语义变化检测技术在河湖岸线监测中的应用研究OA

To support the implementation of the"Clearing Up the Four Disorders"policy and address low efficiency,long cycles,high costs,and limited accuracy in river and lake shoreline monitoring,a comprehensive semantic label change detection dataset,RLShorelineSCD,was constructed.A convolutional attention module was integrated into the Semantic Change Detection(SCD)task to improve a Bi-temporal Semantic Reasoning Network(Bi-SRNet),leading to the proposed Bi-temporal Convolutional Block Attention Module Network(Bi-CBAMNet).The key shoreline section of the Yangtze River in Yangzhou was selected as the study area.Based on historical monitoring data from 2022 to 2023,the improved network was applied to the prediction dataset.Compared with Bi-SRNet,Bi-CBAMNet improved overall accuracy and F1 scores for segmentation and change detection by 3.97%and 1.52%on the RLShorelineSCD dataset,while achieving changes of 0.25%and-0.09%on the SECOND dataset.These results indicated that the network enhances semantic feature extraction across branches and facilitates the fusion of dual-temporal semantic features.The findings demonstrate significant practical value and broad application prospects for river and lake shoreline monitoring.