Journal of Jilin University(Engineering and Technology Edition) ›› 2022, Vol. 52 ›› Issue (6): 1353-1361.doi: 10.13229/j.cnki.jdxbgxb20210380

Previous Articles    

Traffic target detection method based on improved convolution neural network

Ming-hua GAO(),Can YANG   

  1. School of Information,East China Jiaotong University,Nanchang 330013,China
  • Received:2021-04-27 Online:2022-06-01 Published:2022-06-02

Abstract:

In view of the problems of low accuracy and high requirements for equipment performance of existing target detection methods in traffic scenes, a target detection method based on improved YOLOv3 network is proposed.A new IOU(Intersection over Union) function is proposed to improve the CIOU(Complete Intersection over Union)function, as CIOU function can not accurately measure the degree of coincidence between the prediction box and the real box under special circumstances. By increasing the detection scale, the feature mAP is refined and the deep and shallow semantic information fusion is enhanced to enhance the small target detection ability; by adding SE(Squeeze-and-Excitation) attention mechanism, the model pays more attention to learning the feature information of important channels; according to the characteristics of the data set, M-YOLT enhancement method was used to improve the performance of small target detection. Through the structural channel pruning of the network, the equipment performance requirements are reduced, and Matrix NMS is used to replace NMS to improve the detection speed. The experimental results show that the proposed method can improve the detection effect and reduce the detection time. The test on KITTI data set reaches 88.4% mAP, or the average detection time of each image can be reduced to 8.6 ms while maintaining 86.3% mAP. This method can achieve better detection effect with lower hardware requirements.

Key words: intelligent transportation, target detection, deep learning, convolutional neural networks(CNN), YOLO

CLC Number: 

  • TP391.4

Fig.1

Improved YOLOv3 network structure"

Fig.2

SE attention module structure"

Fig.3

SPP module structure"

Fig.4

Influence of relative position on CIOU"

Fig.5

When CIOU is constant, the smaller the θ is, the higher the coincidence degree is"

Fig.6

Original picture and resized picture"

Fig.7

M-YOLT cutting and splicing"

Fig.8

Function comparison"

Table 1

mAP of the trained network"

网络PrecisionRecallmAP/%F1mAP提升/%平均每张图片检测时间/ms
Yolov30.8120.76782.10.8010.012.3
4分支YOLOv3 (A)0.8100.78483.20.8111.116.0
A+SPP(B)0.8110.79184.00.8170.816.5
B+SE+CIOU(C)0.8140.81485.10.8211.116.7
C+PIOU(D)0.8160.81485.40.8220.316.7
D+Mish(E)0.8190.83086.60.8311.216.8
E+M-YOLT(F)0.8340.84788.40.8471.816.9
F+剪枝(G)0.8110.82286.30.825-2.112.1
G+Matrix NMS0.8110.82286.30.8260.08.6

Fig.9

Comparison of target detection results"

Fig.10

Comparison of detection results ofCCTSDB dataset"

Fig.11

PIOU simulation results"

