| 963 | 3 | 505 |
| 下载次数 | 被引频次 | 阅读次数 |
主流道路车辆目标检测算法在复杂环境下对小目标识别精度低,易因遮挡和定位不准确造成漏检、误检。提出了改进版YOLOv5算法。针对道路上的小目标,改进Head检测层结构,添加大尺度目标检测层,提高道路上小目标检测精度。为适应目标的形状和尺度变化多样,在颈部网络引入全维动态卷积(Omni-Dimensional Dynamic Convolution, ODConv),对原卷积模块进行替换,提高特征提取能力。为了充分利用全局信息,在颈部网络引入全局注意力机制(Global Attention Mechanism, GAM),提升特征提取能力。针对定位精度问题,引入MPDIoU损失函数,使预测框与真实框更加符合。实验结果表明,改进的YOLOv5算法在自动驾驶数据集KITTI上平均精度均值(mean Average Precision, m AP)达到88.7%,相较于基准模型提高了2%,每秒帧数(Frames per Second, FPS)提升了12%。改进算法的检测精度更高,检测速度更快,有效改善了复杂道路条件下的目标检测问题。
Abstract：The mainstream road vehicle object detection algorithms have low recognition accuracy for small targets in complex environments, and are prone to missed detections and false detections due to occlusion and inaccurate positioning. An improved version of YOLOv5 algorithm is proposed. For small targets on the road, the Head detection layer structure is improved and a large-scale target detection layer is added to improve the accuracy of small target detection on the road. To adapt to the diverse shape and scale changes of the target, Omni-Dimensional Dynamic Convolution( ODConv) is introduced into the neck network to replace the original convolution module and improve the feature extraction ability. In order to fully utilize global information, a Global Attention Mechanism( GAM) is introduced into the neck network to enhance feature extraction capabilities. To address the issue of positioning accuracy, the MPDIoU loss function is introduced to make the predicted box more consistent with the actual box. The experimental results show that the improved YOLOv5 algorithm achieves a mean Average Precision( m AP) of 88. 7% on the autonomous driving dataset KITTI, an increase of 2% compared to the benchmark model, and a 12% increase in Frames per Second( FPS). The improved algorithm, with higher detection accuracy and faster detection speed, effectively improves the object detection in complex road conditions.
[1] REN S Q, HE K M, GIRSHICK R B, et al. Faster RCNN:Towards Real-time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149.
[2] LIU W, ANGUELOV D, ERHAN D, et al. SSD:Single Shot Multibox Detector[C]∥14th European Conference on Computer Vision. Amsterdam:Springer,2016:21-37.
[3] REDMON J,DIVVALA S,GIRSHICK R,et al. You Only Look Once:Unified, Real-time Object Detection[C]∥2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas:IEEE,2016:779-788.
[4] REDMON J,FARHADI A. YOLOv3:An Incremental Improvement[EB/OL].(2018-04-08)[2024-05-20].https:∥arxiv. org/abs/1804. 02767.
[5] BOCHKOVSKIY A,WANG C Y,LIAO H Y M. YOLO4:Optimal Speed and Accuracy of Object Detection[EB/OL].(2020-04-23)[2024-05-10]. https:∥arxiv. org/abs/2004. 10934.
[6]郁强,王宽,王海.一种多尺度YOLOv3的道路场景目标检测算法[J].江苏大学学报(自然科学版),2021,42(6):628-633.
[7]周晴,谭功全,尹宋麟,等.轻量化网络模型的道路目标检测算法[J].无线电工程,2023,53(3):601-610.
[8] MAHTO P,GARG P,SETH P,et al. Refining YOLOv4for Vehicle Detection[J]. International Journal of Advanced Research in Engineering and Technology,2020,11(5):409-419.
[9] WANG P F,FU S,CAO X J. Improved Lightweight Target Detection Algorithm for Complex Roads with YOLOv5[C]∥2022 International Conference on Machine Learning and Intelligent Systems Engineering(MLISE). Guangzhou:IEEE,2022:275-283.
[10]周晴,谭功全,尹宋麟,等.改进YOLOv5s的道路目标检测算法[J].液晶与显示,2023,38(5):680-690.
[11]杜娟,崔少华,晋美娟,等.改进YOLOv7的复杂道路场景目标检测算法[J].计算机工程与应用,2024,60(1):96-103.
[12]马荣贵,张翼,董世浩.基于无人机影像的改进YOLOv5道路目标检测[J/OL].无线电工程, 1-9[2024-05-20]. http:∥kns. cnki. net/kcms/detail/13. 1097. tn. 20230905. 1929. 014. html.
[13]魏陈浩,杨睿,刘振丙,等.具有双层路由注意力的YOLOv8道路场景目标检测方法[J].图学学报,2023,44(6):1104-1111.
[14] LI C,ZHOU A J,YAO A B. Omni-dimensional Dynamic Convolution[EB/OL].(2022-09-16)[2024-05-10].https:∥arxiv. org/abs/2209. 07947.
[15] LIU Y C,SHAO Z R,HOFFMANN N. Global Attention Mechanism:Retain Information to Enhance Channel-spatial Interactions[EB/OL].(2021-12-10)[2024-05-11]. https:∥arxiv. org/abs/2112. 05561.
[16] WOO S,PARK J,LEE J Y,et al. CBAM:Convolutional Block Attention Module[C]∥15th European Conference on Computer Vision(ECCV). Munich:Springer,2018:3-19.
[17] ZHENG Z H,WANG P,REN D W,et al. Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation[J]. IEEE Transactions on Cybernetics,2021,52(8):8574-8586.
[18]麻斯亮,许勇.最小点距离的边界框回归损失函数及其应用[J/OL].小型微型计算机系统,1-8[2024-03-13]. http:∥kns. cnki. net/kcms/detail/21. 1106. TP.20231103. 1816. 008. html.
[19] VARGHESE R,SAMBATH M. YOLOv8:A Novel Object Detection Algorithm with Enhanced Performance and Robustness[C]∥2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems(ADICS). Chennai:IEEE,2024:1-6.
基本信息:
中图分类号:U463.6;TP183;TP391.41
引用信息:
[1]李康,宋文广.改进YOLOv5的道路车辆目标检测方法[J].无线电工程,2025,55(03):493-499.
基金信息:
国家科技重大专项(2021DJ1006)~~
2024-09-26
2024-09-26
2024-09-26