Infrared UAV target detection algorithm based on improved YOLOv8
In order to solve the problems of serious feature loss,low accuracy and complex model in the process of in-frared UAV target recognition,an improved infrared UAV target detection algorithm based on YOLOv8 is proposed.Firstly,deformable convolution is introduced into the backbone network to enhance the feature representation capability of the target region.Secondly,a lightweight feature pyramid network structure SOD-FPN for small targets is proposed for small targets,which avoids the information loss of small targets by reducing the number of network layers and deleting large target detection headers.Moreover,the multi-scale feature fusion capability of the model is enhanced by cross-scale connection and weighted feature fusion method.Finally,NWD Loss based on Wasserstein distance is se-lected as the bounding box loss function to further improve the convergence and detection accuracy of the model.The experimental results show that the mAP50 of the improved algorithm is 99.4％,which is 2.2％higher than YOLOv8n,and the number of parameters is 72.8％lower.Meanwhile,compared with other advanced target detection algorithms,the accuracy and speed of the improved algorithm are improved,which proves the effectiveness and ad-vancement of the improved algorithm.
万方数据
维普
