An improved change detection approach using tri-temporal logic-verified change vector analysis

摘要: Abstract Change vector analysis (CVA) is an effective and widely used unsupervised change detection algorithm in remote sensing. It separates changed pixels from unchanged by binarizing bi-temporal difference image. However, the results performance are affected image acquisitions at different dates threshold decision rules for magnitudes, resulting serious false missed detections. This paper proposed a novel tri-temporal logic-verified (TLCVA) approach which can identify errors of CVA through logical reasoning judgement with additional temporal assistance. not only achieve reliable modification to original results, but also produce two improved circulation land surface automatically. The method consists three parts: traditional detection, automated sample selection, refined based on SVM posterior probability comparison space. was experimented cover Sentinel-2 Planet Labs images study areas located Ma’anshan, Nanjing Taizhou City. show that accuracies have significant improvements TLCVA approach, omission commission reduce obviously. generalization, sensitivity efficiency were analyzed experiments. concluded methods preliminary work effectively efficiently, small size training samples selected enough performance.