一种高分辨率遥感图像视感知目标检测算法

An effective object detection approach with visual perception for high-resolution remote sensing images is proposed to address the problem that the accuracy and speed of existing object detection algorithms of remote sensing images are low, especially in large-scale and high-resolution remote sensing images. Firstly, some sub-regions are selected in the scene using a saliency map, and then transfer computing resources to the area that may contain objects to reduce the computational complexity. Then, pre-selected objects are obtained by a fast learning model YOLO (you only look once). An object semantic association suppress is proposed to screen the pre-selected objects for effective objects. It reduces the interference of false objects for reducing the false alarm probability. Experimental results on NWPU_VHR-10 dataset show that the proposed algorithm is the best, and the mean average precision (mAP) is 86.53%. The results of the proposed algorithm are much better than those of YOLO on LUT_VHRVOC-2 dataset which contains more large-scale and high-resolution remote sensing images. It is concluded that the performance of the high-resolution remote sensing image is improved by the proposed algorithm.