Object detection in optical remote sensing images based on rotation box representation
Rotated object detection is a key step in the intelligent interpretation of remote sensing images.However,the com-plex background,arbitrary target direction,and large scale differences of remote sensing images pose certain difficulties in achiev-ing accurate rotated object detection.The proposed ERDet combines the explicit visual center to extract global and local informa-tion from remote sensing images,and combines the adaptive threshold sample selection horizontal target detection algorithm and long edge definition method to predict the category,position,and rotation angle of remote sensing image targets.Experiments on the DOTA-v1.0 dataset show that this method can accurately extract targets of different scales and directions,achieving accurate detec-tion of remote sensing targets.
万方数据
维普
