Image Classification and Accuracy Assessment Using the Confusion Matrix, Contingency Matrix, and Kappa Coefficient

Authors: F. F. Howard, C. B. Boye, I. Yakubu, J. S. Y. Kuma

Abstract:

One of the ways that could be used for the production of land use and land cover maps by a procedure known as image classification is the use of the remote sensing technique. Numerous elements ought to be taken into consideration, including the availability of highly satisfactory Landsat imagery, secondary data and a precise classification process. The goal of this study was to classify and map the land use and land cover of the study area using remote sensing and Geospatial Information System (GIS) analysis. The classification was done using Landsat 8 satellite images acquired in December 2020 covering the study area. The Landsat image was downloaded from the USGS. The Landsat image with 30 m resolution was geo-referenced to the WGS_84 datum and Universal Transverse Mercator (UTM) Zone 30N coordinate projection system. A radiometric correction was applied to the image to reduce the noise in the image. This study consists of two sections: the Land Use/Land Cover (LULC) and Accuracy Assessments using the confusion and contingency matrix and the Kappa coefficient. The LULC classifications were vegetation (agriculture) (67.87%), water bodies (0.01%), mining areas (5.24%), forest (26.02%), and settlement (0.88%). The overall accuracy of 97.87% and the kappa coefficient (K) of 97.3% were obtained for the confusion matrix. While an overall accuracy of 95.7% and a Kappa coefficient of 0.947 were obtained for the contingency matrix, the kappa coefficients were rated as substantial; hence, the classified image is fit for further research.

Keywords: Confusion Matrix, contingency matrix, kappa coefficient, land used/ land cover, accuracy assessment.

References:

[1] Angima, S. D., Stott, D. E., O’neill, M. K., Ong, C. K., & Weesies, G. A. (2003). Soil erosion prediction using RUSLE for central Kenyan highland conditions. Agriculture, ecosystems & environment, 97(1-3), pp. 295-308.
[2] Congalton RG. (2004), “Putting the map back in map accuracy assessment. In: Geospatial Data
[3] Congalton RG, Green K. (1999), “Assessing the accuracy of remotely sensed data: Principles and Practices”. Lewis Publisher, Boca Raton,
[4] Congalton RG. (1991) “A review of assessing the accuracy of classifications of remotely sensed Data”. Remote Sensing of Environment; 37(1): pp. 35–46.
[5] Disperati, L., Gonario S. and Virdis, P. (2015) “Assessment of Land-Use and Land-Cover Changes from 1965 to 2014 in Tam Giang-Cau Hai Lagoon, Central Vietnam”. Applied Geography, 58, pp. 48-64.
[6] Dominati, E., Patterson, M. and Mackay, A. (2010) A Framework for Classifying and Quantifying the Natural Capital and Ecosystem Services of Soils. Ecological Economics, 69, pp. 1858-1868.
[7] Hammond TO, Verbyla DL (1996). Optimistic bias in classification accuracy assessment. International Journal of Remote Sensing; 17(6): pp. 1261-66.
[8] Manisha B. G Chitra & N Umrikar (2012). “Image Classification Tool For Land Use/ Land Cover Analysis”: A Comparative Study of Maximum Likelihood and Minimum Distance Method. Int. J. Geo. earth. environ. , 6, pp. 189-96.
[9] Kuma J.S, Jerome A. Yendaw (2010), InternationalJournal of Geosciences, 2010, 1, 113-120 doi:10.4236/ijg.2010.13015 Published Online November, 2010 (http://www.SciRP.org/journal/ijg).
[10] Pouyat, R., Groffman, P., Yesilonis, I., & Hernandez, L. (2002). Soil carbon pools and fluxes in urban ecosystems. Environmental pollution, 116, pp. S107-S118.
[11] Story M, Congalton RG (1986). “Accuracy Assessment”: A user’s perspective. Protogrammetric Engineering and Remote Sensing; 52(3): 397-99.
[12] Richards, J., and Jia, X. (2006) Remote Sensing Digital Image Analysis: An Introduction. Springer, Berlin.