Research On Night Vision Assisted Driving System Based On Davinci Technology

With the rapid development of the economy,the amount of car ownership is increasing along with the increasing traffic accidents.Due to low visibility and the driver's proneness to fatigue at night,most traffic accidents occur,so it is very necessary to improve the safety of driving at night.Therefore,it is very necessary to improve the safety of driving at night.Vehicle safety technology is divided into active safety technology and passive safety technology.Active safety is to warn drivers before possible danger,so as to avoid hurting people and objects.This is a hot research direction of the current automobile safety issues.The vehicle night vision auxiliary driving system is an effective active safety system.This paper carries out the research of vehicle night vision aided driving system based on DaVinci technology,and designs a night vision aided driving system based on TMS320DM642.By using this system in the night drive,the driver can see the possible danger of road surface ahead of time,so that take effective measures in time.This system mainly includes vehicle night vision image acquisition,image enhancement,vehicle recognition,pedestrian detection and image display.Main tasks of this paper are listed as following:(1)The advantages of DSP in image processing are expounded,and the digital signal processor TMS320DM642 is selected as the core processor by introducing the integrated development environment: CCS.The hardware of this system is composed of night vision camera,DSP module,power supply module and display module used for image acquisition,processing and display.(2)Because the images collected at night are in low quality and the information loss is serious,so we need to enhance the night image.Considering that the histogram distribution of the night object and the fog image is symmetrical,the night image gray level is reversed first,then the dark channel prior algorithm used in the fog image is improved and applied to the image after the inversion,and the image is enhanced after the image is reversed.(3)Nighttime vehicle identification.On the basis of two methods based on gradient in optical flow method,namely,Horn-Schunck algorithm and Lucas-Kanade,a combination algorithm of Pyramid optical flow method and Vibe is proposed to identify vehicles at night.(4)Night pedestrian detection.On the basis of introducing the three classifiers,it is proposed that the SVM and Adaboost are fused to design the classifier,and the extracted HOG features are discriminate to determine whether they meet the pedestrian characteristics,thus the night pedestrian detection is realized.