Global-scale explainable AI assessment for OBIA-based classificationusing Deep Learning and Machine Learning methods

Abstract

Over the past decade, object-based image analysis (OBIA) has gained prominenceas a widely adopted method for generating land use/land cover (LULC) maps. This study aimsto evaluate the performance of various classification algorithms within the OBIA frameworkusing SPOT-6 satellite imagery. The research methodology involved segmenting the imageswith the multi-resolution segmentation (MRS) algorithm, followed by the application ofconvolutional neural networks (CNN), random forest (RF), and support vector machine(SVM) algorithms for classification. The study was conducted in the Perpignan province,located in the Pyr & eacute;n & eacute;es-Orientales region of France. After the segmentation stage, CNN,RF, and SVM classifiers were employed to classify the image segments based on bothspectral and spatial attributes. The accuracy of the resulting thematic maps was assessedusing standard metrics, including overall accuracy (OA), the Kappa coefficient (KC), and theF-score(FS). Of the three classifiers, CNN achieved the highest overall accuracy at 91.28%,outperforming SVM, which attained an OA of 90.50%, and RF, which recorded an OA of87.28%. Additionally, this study explored the integration of explainable artificial intelligence(AI) techniques, specifically the Shapley Additive Explanations (SHAP) algorithm, toenhance the interpretability of the machine learning models. This approach fosters greatertrust, accountability, and acceptance in decision-making processes. By leveraging SHAPvalues, the study provides deeper insights into the decision-making processes of the CNN,SVM, and RF classifiers, ultimately enhancing the transparency and comprehensibility ofthese models.