Multitask Learning Architecture For Vehicle Over Speed As Traffic Violations Detection And Automated Safety Violation Fine Ticketing Using Convolution Neural Network And Yolo V4 Techniques

Authors

  • Navatha S Assistant Professor, Department of Computer Application, Nagarjuna Degree College, Bangalore City University, Bangalore – 64, Karnataka
  • Kavitha M Assistant Professor, Department of Computer Application, Nagarjuna Degree College, Bangalore City University, Bangalore – 64, Karnataka
  • K L S D T Keerthi Vardhan Assistant Professor, Department of Computer Science and Engineering (AIML), Siddhartha Institute of Engineering and Technology, Ibhrahimpatnam, Ragareddy Dist., Telangana – 501506
  • P Raghu Assistant Professor, Department of Computer Science and Engineering (AIML), Siddhartha Institute of Engineering and Technology, Ibhrahimpatnam, Ragareddy Dist., Telangana – 501506
  • E. Boopathi Kumar Assistant Professor, Department of Information Technology, Bharathiar University, Coimbatore – 641046, Tamilnadu

Keywords:

Speed Classification, Yolo V5 Algorithm, Automated Fining, Convolution Neural Network

Abstract

The government and insurance companies are increasingly concerned about traffic violations since they put their lives and the lives of others in danger. The Department of Traffic Enforcement has proposed significant preventive measures to stop traffic offences. There are situations where traffic restrictions are broken despite strict enforcement and actual police observation. A deep learning model must be used in conjunction with automated technique to detect safety and traffic violations. Using convolution neural network and yolo v4 methods, a multitask learning architecture for vehicle passenger safety and traffic violations detection and automated violation fine ticketing is proposed in this paper to identify persons on over speeding on capturing the video from surveillance cameras and imposing the automatic traffic and safety violation fine ticketing on identifying vehicle details from processing the license plate of the device. Initially the suggested model works with the video that was taken from the Coco dataset. Image frames are created from extracted footage. Using the Yolov4 approach, image frames are processed for feature extraction and detection. Multiple objects can be detected by the Yolo4 object identification approach in a single frame. Yolov4 creates an object-bounding box.Yolov5 is made up of Yolov4 as the head, PANet as the path aggregation neck, and Resnet34 as the backbone. Resnet34 has 34 CNN layers with 725*725 receptive fields. On the basis of vehicle speed, Backbone is used to extract the feature connected to the license plate. However, the neck is used for feature mapping based on vehicle employment position. Finally, a head is used to measure the individual vehicle's speed. Automated detection of abnormality on the individual vehicle. Proposed model is evaluated with respect to detection accuracy, average precision and frames per second on detecting the speed of the tracked vehicle. Experimental results proves that proposed model outperforms other state of art approaches on accuracy and efficiency on processing road safety related data.

Downloads

Published

2023-10-12

How to Cite

Navatha S, Kavitha M, K L S D T Keerthi Vardhan, P Raghu, & E. Boopathi Kumar. (2023). Multitask Learning Architecture For Vehicle Over Speed As Traffic Violations Detection And Automated Safety Violation Fine Ticketing Using Convolution Neural Network And Yolo V4 Techniques. Chinese Journal of Computational Mechanics, (5), 431–435. Retrieved from http://jslxxb.cn/index.php/jslxxb/article/view/4379

Issue

Section

Articles