A Novel Approach of Fuzzy Dempster–Shafer Theory for Spatial Uncertainty Analysis and Accuracy Assessment of Object-Based Image Classification

Abstract

Accuracy assessment is a fundamental step in remote-sensing image processing. The accuracy assessment techniques aim to compute classification accuracy and characterize errors, and can, thus, be used to refine the classification or estimates derived from the assessment itself. With regard to their technical capabilities, these techniques have been criticized for their inherent uncertainty and inability to evaluate image classification accuracies. To overcome this issue, the main objective of this letter was to introduce a new approach for the accuracy assessment of object-based image analysis (OBIA). To this end, an integrated approach of fuzzy synthetic evaluation and Dempster–Shafer theory (FSE-DST) was adapted and proposed as an effective approach for object-based image classification accuracy assessment. Two experiments were established to examine the capability of the proposed approach. OBIA was applied to develop a land-use land-cover map of Ahar city and the Ousko area. The proposed FSE-DST was applied for a spatially explicit accuracy assessment. Results indicate that FSE-DST can be effectively applied in spatial accuracy assessments for OBIA and for spatial accuracy assessments in remote-sensing-based classifications. The results of this letter are important to the development of OBIA and can serve as the basis for progressive research in remote sensing by supporting future researchers in obtaining more accurate results from OBIA-based classifications and spatially analyzing the reliability of results.