Night High Beam Detection Based On Multi-Object Tracking And Evidence Chain Forensics

The behavior of using high beam on urban roads at night leads to huge traffic safety hazards.Intelligent monitoring is an important means to manage vehicles using high beam illegally.The main content of this paper is to use machine vision technology to fuse multitarget tracking and evidence chain forensics for night high beam detection.Through the detection and tracking of night vehicles,the vehicle coordinates are obtained,and the value of light halo range in each frame image is associated to establish vehicle information evidence chain to support the determination of high beam.The main contents of this paper are as follows:(1)This paper describes the research background and significance of this topic,analyzes the research status of target detection at home and abroad,and analyzes and compares the target detection network based on deep learning.At the same time,it analyzes and compares the performance of detectors in several deep learning areas,selects the vehicle detector suitable for the project at night,and verifies the performance of the detector through the experiment of the vehicle data set at night;(2)This paper analyzes the problems of detection frame position offset and insufficient width when the detector detects vehicles at night,and puts forward a lamp detection scheme based on traditional image processing.The lamp is detected and matched,and the lamp detection frame is generated.Through the traditional image processing,the lamp position coordinates and lamp detection frame width are obtained,and the center point coordinates and detection frame width generated by the detector based on deep learning are corrected.At the same time,the region of interest is divided in the image,and only the vehicles in the region are detected to improve the detection accuracy of the deep learning target detector;(3)In order to support the determination of vehicle high beam behavior,this paper establishes a vehicle exclusive evidence chain,which corresponds to the continuous frames of the vehicle image,stores the position coordinates of each frame of the vehicle,the value of the vehicle light halo range and the vehicle close-up image in the vehicle detection frame,provides data information that can be read directly for the determination calculation model of high beam,and provides image evidence for the determination of high beam;(4)In order to judge the behavior of high beam,this paper puts forward the index of halo range and the index of halo change which can be used to judge the high beam and its calculation method,and further puts forward the light intensity index.The data fusion of the index of halo range and the index of halo change is carried out to increase the data difference,so that the data characteristics of high beam and low beam are more different,which is convenient for the judgment of high beam To improve the accuracy of high beam judgment.