利用多尺度挤压激励网络提取煤矿采场OA

Due to the large scale of mining area targets and the high complexity of scenes,traditional methods often fall short in achieving complete and accurate extractions.To tackle this issue,we presented a coal mining area extraction method from high-resolution remote sensing images based on deep convolutional neural multi-scale squeeze-and-excitation(SE)network.This method takes U-Net semantic segmentation network as its framework and EfficientNet-B3 network as the backbone.By meticulous design and introducing the multi-scale SE structure,it is aimed at more effectively mining multi-level and multi-scale feature information,thus enhancing the accuracy and integrity of coal mining area extraction.To thoroughly analyze the feature extraction effect of introducing SE module,we carried out experiments on three designed networks,one without the multi-scale strategy,one with multi-scale strategy 1,and one with both multi-scale strategy 1 and 2.The results show that the mean intersection over unions of three models are 0.751 9,0.799 8 and 0.810 4,and the verification accuracy are 0.969 2,0.978 5 and 0.989 0,respectively,which illustrates that this model can effectively enhance the integrity and accuracy of coal mining area extraction.