CCE-YOLO:一种基于改进YOLO11n的水下目标检测算法OA

Underwater target detection holds significant application value in fields like marine resource exploration and ecological monitoring.However,in complex underwater environments,challenges such as frequent target occlusion,severe light attenuation,and insufficient multi-scale target representation are prevalent.To address these challenges,this study proposes an improved underwater target detection model named CCE-YOLO,based on YOLO11n.First,a cascade group attention module(C2CGA)is in-troduced to enhance the original C2PSA by improving the PSA mechanism.This module strengthens cross-level feature interaction capabilities through group attention and cascaded information transmis-sion,thereby improving the model's discriminative performance for occluded and weak-texture targets.Second,a partial multi-scale feature aggregation module(CSP-PMSFA)is designed,utilizing progres-sive channel splitting and multi-scale convolution fusion strategies to enhance feature representation for multi-scale underwater targets.Finally,an efficient upsampling module based on shifted channel mixing(EUCB-SC)is constructed to strengthen spatial-channel interactions and mitigate the parameter redun-dancy and artifact issues associated with traditional upsampling.Experimental results on the DUO dataset show that CCE-YOLO achieves 85.8%in mAP@0.5 and 67.2%in mAP@0.5:0.95,representing improve-ments of 1.4 and 1.3 percentage points over the baseline model,respectively.Meanwhile,its generaliza-tion and robustness are verified on the RUOD dataset,demonstrating effectiveness in complex underwa-ter environments.