基于改进Yolov4的车辆目标检测

doi:10.19678/j.issn.1000-3428.0062943

摘要/Abstract

摘要： 交通场景中小目标及遮挡目标的检测对智能交通具有重要意义。目前基于深度学习的方法在检测车辆目标的特征提取方面取得了较好的效果，但是这些方法都缺乏鲁棒性，在交通场景中对小目标及遮挡目标的检测存在漏检、错检等情况。提出一种改进Yolov4的车辆目标检测算法，在主干网络的残差模块中嵌入通道注意力机制ECA-Net，通过对每个通道的重要程度赋予不同的权重来获取跨通道的交互信息，实现通道间的信息关联，加强特征聚合，抑制无效特征。将主干网络输出的浅层特征细节信息与深层特征语义信息相融合，使每层具有当前层语义信息的同时融合丰富的细节信息，增强小目标及遮挡目标的特征。在此基础上，利用深度可分离卷积替换网络特征融合模块的普通卷积，提高网络速度，降低网络计算量。实验结果表明，改进后的Yolov4算法在KITTI和UA-DETRAC数据集上比原Yolov4算法分别提高了1.30和2.16个百分点，检测速度达到55帧/s，相比其他主流模型，能更好地检测小目标及遮挡目标。

关键词: 车辆检测, 小目标, 遮挡目标, 特征融合, 深度可分离卷积

Abstract: The detection of small targets and occlusion objects in traffic scenes is crucial for intelligent transportation. Recently, deep learning-based methods have outperformed other feature extraction techniques for detecting vehicle targets.However, deep learning-based methods have low accuracy in detecting small targets and occluded objects, which lead to a high rate of missed detection.This study proposes an improved Yolov4 vehicle target detection framework to solve these challenges.First, the ECA-Net channel attention mechanism is introduced to the residual module in the backbone.This mechanism assigns different weights according to the importance of each channel to obtain interactive information among channels.In this way, the information association among channels can be realized, which strengthens feature aggregation and suppresses invalid features.Second, the feature information of the shallow and deep layers is merged layer by layer.Each layer has the semantic information of the current layer and incorporates detailed information.This structure is used to enhance the feature information of small targets and occluded targets.Finally, instead of using traditional convolutional operation, the deep separable convolution is dedicated to reducing the training time and computation.The detection results show that the proposed method achieves 1.3 and 2.16 percentage points higher than the baseline Yolov4 on KITTI and UA-DETRAC datasets, respectively, and the detection speed reaches up to 55 frame/s.The proposed method achieves a higher performance than other state-of-the-art techniques for small targets and occluded objects.

Key words: vehicle detection, small targets, occlusion targets, feature fusion, depth separable convolution

中图分类号:

TP391

李松江, 耿兰兰, 王鹏. 基于改进Yolov4的车辆目标检测[J]. 计算机工程, 2023, 49(4): 272-280.

LI Songjiang, GENG Lanlan, WANG Peng. Vehicle Target Detection Based on Improved Yolov4[J]. Computer Engineering, 2023, 49(4): 272-280.