1 Ren S Q, He K M, Girshickr 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.
2 邵毅明,屈治华,邓天民,等. 基于加权密集连接卷积网络的快速交通标志检测[J]. 交通运输系统工程与信息, 2020, 20(2): 48-54.
Shao Yi-ming, Qu Zhi-hua, Deng Tian-min, et al. Fast traffic sign detection based on weighted densely connected convolutional network[J]. Journal of Transportation Systems Engineering and Information Technology, 2020, 20(2): 48-54.
3 张富凯, 杨峰, 李策. 基于改进YOLOv3的快速车辆检测方法[J]. 计算机工程与应用, 2019, 55(2): 12-20.
Zhang Fu-kai, Yang Feng, Li Ce. Fast vehicle detection method based on improved YOLOv3[J]. Computer Engineering and Application, 2019, 55(2): 12-20.
4 张江永, 徐智勇, 张建林, 等. 基于敏感度的YOLO网络集成剪枝算法[J]. 计算机工程, 2021, 47(9): 59-68.
Zhang Jiang-yong, Xu Zhi-yong, Zhang Jian-lin, et al. Sensitivity-based Integrated Pruning Algorithm for YOLO Network[J]. Computer Engineering, 2021, 47(9): 59-68.
5 贺超雷, 毕秀丽, 肖斌. 一种基于Zernike矩的局部特征检测方法[J]. 计算机科学, 2020, 47(2):135-142.
He Chao-lei, Bi Xiu-li, Xiao Bin. Zernike moment based approach for local feature detection [J]. Computer Science, 2020,47 (2): 135-142.
6 Zheng Z H, Wang P, Liu W, et al. Distance-IoU loss: faster and better learning for bounding box regression[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 12993-13000.
7 宋欣,李奇,解婉君,等. YOLOv3-ADS:一种基于YOLOv3的深度学习目标检测压缩模型[J]. 东北大学学报: 自然科学版, 2021, 42(5): 609-615.
Song Xin, Li Qi, Xie Wan-jun, et al. YOLOv3-ADS:a compression model for deep learning object detection based on YOLOv3[J]. Journal of Northeastern University(Natural Science), 2021,42(5): 609-615.
8 Ophoff T, Puttemans S, Kalogirou V, et al.Vehicle and vessel detection on satellite imagery: a comparative study on single-shot detectors[J]. Remote Sensing, 2020, 12(7): 1217-1238.
9 李宇杰, 李煊鹏, 张为公. 基于视觉的三维目标检测算法研究综述[J]. 计算机工程与应用, 2020, 56(1): 11-24.
Li Yu-jie, Li Xuan-peng, Zhang Wei-gong. Survey on vision-based 3D object detection methods [J]. Computer Engineering and Applications, 2020, 56(1): 11-24.
10 张飞翔, 余学儒, 何卫锋, 等. 结合改进的损失函数与多重范数的人脸识别[J]. 计算机工程与应用, 2020, 56(24): 144-150.
Zhang Fei-xiang, Yu Xue-ru, He Wei-feng, et al. Face recognition with improved loss function and multiple-norm[J]. Computer Engineering and Applications, 2020, 56(24): 144-150.
[1] Ji-hong OUYANG,Ze-qi GUO,Si-guang LIU. Dual⁃branch hybrid attention decision net for diabetic retinopathy classification [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 648-656.
[2] Lin SONG,Li-ping WANG,Jun WU,Li-wen GUAN,Zhi-gui LIU. Reliability analysis based on cyber⁃physical system and digital twin [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 439-449.
[3] Jie CAO,Jia-lin MA,Dai-lin HUANG,Ping YU. A fault diagnosis method based on multi Markov transition field [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 491-496.
[4] Gui-xia LIU,Zhi-yao PEI,Jia-zhi SONG. Prediction of protein-ATP binding site based on deep learning [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(1): 187-194.
[5] You QU,Wen-hui LI. Single-stage rotated object detection network based on anchor transformation [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(1): 162-173.
[6] Xian-yan KUANG,Hui-chao LUO,Rui ZHONG,Peng OUYANG. Bus arrival time prediction based on wavelet neural network optimized by Beetle Antennae Search [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(1): 110-117.
[7] Jie ZHANG,Wen JING,Fu CHEN. Vulnerability detection of instant messaging network protocol based on passive clustering algorithm [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(6): 2253-2258.
[8] Li-li DONG,Dan YANG,Xiang ZHANG. Large⁃scale semantic text overlapping region retrieval based on deep learning [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(5): 1817-1822.
[9] Li-sheng JIN,Bai-cang GUO,Fang-rong WANG,Jian SHI. Dynamic multiple object detection algorithm for vehicle forward based on improved YOLOv3 [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(4): 1427-1436.
[10] Xue-yun CHEN,Tao XU,Xiao-qiao HUANG. Detection method of medical cell image generation based on conditional generative adversarial network [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(4): 1414-1419.
[11] De-lun PAN,Jun JI,Yue-jin ZHANG. Dynamic object detection method of video surveillance based on motion vector space coding [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(4): 1370-1374.
[12] Feng-chong LAN,Ji-wen LI,Ji-qing CHEN. DG-SLAM algorithm for dynamic scene compound deep learning and parallel computing [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(4): 1437-1446.
[13] Jin-qing LI,Jian ZHOU,Xiao-qiang DI. Learning optical image encryption scheme based on CycleGAN [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(3): 1060-1066.
[14] Zhen SONG,Jun-liang LI,Gui-qiang LIU. Constant flow prediction method of variable speed hydraulic power source based on deep learning and limitation fuzzy [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(3): 1106-1110.
[15] Zhe-ming YUAN,Hong-jie YUAN,Yu-xuan YAN,Qian LI,Shuang-qing LIU,Si-qiao TAN. Automatic recognition and classification of field insects based on lightweight deep learning model [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(3): 1131-1139.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!