Research On Display Status Monitoring Technology Of Bus Automobile Dashboard Based On Machine Vision

With the fast development of automobile electronic technology,embedded technology and liquid crystal display technology,automobile dashboard can display more and more information,it can provide driver the system condition information of vehicle,such as vehicle speed,engine speed,gear information,warning information,which makes the increasing demand for digital intelligent automobile instrumentation.Bus automobile instrument has the advantages of simple transmission line and fast transmission speed.Therefore,it becomes the most popular bus digital automobile dashboard currently.In order to ensure that the automobile dashboard can reflect the state of the vehicle quickly and accurately,it is necessary to implement a comprehensive test to see if the state information displayed on the dashboard is right under corresponding working conditions before installation.The traditional method of testing is through manual visual inspection.This method has many problems such as low efficiency,poor accuracy and boring operation.Although many scholars have done related research of automation test and have made some development.There are still existing the problem of low test accuracy and poor real-time.For this,in this thesis,display status monitoring system of bus automobile dashboard based on machine vision is designed.A straight line extraction method based on the combination of the minimum distance method and the center point correction is proposed to improve the recognition speed and accuracy of the pointer table;To achieve the identification of the gear icon,Hu invariant matrix is used;To improve the speed and accuracy of the warning icon recognition,the improved SIFT(scale invariant feature transformation)method is used;To achieve the identification of the character,OCR(Optical Character Recognition)optical character recognition technology is used;To achieve the identification of status of light,gray average of the area of indicator light is extracted;To realize the identification of the color information of indicator light,color space model is used.The specific research content of this thesis is as follows:(1)The display content characteristics of automobile dashboard are analyzed.Determine the visual hardware device.Automobile dashboard image acquisition test platform is designed and built.(2)Image pre-processing algorithm for automobile dashboard is studied,including image gray scale,image enhancement,image filtering,and image binarization.The comparative experiment of the algorithm is carried out,and the appropriate pre-processing algorithm is analyzed and selected.(3)The display state characteristics of automobile dashboard are studied.Including the reading of the pointer table information,the icon and the characters of TFT LCD screen information,the status and color of the lights.(4)The method of extracting the pointer line based on the central projection method and combination of the minimum distance method and the center point correction is proposed.And a comparison experiment is done using traditional Hough transform to extract pointer line.Result shows that extracting line method based on combination of the minimum distance method and the center point correction is better than the other two methods.At the mean time,the average recognition error of pointer reading is reduced from 0.98% to 0.49%,and average time of detection is reduced from 254.08 ms to 49.42 ms.(5)To achieve the identification of the gear icon,Hu invariant matrix is used.To achieve the identification of the warning icon,improved SIFT is used,the SIFT feature descriptor is linearly reduced to a lower dimension space,appropriate weight is assigned to descriptor to enhance the distinctiveness between descriptors,and the speed and accuracy of the warning icon recognition is improved,accuracy of the icon recognition result can reach 99%.To achieve the recognition of characters on LCD screen,OCR optical character recognition technology is used and do revise of recognition result,accuracy of recognition result can reach 96%.(6)To achieve the identification of status of indicator light,gray average of the area of the indicator light is extracted.The flashing frequency of light is calculated according to the interval time of light on and off.And color space model is used to identify the color information of the light.The recognition method is simple and the recognition accuracy can reach 99%.(7)Writing software of instrument panel image processing,feature parameter extraction,recognition,result analysis.And human-computer interaction interface is designed.