http://www.ecice06.com/CN/Y2023/V49/I4/272

图/表 15

20230417190331

20230417190334

20230417190337

20230417190340

20230417190343

20230417190346

20230417190349

20230417190352

20230417190355

20230417190359

20230417190402

20230417190405

20230417190408

20230417190412

20230417190416

参考文献

[1] 张小瑞, 陈旋, 孙伟, 等.基于深度学习的车辆再识别研究进展[J].计算机工程, 2020, 46(11):1-11. ZHANG X R, CHEN X, SUN W, et al.Progress of vehicle re-identification research based on deep learning[J].Computer Engineering, 2020, 46(11):1-11.(in Chinese)
[2] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once:unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:779-788.
[3] REDMON J, FARHADI A.YOLO9000:better, faster, stronger[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:6517-6525.
[4] REDMON J, FARHADI A.YOLOv3:an incremental improvement[EB/OL].[2021-09-10].https://arxiv.org/abs/1804.02767.
[5] LIU W, ANGUELOV D, ERHAN D, et al.SSD:single shot MultiBox detectorr[C]//Proceedings of ECCVʼ16.Berlin, Germany:Springer, 2016:21-37.
[6] LIN T Y, GOYAL P, GIRSHICK R, et al.Focal loss for dense object detection[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:2999-3007.
[7] GIRSHICK R.Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2016:1440-1448.
[8] REN S Q, HE K M, GIRSHICK R, et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[9] HE K M, GKIOXARI G, DOLLÁR P, et al.Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:2980-2988.
[10] JU M R, LUO J N, LIU G Q, et al.A real-time small target detection network[J].Signal, Image and Video Processing, 2021, 15(6):1265-1273.
[11] ZHANG J, MENG Y Z, CHEN Z P.A small target detection method based on deep learning with considerate feature and effectively expanded sample size[J].IEEE Access, 2021, 9:96559-96572.
[12] LIAN J, YIN Y, LI L, et al.Small object detection in traffic scenes based on attention feature fusion[J].Sensors, 2021, 21(9):3031.
[13] WANG A T, SUN Y H, KORTYLEWSKI A, et al.Robust object detection under occlusion with context-aware compositional nets[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:12642-12651.
[14] 阳珊, 王建, 胡莉, 等.改进RetinaNet的遮挡目标检测算法研究[J].计算机工程与应用, 2022, 58(11):209-214. YANG S, WANG J, HU L, et al.Research on occluded object detection by improved RetinaNet[J].Computer Engineering and Applications, 2022, 58(11):209-214.(in Chinese)
[15] 候少麒, 梁杰, 殷康宁, 等.基于空洞卷积金字塔的目标检测算法[J].电子科技大学学报, 2021, 50(6):843-851. HOU S Q, LIANG J, YIN K N, et al.Object detection algorithm based on atrous convolutional pyramid[J].Journal of University of Electronic Science and Technology of China, 2021, 50(6):843-851.(in Chinese)
[16] BOCHKOVSKIY A, WANG C Y, LIAO H Y M.YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2021-09-10].https://arxiv.org/abs/2004.10934.
[17] WANG C Y, MARK LIAO H Y, WU Y H, et al.CSPNet:a new backbone that can enhance learning capability of CNN[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops.Washington D.C., USA:IEEE Press, 2020:1571-1580.
[18] HE K M, ZHANG X Y, REN S Q, et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9):1904-1916.
[19] LIU S, QI L, QIN H F, et al.Path aggregation network for instance segmentation[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:8759-8768.
[20] WANG Q L, WU B G, ZHU P F, et al.ECA-Net:efficient channel attention for deep convolutional neural networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:11531-11539.
[21] ZHANG H, ZU K, LU J, et al.EPSANet:an efficient pyramid squeeze attention block on convolutional neural network[EB/OL].[2021-09-10].https://arxiv.org/abs/2105.14447.
[22] TONG K, WU Y, ZHOU F.Recent advances in small object detection based on deep learning:a review[J].Image and Vision Computing, 2020, 97:103910
[23] BENDERSKY E.Depth wise separable convolutions for machine learning[EB/OL].[2021-09-10].https://eli.thegreenplace.net/2018/depthwise-separable-convolutions-for-machine-learning.
[24] CHU J, GUO Z X, LENG L.Object detection based on multi-layer convolution feature fusion and online hard example mining[J].IEEE Access, 2018, 6:19959-19967.
[25] HAN G, SU J, ZHANG C.A method based on multi-convolution layers joint and generative adversarial networks for vehicle detection[J].KSII Transactions on Internet and Information Systems, 2019, 13(4):1795-1811.
[26] CAI Z W, VASCONCELOS N.Cascade R-CNN:delving into high quality object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:6154-6162.
[27] HINTON G, VINYALS O, DEAN J.Distilling the knowledge in a neural network[EB/OL].[2021-09-10].https://arxiv.org/abs/1503.02531.

选择文件类型/文献管理软件名称

选择包含的内容