改进YOLOv5s的交通多目标检测方法

Traffic multi-target detection method of YOLOv5s is improved

为了提高交通目标检测的精度和效率,提出一种改进YOLOv5s的交通场景多目标检测方法,在YOLOv5s的主干网络中引入高效的层聚合网络结构来提高模型学习目标特征的能力,引入了通道注意力和空间注意力结合的卷积注意力模块(BAM)机制,进一步提高网络模型的特征提取能力,通过采用α-IoU作为边界框回归损失函数,提高了边界框回归精度。实验结果表明,改进的目标检测模型相较于YOLOv5s原模型在检测精度上提升了2.4%,模型参数量和模型大小分别降低了20.9%和19.1%。实现了在不同时间段准确且高效的检测交通场景的多种目标,保证了实时检测的应用需求。

In order to improve the accuracy and efficiency of traffic intersection target detection,this paper proposes an improved traffic scene target detection model based on YOLOv5s.An efficient layer aggregation network structure is introduced into the backbone network of YOLOv5s to improve the ability to learn target features.The attention mechanism CBAM of channel attention and spatial attention is introduced to further improve the feature extraction ability of the network model.Theα-IoU is used as the bounding box regression loss function to improve the bounding box regression accuracy.The experimental results show that compared with the original YOLOv5s model,the improved object detection model proposed in this paper has the detection accuracy increased by 2.4%and the model parameter number and model size reduced by 20.9%and 19.1%respectively.It realizes all kinds of targets of accurate and efficient detection of traffic intersection scenes in different time periods,and ensures the application requirements of real-time detection.