ntegration of LIME with segmentation techniques for SVMclassification in Sentinel-2 imagery

Abstract

Recent advancements in remote sensing technology have facilitated the acquisitionof images with higher spatial resolution. In response to this rapid technological evolution,the paradigm of OBIA has emerged as a key approach. An essential component of OBIA isimage segmentation, where the careful selection of an appropriate segmentation algorithmand its parameters significantly influences the quality of the segmentation output. This studyaims to conduct LULC analysis on Sentinel-2 imagery and compare the accuracy of theSVM classifier across different segmentation methods produced by MRS, SLIC, Mean Shift,and Quick Shift algorithms. The selected study area is located in the Marmara region ofTurkey and characterized by seven major LULC classes. The segmentation was conductedthough four algorithms, with 60 segment features being extracted for each output, consideringspectral, textural, and geometric attributes separately. Following the classification processwith SVM, overall accuracies of 96.14% for the MRS, 91.00% for the SLIC, 89.95% for theMean Shift and 87.95% for the Quick Shift approach were estimated. These results underscorethe superior performance of the MRS algorithm with significant level of improvement. Thishigh level of accuracy holds significant potential for delivering more dependable and preciseoutcomes in planning and decision-making processes. Moreover, integrating XAI, specificallythe LIME algorithm, enhances the transparency and comprehensibility of classificationanalysis within the OBIA framework. Features associated with the NIR and SWIR bandswere found to have predominantly positive effects. This integration contributes to improvedtransparency, enabling more informed and reliable decision-making processes.