Unmanned aerial vehicle image target detection method based on ASP-YOLOv4 multi-scale fusion

The invention discloses an unmanned aerial vehicle image target detection method based on ASP-YOLOv4 multi-scale fusion, and the specific operation of building an unmanned aerial vehicle image target detection model based on ASP-YOLOv4 multi-scale fusion comprises the following steps: building a YOLOv4 network structure; then, the improved SENet network is embedded into each CSP residual error block of a trunk feature extraction network CSPDarknet53 in a YOLOv4 network structure; then, feature maps with the resolution ratios of 208 * 208 and 104 * 104 are output behind the first CSP residual block and the second CSP residual block respectively, and the feature maps are fused into a feature pyramid network PANet after being subjected to down-sampling dimension reduction through improved ASPP, so that an unmanned aerial vehicle image target detection model based on ASPP-YOLOv4 multi-scale fusion is obtained; the small target in the unmanned aerial vehicle image can be detected by using the unmanned aerial vehic