Small Target Detection Method in Remote Sensing Images Based on Atrous Convolution and Spatial Attention
Aiming at the problems of low accuracy and poor robustness in the detection of small targets in remote sensing images,a deep learning detection model was designed. The backbone network of the model is based on CSPDarknet53,and the convolution ker-nel in the original network is replaced with an atrous convolution kernel to increase the receptive field. At the same time,the spatial attention mechanism is introduced to improve the learning weight of the model for small-sized positive sample features;in order to im-prove the richness of information in the feature map delivered to the detection end,a multi-scale aggregation network is used to aggre-gate feature information under multiple receptive fields,and then combined with upper and lower size sampling and feature map mosa-icking of the same size to output feature maps of three sizes to participate in detection. The experimental results show that the detection accuracy of the model proposed in this paper on the test dataset is significantly higher than that of similar comparison models,and it has good robustness in the face of a variety of complex scenes,and can achieve real-time detection of targets in the test environment.
万方数据
维普
