Improve Ultra-low Latency Vehicle Identification and Traffic Precision Based Yolo-v4

Ikhyeon Jo; Jaemin Sa; Yongyu Kim; Baek Nam

doi:10.7470/jkst.2022.40.6.908

All Issue

2022 Vol.40, Issue 6 Preview Page

Article

Improve Ultra-low Latency Vehicle Identification and Traffic Precision Based Yolo-v4 Yolo-v4 기반 초저지연 차량 식별과 통행량 추적 정밀도 개선

31 December 2022. pp. 908-920

PDF XML

Abstract

With the advancement and commercialization of autonomous driving technology, the demand for infrastructure-based cooperative driving technology to expand the recognition range of the vehicle's own sensor is increasing. In particular, research on edge infrastructure technology that can be installed on the roadside to identify and record traffic conditions in real time and provide it to vehicles is actively underway, and research on sensors for information collection is active. According to this trend, research on traffic information collection technology through CNN-based image analysis is also being actively conducted. It is attracting attention that it can be used in infrastructure construction because it can provide information about objects on the road that cannot be recognized. However, most of the existing C-ITS projects are processing video analysis on high-spec servers at the center rather than on-site due to the high amount of computation required for deep learning algorithms. There is a risk. Therefore, in order to minimize this risk, it is necessary to minimize communication delay through real-time image analysis at the roadside rather than the center. In this study, we designed a vehicle detection system that is located on the roadside and analyzes images in real time to calculate object information located on the road. To apply the algorithm to the roadside environment, we present a method to reduce the amount of deep learning computation by simplifying the configuration of the neural network.

Keywords

cooperative driving infra sensor

deep learning

object detection

object tracking

self driving

자율주행 기술의 고도화 및 상용화에 따라 차량 자체 센서의 인지 범위의 확장을 위한 인프라 기반 협력주행기술에 대한 수요가 증가하고 있다. 특히 노변에 설치되어 교통 상황을 실시간으로 파악하고 기록하여 차량에 제공할 수 있는 엣지 인프라 기술 연구가 활발하게 진행되고 있으며, 정보 수집을 위한 센서에 대한 연구가 활발한 상황이다. 이러한 흐름에 따라, 최근 CNN 기반의 영상분석을 통한 교통정보 수집 기술에 대한 연구도 활발히 진행되고 있는데, 이러한 CNN 기반의 알고리즘은 영상 내의 외양 특징을 기반으로 개별 객체를 식별할 수 있어 자율주행 자동차에서 인식하지 못하는 도로상의 객체 정보를 제공할 수 있어 인프라 구축에 활용이 가능할 것으로 주목받고 있다. 하지만 기존 C-ITS 사업 대부분은 딥러닝 알고리즘의 높은 요구 연산량으로 인하여 영상분석을 현장이 아닌 센터의 높은 사양 서버에서 처리하고 있는데, 이는 인프라 시스템에서의 정보 제공 지연에 따른 사고의 발생 위험을 증대시킬 위험이 있다. 따라서 이러한 위험을 최소화하기 위해서는 센터가 아닌 노변에서의 실시간 영상분석을 통해 통신 지연을 최소화할 필요가 있다. 본 연구에서는 노변에 위치하여 실시간으로 영상을 분석하여 도로에 위치한 객체 정보를 산출할 수 있는 차량검지 시스템을 설계하였으며, 알고리즘의 노변 환경 적용을 위해 신경망의 구성을 간소화하여 딥러닝 연산량을 줄일 수 있는 방법을 제시한다.

키워드

협력주행 인프라 센서

딥러닝

객체 검출

객체 추적

자율주행

References

Alex Bewley, Zongyuan Ge, Lionel Ott, Fabio Ramos, Ben Upcroft (2016), Simple Online and Realtime Tracking, 2016 IEEE International Conference on Image Processing (ICIP), 10.1109/ICIP.2016.7533003, IEEE. 10.1109/ICIP.2016.7533003

Bewley A., Ge Z., Ott L., Ramos F., Upcroft B. (2016), Simple Online and Realtime Tracking, In 2016 IEEE international conference on image processing (ICIP), 3464-3468, IEEE. 10.1109/ICIP.2016.7533003

Bochkovskiy A., Wang C. Y., Liao H. Y. M. (2020), Yolov4: Optimal Speed and Accuracy of Object Detection, arXiv preprint arXiv:2004.10934.

