Infrared Ship Target Detection Algorithm PEW_YOLOv8 in Complex Environments

In the infrared ship detection task in complex environments, issues such as high rates of missed and false detections occur due to noise, occlusion, and the indistinct features of small targets. To address these problems, a ship target detection algorithm PEW_YOLOv8 based on YOLOv8 is proposed. Firstly, the FFA-Net algorithm is used to pre-process the images to improve the image contrast and clarity. Secondly, an MPFF-Backbone main network is designed. Through the multi-path fusion technology, features at different scales can complement each other. Meanwhile, combined with the cross-scale information interaction mechanism, the detail expression ability of small targets is enhanced. Then, an enhanced multi-scale attention neck network EMA-Neck is designed. Through the attention mechanism, background noise is suppressed, and the feature information related to the target is enhanced, improving the distinguishability between the target and the background. Finally, a weighted intersection-over-union loss function WCIoU Loss is designed. By introducing a weighting mechanism and a more comprehensive method for evaluating the target bounding box, the model can better handle inter-target overlap, occlusion, and other interferences in complex scenes.Under the same experimental conditions, compared with YOLOv8, the detection accuracy of the FSPEW_YOLOv8 algorithm on the InfiRay infrared ship dataset reaches 92.7%, increasing mAP50 and mAP50:95 by 4.4% and 3.1% respectively.