一种适用于交通流多目标检测的CDD-YOLOv8n模型OA

In view of the limited device resources and the low efficiency and accuracy of multi-object detection in complex traffic scenarios,this paper proposes an improved multi-object detection model CDD-YOLOv8n.Firstly,the traditional C2f is replaced with the CAC2f module to effectively aggregate features at different levels and improve the model 's detection capability for objects of varying scales.Secondly,the improved attention mechanism DWECA is integrated into the ends of the backbone network and neck network.The efficient channel attention(ECA)is enhanced by adopting the decomposition operation of depthwise separable convolution.By adjusting parameters such as the size of the convolution kernel,the step size,and the dilation rate,the model's ability to extract and fuse features of different scales is enhanced,while the model complexity is reduced.Finally,in the neck network,UpSample is replaced by DySample,which automatically adjusts the sampling set according to the size and position of the object,enhancing the model's detection ability for objects of different sizes.The experiments show that compared with the original YOLOv8n model,the algorithm proposed in this paper improves precision,recall rate,and mAP@0.5 by 4.7%,4.4%,and 4%,respectively,reduces the model size by 0.3 MB,and improves model detection speed by 5 f/s.This model can quickly and accurately detect multiple objects in different scenes while reducing the computational burden and complexity.It can also effectively detect occluded small objects.By combining this model with the DeepSort-ID algorithm,traffic statistics and vehicle tracking can be achieved,improving the efficiency and safety of traffic management.