基于YOLOv11的水下中华鲟视频智能检测方法

Intelligent Video Detection Method of Underwater Chinese Sturgeon Based on YOLOv11

摘要: 为克服水下中华鲟检测存在的光照不均、水下水体浑暗、目标遮挡影响问题，本文对YOLOv11检测算法，引入全局注意力机制增强模型对关键特征区域的表达能力，在特征融合阶段采用混洗卷积结构以提升多尺度特征融合效果，从而提高中华鲟检测精度。改进的YOLOv11算法用于水下中华鲟视频检测的试验结果表明：改进的YOLOv11模型精度指标mAP@0.5相比YOLOv11模型提高了3.2%，达到了91.9%。因此，改进YOLO11对实际应用场景中华鲟视频检测具有更高的精度与更强的鲁棒性。

Abstract: In order to overcome the problems of uneven illumination, dark underwater water and target occlusion in underwater Chinese sturgeon video image detection, the global attention mechanism was introduced to YOLOv11 detection algorithm to enhance the expression ability of the model for key feature regions, and the group shuffle convolution structure was used in the feature fusion stage to improve the multi-scale feature fusion effect, so as to improve the detection accuracy of Chinese sturgeon. The experimental results of the improved YOLOv11 algorithm for underwater Chinese sturgeon video detection show that the accuracy index of the improved YOLOv11 model, mAP@0.5, compared with YOLOv11 model, increased by 3.2% and reached 91.9%. Therefore, the improved YOLO11 has higher accuracy and stronger robustness for the actual application scene of Chinese sturgeon video detection.