Counting Vehicle with High-Precision in Brazilian Roads Using YOLOv3 and Deep SORT

The Brazilian National Department of Transport Infrastructure (DNIT) maintains the National Traffic Counting Plan (PNCT). The main goal of PNCT is to evaluate the current flow of traffic on federal highways aiming to define public policies. However, DNIT still performs the quantitative classificatory surveys not automated or with invasive equipment. It is crucial for conducting traffic studies to search for more modern solutions to accomplish a higher number of automated non-invasive, and low-cost classificatory surveys. This paper proposes a system that uses YOLOv3 for object detection and the Deep SORT for multiple objects tracking algorithms. From the results over real-world videos collected in Brazilian roads, we obtained a precision above 90 % in the global vehicle count. We also show that our proposal outperformed other previously proposed tools with 99.15% precision in public datasets. We believe this paper’s proposal allows the development of a traffic analysis tool to be used for the automation of the volumetric traffic surveys, enabling to improve the DNIT agility and generating economy for the public coffers.

[1]  Antonio Fernández-Caballero,et al.  Vehicle Tracking by Simultaneous Detection and Viewpoint Estimation , 2013, IWINAC.

[2]  Mohamed A. Abdelwahab.  Accurate Vehicle Counting Approach Based on Deep Neural Networks , 2019, 2019 International Conference on Innovative Trends in Computer Engineering (ITCE).

[3]  Fatih Murat Porikli,et al.  CDnet 2014: An Expanded Change Detection Benchmark Dataset , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops.

[4]  Shiru Qu,et al.  Real-time vehicle detection and counting in complex traffic scenes using background subtraction model with low-rank decomposition , 2017 .

[5]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[6]  Ross B. Girshick,et al.  Focal Loss for Dense Object Detection , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[7]  Kaiming He,et al.  Feature Pyramid Networks for Object Detection , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[8]  Kazutoshi Fujikawa,et al.  Passenger Counter Based on Random Forest Regressor Using Drive Recorder and Sensors in Buses , 2019, 2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops).

[9]  Fabio Tozeto Ramos,et al.  Simple online and realtime tracking , 2016, 2016 IEEE International Conference on Image Processing (ICIP).

[10]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[11]  Abhinav Dadhich.  Practical Computer Vision: Extract insightful information from images using TensorFlow, Keras, and OpenCV , 2018 .

[12]  Mohan M. Trivedi,et al.  Looking at Vehicles on the Road: A Survey of Vision-Based Vehicle Detection, Tracking, and Behavior Analysis , 2013, IEEE Transactions on Intelligent Transportation Systems.

[13]  A. A. dos Santos,et al.  Detecção automática de uvas e folhas em viticultura com uma rede neural YOLOv2. , 2018 .

[14]  Ali Farhadi,et al.  YOLOv3: An Incremental Improvement , 2018, ArXiv.

[15]  Dietrich Paulus,et al.  Simple online and realtime tracking with a deep association metric , 2017, 2017 IEEE International Conference on Image Processing (ICIP).