Deep learning-based red light violation detection method

Abstract

The invention discloses a deep learning-based red light violation detection method, which relates to the field of traffic violation detection. The positions of lanes and traffic lights at a traffic light intersection are determined in advance, and the traffic lights are matched with the corresponding lanes. The condition at the traffic light intersection is photographed and recorded through a camera, a photo set transmitted by the camera later is processed centrally, and whether a vehicle has a red light violation behavior is judged based on the positions of the same vehicle under the same traffic light on different photo frames. Besides, as for an off-line vehicle that is unrecognizable due to a long distance, whether to be a known vehicle is recognized in a vehicle similarity comparisonmethod. Various algorithms and models are combined, and the novel traffic violation detection method meets the requirements of violation detection accuracy and flexibility by urban road traffic management.

Description

technical field [0001] The invention relates to the field of traffic violation detection, in particular to a method for detecting violations of traffic violations at traffic light intersections based on deep learning convolutional neural network image recognition and image similarity comparison. Background technique [0002] With the development of society and economy, cars are playing an increasingly important role in our lives. While facilitating people's daily life, it brings a series of problems such as traffic violations and congestion. Most of them are caused by motor vehicle drivers breaking the rules or running red lights. [0003] At present, in the traffic management system of cities in my country, there are mainly two ways to capture illegal vehicles: one is to bury induction coils underground, and to set up digital cameras on crossbars to capture images of running red lights. This method has several significant disadvantages of high cost, inconvenient installat...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the disadvantages of high cost, low efficiency, inflexibility, and low accuracy of the above two traditional illegal vehicle monitoring methods, and provide a detection method based on deep learning for running a red light to meet the needs of urban road traffic management. The need for accuracy and flexibility in illegal detection

Images