改进YOLOv11s的距离选通图像人脸检测算法OA

Laser range-gated imaging technology enables long-range object detection,partial penetration imaging,and reliable operation in adverse weather conditions such as rain,snow,and fog.To address challenges including local feature loss and noise interference in complex environments like through windows,occlusions,smoke,or flames,this study devel-ops an enhanced face detection algorithm based on YOLOv11s,optimized for range-gated imaging characteristics.First,the CSDSM module replaces C2PSA to improve fine-grained feature preservation while maintaining training efficiency.Then,the MultiSEAM(multi-scale separation and enhancement attention module)enhances the neck network's ability to handle occlusions and understand contextual features.Furthermore,the SPD-Conv module strengthens low-resolution fea-ture extraction in the backbone network.Finally,an improved EMF module expands local perception and semantic repre-sentation for small objects.Experimental results demonstrate performance gains of 3.3 percentage points in mAP@0.5 and 1.7 percentage points in mAP@0.5:0.95 on range-gated image datasets,confirming the method's effectiveness.To fur-ther verify the generalization and universality of the improved algorithm,the VOC2007 dataset is also selected for testing.The results show that the mAP@0.5 and mAP@0.5:0.95 metrics are improved by 1.8 percentage points and 2.2 percent-age points,respectively.