摘要
为了解决人工识别车辆耗时耗力的问题,开展了面向道路场景的智能车辆检测算法研究,旨在实现智能化的车辆检测。提出了一种基于深度学习的道路车辆检测模型,通过采用轻量化且易于部署和开发的YOLOv5s模型作为基础模型,同时引入CA、SE和CBAM三个经典的注意力模块来替换YOLOv5的主干网络中的C3模块,使网络模型能够更好地聚焦于车辆区域,提升了车辆检测的准确性。这使得模型在保持高效性和易用性的同时,能够更好地适应复杂道路场景下的车辆检测需求。实验结果显示,将CBAM注意力模块引入网络模型后,在UA-DETRAC数据集上进行车辆检测的平均精度均值可达92.3%,相比其他注意力模块,其表现更为出色。这一研究结果对于实现智能化的车辆检测具有重要意义,有望为道路交通监控、驾驶辅助系统等应用提供更可靠的解决方案。
In order to solve the problem of time-consuming and labor-intensive manual identification of vehicles,the research on intelligent vehicle detection algorithm for road scenes is carried out to achieve intelligent vehicle detection.A road vehicle detection model based on deep learning is proposed.By using the lightweight and easy to deploy and develop YOLOv5s model as the basic model,three classic attention modules,CA,SE and CBAM,are introduced to replace the C3 module in the backbone network of YOLOv5,so that the network model can better focus on the vehicle area and improve the accuracy of vehicle detection.This enables the model to better adapt to vehicle detection requirements in complex road scenarios while maintaining efficiency and ease of use.The experimental results show that after the CBAM attention module is introduced into the network model,the mean average precision(mAP)of vehicle detection on the UA-DETRAC dataset can reach 92.3%,which is better than other attention modules.The results of this research are of great significance for the realization of intelligent vehicle detection,and are expected to provide more reliable solutions for applications such as road traffic monitoring and driving assistance systems.
作者
义凯
YI Kai(Shanxi Electromechanical Design Institute Co.,Ltd.,Taiyuan 030009,China)
出处
《机械工程与自动化》
2024年第2期15-17,共3页
Mechanical Engineering & Automation