Research on improved YOLOv8 remote sensing target detection algorithm based on multi-receptive field feature enhancement

In view of the problem that small targets in remote sensing image are densely arranged and easily blocked, the YOLOv8 model is improved. In the backbone network, a Dilated Convolution module is used to replace the original convolutional module, maintain the resolution and size of the feature map, increase the receptive field of the convolutional kernel, and capture more context information. Channel and spatial multi-dimensional attention mechanism are introduced, correlation between spatial direction and location is mined, and feature extraction ability and long-distance dependency capturing ability are improved. Experiments show that: For remote sensing target data set UC Merced Land Use, data set preprocessing was first carried out to increase the complexity of image processing, and then the average accuracy of model (mAP) was increased by 6.8% after the improvement. The average accuracy rate (AP) of all categories has been improved, and the average accuracy value of most categories has been increased to more than 5%, which alleviates the problems of missed detection and false detection, and enhances the detection effect of the model for remote sensing images with mutual occlusion.