redundant data. On the other side of the advantages, it is difficult to handle the data because the data format is completely different from RGB image; for example, acquired signals are asynchronous and sparse, and each signal is composed of x-y coordinate, polarity (two values: +1 or -1) and time stamp, it does not include intensity such as RGB values. Therefore, as we cannot use existing algorithms straightforwardly, we have to design a new processing algorithm to cope with DVS data. In order to solve difficulties caused by data format differences, most of the prior arts make a frame data and feed it to deep learning such as Convolutional Neural Networks (CNN) for object detection and recognition purposes. However, even though we can feed the data, it is still difficult to achieve good performance due to a lack of intensity information. Although polarity is often used as intensity instead of RGB pixel value, it is apparent that polarity information is not rich enough. Considering this context, we proposed to use the timestamp information as a data representation that is fed to deep learning. Concretely, at first, we also make frame data divided by a certain time period, then give intensity value in response to the timestamp in each frame; for example, a high value is given on a recent signal. We expected that this data representation could capture the features, especially of moving objects, because timestamp represents the movement direction and speed. By using this proposal method, we made our own dataset by DVS fixed on a parked car to develop an application for a surveillance system that can detect persons around the car. We think DVS is one of the ideal sensors for surveillance purposes because this sensor can run for a long time with low energy consumption in a NOT dynamic situation. For comparison purposes, we reproduced state of the art method as a benchmark, which makes frames the same as us and feeds polarity information to CNN. Then, we measured the object detection performances of the benchmark and ours on the same dataset. As a result, our method achieved a maximum of 7 points greater than the benchmark in the F1 score.

Keywords: event camera, dynamic vision sensor, deep learning, data representation, object recognition, low energy consumption

Procedia PDF Downloads 97

Authors: Tauqir A. Moughal, Fusheng Yu, Abeer Mazher

Abstract:

Land-cover and land-use change information are important because of their practical uses in various applications, including deforestation, damage assessment, disasters monitoring, urban expansion, planning, and land management. Therefore, developing change detection methods for remote sensing images is an important ongoing research agenda. However, detection of change through optical remote sensing images is not a trivial task due to many factors including the vagueness between the boundaries of changed and unchanged regions and spatial dependence of the pixels to its neighborhood. In this paper, we propose a binary change detection technique for bi-temporal optical remote sensing images. As in most of the optical remote sensing images, the transition between the two clusters (change and no change) is overlapping and the existing methods are incapable of providing the accurate cluster boundaries. In this regard, a methodology has been proposed which uses the fuzzy c-means clustering to tackle the problem of vagueness in the changed and unchanged class by formulating the soft boundaries between them. Furthermore, in order to exploit the neighborhood information of the pixels, the input patterns are generated corresponding to each pixel from bi-temporal images using 3×3, 5×5 and 7×7 window. The between images and within image spatial dependence of the pixels to its neighborhood is quantified by using Pearson product moment correlation and Moran’s I statistics, respectively. The proposed technique consists of two phases. At first, between images and within image spatial correlation is calculated to utilize the information that the pixels at different locations may not be independent. Second, fuzzy c-means technique is used to produce two clusters from input feature by not only taking care of vagueness between the changed and unchanged class but also by exploiting the spatial correlation of the pixels. To show the effectiveness of the proposed technique, experiments are conducted on multispectral and bi-temporal remote sensing images. A subset (2100×1212 pixels) of a pan-sharpened, bi-temporal Landsat 5 thematic mapper optical image of Los Angeles, California, is used in this study which shows a long period of the forest fire continued from July until October 2009. Early forest fire and later forest fire optical remote sensing images were acquired on July 5, 2009 and October 25, 2009, respectively. The proposed technique is used to detect the fire (which causes change on earth’s surface) and compared with the existing K-means clustering technique. Experimental results showed that proposed technique performs better than the already existing technique. The proposed technique can be easily extendable for optical hyperspectral images and is suitable for many practical applications.

Keywords: burned area, change detection, correlation, fuzzy clustering, optical remote sensing

Procedia PDF Downloads 169

Authors: Miguel Contreras, David Long, Will Bachman

Abstract:

Background: Microscopy plays a central role in cell and developmental biology. In particular, fluorescence microscopy can be used to visualize specific cellular components and subsequently quantify their morphology through development of virtual-cell models for study of effects of mechanical forces on cells. However, there are challenges with these imaging experiments, which can make it difficult to quantify cell morphology: inconsistent results, time-consuming and potentially costly protocols, and limitation on number of labels due to spectral overlap. To address these challenges, the objective of this project is to develop an automatic computational machine learning pipeline to predict cellular components morphology for virtual-cell generation based on fluorescence cell membrane confocal z-stacks. Methods: Registered confocal z-stacks of nuclei and cell membrane of endothelial cells, consisting of 20 images each, were obtained from fluorescence confocal microscopy and normalized through software pipeline for each image to have a mean pixel intensity value of 0.5. An open source machine learning algorithm, originally developed to predict fluorescence labels on unlabeled transmitted light microscopy cell images, was trained using this set of normalized z-stacks on a single CPU machine. Through transfer learning, the algorithm used knowledge acquired from its previous training sessions to learn the new task. Once trained, the algorithm was used to predict morphology of nuclei using normalized cell membrane fluorescence images as input. Predictions were compared to the ground truth fluorescence nuclei images. Results: After one week of training, using one cell membrane z-stack (20 images) and corresponding nuclei label, results showed qualitatively good predictions on training set. The algorithm was able to accurately predict nuclei locations as well as shape when fed only fluorescence membrane images. Similar training sessions with improved membrane image quality, including clear lining and shape of the membrane, clearly showing the boundaries of each cell, proportionally improved nuclei predictions, reducing errors relative to ground truth. Discussion: These results show the potential of pre-trained machine learning algorithms to predict cell morphology using relatively small amounts of data and training time, eliminating the need of using multiple labels in immunofluorescence experiments. With further training, the algorithm is expected to predict different labels (e.g., focal-adhesion sites, cytoskeleton), which can be added to the automatic machine learning pipeline for direct input into Principal Component Analysis (PCA) for generation of virtual-cell mechanical models.

Keywords: cell morphology prediction, computational machine learning, fluorescence microscopy, virtual-cell models

Procedia PDF Downloads 205