Accuracy Assessment of Land Use/Land Cover Classification Data from Sentinel-2 & ASTER Imagery Interpretation using Unsupervised Classification Algorithm
Keywords:Keywords: remote sensing, land use/land cover classification, unsupervised classification algorithm, accuracy assessment.
Abstract. Accurate land use/land cover information is important for various spatial planning decision making. Remote sensing is an effective mapping technique such as those depicting land cover as it provides a map-like representation of the Earth’s surface that is spatially highly consistent. This study compared the classification accuracies of land cover/land use maps created from Sentinel-2 and ASTER imagery with the Kalimanah Sub-district as a research area. Both images are clustered into 52 spectral clusters using Learning Vector Quantization (LVQ) and K-means unsupervised classification algorithm. Each spectral cluster from each image was assigned into four land use/land cover classes, i.e. urban, agricultural, forest, and barren land. 240 data references were generated from Google Earth imagery as the sample data set is compared with the classification maps that is being assessed. With the kappa analysis approach, error matrices are made based on the same data references for each of the two images to assess the classification quality and to find out the best imagery that yields the most accurate land use/land cover data. Overall accuracy of LVQ algorithm for the Sentinel-2 and ASTER imageries was 78.33% and 69.17%, respectively; while the kappa coefficient of LVQ algorithm for the Sentinel-2 and ASTER imageries were 0.71 and 0.55, respectively. In different circumstances, overall accuracy of K-means algorithm for the Sentinel-2 and ASTER imageries were 81.25% and 72.68%, respectively; while the kappa coefficient of K-means algorithm for both imageries were 0.74 and 0.61, respectively. At the 95% confidence level, for both LVQ and K-means classification algorithms, image classification accuracies of Sentinel-2 dataset are better than the ASTER dataset. Thus, Sentinel-2 imagery provides better accuracy than ASTER imagery in land use/land cover classification data from any unsupervised classification algorithms.
Kaura, S., & Bansal, R. K. (2018). Mixed Pixel Classification by Using Hybridization of Evolutionary Method With Neural Networks. International Journal of Applied Engineering Research, 13(8), 5736-5744. Lillesand, T., Kiefer, R. W., & Chipman, J. (2015). Remote Sensing and Image Interpretation. John Wiley & Sons, Hoboken. Rwanga, S. S., & Ndambuki, J. M. (2017). Accuracy assessment of land use/land cover classification using remote sensing and GIS. International Journal of Geosciences, 8(04), 611. Su, T. C. (2016). A filter-based post-processing technique for improving homogeneity of pixel-wise classification data. European
Journal of Remote Sensing, 49(1), 531-552. Yadahalli, G. S., & Bellakki, M. A. (2018). Accuracy Assessment of Supervised and Unsupervised Classification using Landsat-8 Imagery of D-7 Shahapur Branch Canal of UKP Command Area Karnataka, India. Int. J. Curr. Microbiol. App. Sci, 7(7), 205-216. Zhou, Z., Ma, L., Fu, T., Zhang, G., Yao, M., & Li, M. (2018). Change Detection in Coral Reef Environment Using High-Resolution Images: Comparison of Object-Based and Pixel-Based Paradigms. ISPRS International Journal of Geo-Information, 7(11), 441.