Research On Automobile Headlight Detection Technology Based On Computer Vision

Automobile lights play an important role in road lighting,driving state indication,automobile appearance and other aspects,so the detection of automobile lamps is also an important link in the factory inspection of automobiles.The detection of headlamp is mainly to detect the lamp distribution performance of automobile headlamp and the function state of vehicle lamp.Among them,the detection of lamp distribution performance mainly includes the detection of luminous intensity and the detection of optical axis position,while the detection of lamp function mainly takes the luminous state of the automobile lamp as the main detection index.The existing headlamp detection instruments have shortcomings such as low detection accuracy,low speed and large detection error,while the detection of lamp function still adopts the method of manual observation,with low detection efficiency and low accuracy.Aiming at the above problems,this paper combined with computational vision technology to carry out the following research:Based on CCD camera exposure time and luminous intensity image gray scale,object relations,we propose a way to use the gray level and exposure time ratio and luminous intensity calibration algorithm,linear relationship between using Kalman filter to control the average level of the image in a dynamic range of gray.Calculate the ratio of grayscale to camera exposure time,which will be used to calibrate the intensity of high beam.This algorithm can prevent the detection error caused by over-exposure of camera aperture or pixel distortion and improve the robustness of high beam intensity detection.Marr-Hildreth edge detection algorithm is used to carry out edge detection on the optical axis,and clear chiaroscuro cutoff line information can be extracted.By fitting calculation of chiaroscuro cutoff line,the information of the Angle and inflection point of the optical axis can be obtained.Compared with the results of other edge detection algorithms,the detection results of this algorithm are more accurate and stable.The deep learning convolutional neural network model of YOLOv5 algorithm is used to make and train automobile lamp data sets for the main types of automobile headlamps,so as to realize the automatic detection and recognition of the working state of the headlamps.In this paper,the headlamp light distribution detection algorithm based on computer vision technology has formed the corresponding software and hardware equipment,and the detection experiment has been completed in the environment of automobile inspection workshop.The results show that the detection effect is good and the robustness is strong,which has been unanimously recognized by manufacturers and has been put into use.The detection method of lighting function based on YOLOV5 has high identification accuracy,fast speed and lightweight detection model,which is suitable for the inspection of finished vehicles on the factory line.Based on the continuously enriched data,it has important promotion significance.