摘要
针对当前交通目标检测方法计算量大、检测精度低的问题,提出一种基于改进YOLOv5的行人车辆检测方法。使用CECA模块替换YOLOv5中的C3模块,并在Backbone结构中添加CBAM注意力机制,使得网络能够提高关键特征信息的权重,降低其他信息的关注度,使用双线性插值进行上采样,优化模型对于较小目标和被遮挡区域的检测性能,以提高检测结果的准确度。实验结果表明,YOLOv5-EC网络的平均检测精度比原始网络提升了1.12%,达到了92.63%,与同类型算法相比都有明显提高。实验结果证明了算法在小目标和遮挡情况下检测时更加有效、准确,因此更适合于实际交通场景中的目标检测任务。
Aiming at the problems of high computational complexity and low detection accuracy in current traffic target detection methods,a pedestrian and vehicle detection method based on improved YOLOv5 is proposed.The CECA module is used to replace the C3 module in YOLOv5,and the CBAM attention mechanism is added to the Backbone structure so that the network can increase the weight of key feature information,reduce the attention of other information,use bilinear interpolation for up sampling,optimize the detection performance of the model for small targets and occluded areas,and improve the accuracy of detection results.The experimental results show that the average detection accuracy of YOLOv5-EC network is 1.12% higher than that of the original network,reaching 92.63%,which is significantly improved compared with the same type of algorithm.The comparative experimental results show that the algorithm in this paper is more effective and accurate in detecting small targets and occlusion,so it is more suitable for target detection tasks in actual traffic scenes.
作者
于晓
叶健
汤计洁
Yu Xiao;Ye Jian;Tang Jijie(School of Electrical Engineering and Automation,Tianjin University of Technology,Tianjin 300384,China;Police Aviation Force,Tancheng Public Security Bureau,Linyi,Shandong 276100,China)
出处
《黑龙江工业学院学报（综合版）》
2023年第8期80-86,共7页
Journal of Heilongjiang University of Technology(Comprehensive Edition)
基金
国家自然科学基金项目“边带模糊红外图像目标的最优可免域免疫因子网络提取研究”(项目编号:61502340)
天津市自然科学基金项目“基于仿生免疫网络的深度模糊红外图像目标提取算法研究”(项目编号:18JCQNJC01000)。
