Method of Automated Detection of Traffic Violation with a Convolutional Neural Network

Abstract

This article describes the relevance of developing methods and systems for detection photo-video violations of the Rules of the road. The proposed method includes several steps: 1) detecting of the three classes of objects on a video sequence (pedestrian crossing, a motor vehicle and a human on the pedestrian crossing; 2) tracking the trajectories of the vehicle and the human on the pedestrian crossing; 3) comparing the paths of the pedestrian and the vehicle and determining whether there has been a violation of the Rules of the road for a certain period of time. For real-time object detection, we used neural network YOLO V3.

Detection of Traffic Rule Violations Using Machine Learning: An Analytical Review

“…Many works in the field of traffic violations, particularly detecting the pedestrian lane, can be seen. The authors of that work [25] introduced a system that is able to detect traffic violations at pedestrian crossings. To detect pedestrian crossings, the authors also used a segmentation method depending on user marks.…”

“…[28] This algorithm divides digital images into a number of grids that makes the object detection process easier. For automatic traffic violation detection, the neural network YOLOv3 is applied in the paper [25]. YOLOv3 algorithm is faster than other algorithms and the output has three layers that detect the object of different sizes.…”

“…Global Camera Sensors (GCS) and Local Camera Sensors (LCS) are being used in various cases to detect vehicle number plates with the help of video cameras [30]. In the detection of humans and vehicles, the authors of [25] have used video cameras with high frame rates. They have also worked with pedestrian crossings with these cameras.…”

Restoration of the Lost Map Area of the Underlying Image Surface Using the Reconstruction Method

“…To restore the blind zone in the direction of movement of the carrier radar in this work, is use the image recovery method based on the search for similar blocks [5][6][7].…”

“…In this case RMSE = 9.84. (5) Analysis of the results shows that the higher the spatial correlation, the smaller the recovery error. (7) An analysis of the results shows that the higher the spatial correlation, the smaller the reconstruction error.…”