基于差分融合与改进YOLOv11的运动小目标群检测算法OA

To address the issue of insufficient detection accuracy for small moving target groups—where images occupy minimal area with sparse features,leading to false positives in similar backgrounds—this study proposes a differential fusion and im-proved YOLOv11 detection method.This approach enhances detection accuracy for small moving target groups in complex scenes,providing technical support for relevant applications.Using plastic pellets as detection targets,a dataset is constructed using high-speed camera imagery to design the differential fusion and enhanced YOLOv11 detection scheme.First,motion fea-ture maps of small moving targets are extracted via frame difference and background subtraction methods.Next,RGB tri-channel fusion technology combined the frame difference image,background difference image,and grayscale original image to enhance feature representation.Finally,detection is performed using the enhanced YOLOv11 model.By adding a small object detection layer,embedding the MSCA attention mechanism,and incorporating the DFF module,the model's performance in capturing and identifying small objects is optimized.Experimental results show that the model's mAP50-95 improved from 0.456 before differential fusion to 0.552 after fusion;Adding the MSCA attention mechanism and DFF module to the fusion further elevated mAP50-95 to 0.580,demonstrating significant detection accuracy improvement.This approach combining differential fusion with the enhanced YOLOv11 effectively compensates for the feature deficiency in moving small target groups,substantially re-duces false detection rates,and markedly improves detection accuracy.It holds significant research prospect for the field of mov-ing small target group detection.