Choi J. R. (2021), Derivation of Autonomous Driving Safety Requirements Based on Hazard Analysis, The 85th Conference of KST, Korean Society of Transportation, 865-878.

DGT H. S. P., Udesang K. J. (2014), A Survey on Object Detection and Tracking Methods, International Journal of Innovative Research in Computer and Communication Engineering, 2(2).

Eun J. H., Oh Y. T., Yoon I. S., Lee C. l., Kim N. S., Han W. G. (2011), Development of The Left-Turn Actuated Traffic Signal Control Strategy using Image Detectors, J. Korean Soc. Transp., 28(2), Korean Society of Transportation, 111-121.

Hwang B. J., Ahn H. J., So J. H., Kim T. H. (2019), Plan of Local Dynamic Map (LDM) Information System for Cooperative Automated Driving, The 81st Conference of KST, Korean Society of Transportation, 512-513.

Held D., Thrun S., Savarese S. (2016), Learning to Track at 100 fps with Deep Regression Networks, In European Conference on Computer Vision, 749-765, Springer, Cham. 10.1007/978-3-319-46448-0_45

Hou X., Wang Y., Chau L. P. (2019), Vehicle Tracking Using Deep Sort with Low Confidence Track Filtering, In 2019 16th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), 1-6, IEEE. 10.1109/AVSS.2019.8909903

Jin H. Y., Kim S. H., Kim T. H. (2022), A Study on Safety of Urban Road Traffic Flow of Autonomous Vehicles in Adverse Conditions, J. Korean Soc. Transp., 40(2), Korean Society of Transportation, 161-177. 10.7470/jkst.2022.40.2.161

Kim H. J., Baek S. C., Yun D. G., Park J. J. (2021), A Study on the Scenario Development for the Analysis of Mixed Traffic Flow Characteristics Following the Introduction of Freeway Exclusive Lanes for Autonomous Vehicles, J. Korean Soc. Transp., 39(6), Korean Society of Transportation, 838-848. 10.7470/jkst.2021.39.6.838

Maaloul S., Aniss H., Kassab M., Berbineau M. (2021), Classification of C-its Services in Vehicular Environments, IEEE Access, 9, 117868-117879. 10.1109/ACCESS.2021.3105815

Mun C. (2022), Development of Advanced Technology for Road Situation Awareness Based on Infrastructure Sensors, The 86th Conference of KST, Korean Society of Transportation, 547-551.

Park J. W., Baek J. W., Lee S. W., Seo W. C., Seo D. W. (2019), Edge Camera based C-ITS Pedestrian Collision Avoidance Warning System, The Journal of The Korea Institute of Intelligent Transportation Systems, 18(6), 176-190. 10.12815/kits.2019.18.6.176

Roh C. G. (2020), Configuration and Evaluation of Local Dynamic Map Platform for Connected Automated Driving in Urban Roads, J. Korean Soc. Transp., 38(1), Korean Society of Transportation, 42-57. 10.7470/jkst.2020.38.1.042

Tan M., Pang R., Le Q. V. (2020), Efficientdet: Scalable and Efficient Object Detection, In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10781-10790. 10.1109/CVPR42600.2020.01079

Wang C. Y., Alexey B., Hong Y. M. L. (2021), Scaled-yolov4: Scaling Cross Stage Partial Network, Proceedings of the IEEE/cvf Conference on Computer Vision and Pattern Recognition. 10.1109/CVPR46437.2021.01283

Wojke N., Bewley A., Paulus D. (2017), Simple Online and Realtime Tracking with a Deep Association Metric, In 2017 IEEE International Conference on Image Processing (ICIP), 3645-3649, IEEE. 10.1109/ICIP.2017.8296962

Yoon C. Y. (2020), Vehicle Tracking Algorithm Using YOLOv3 based SORT, Proceeding of Conference on Information and Control Systems, 409-410.

Information

Publisher :Korean Society of Transportation
Publisher(Ko) :대한교통학회
Journal Title :Journal of Korean Society of Transportation
Journal Title(Ko) :대한교통학회지
Volume : 40
No :6
Pages :908-920
Received Date : 2022-11-10
Revised Date : 2022-11-25
Accepted Date : 2022-11-25
DOI :https://doi.org/10.7470/jkst.2022.40.6.908

Journal of Korean Society of Transportation ISSN:1229-1366(Print) 2234-4217(Online) 대한교통학회지

All Issue