Research On Target Recognition Of Intelligent Driving Vehicle Based On Machine Vision

Traffic has always been an integral part of human life,accompanied by all aspects of human life.With the increase in population,travel volume,traffic pressure and traffic accident annually,intelligent transportation is expected to ensure the security of people's daily travel.Therefore,the continuous breakthrough in artificial intelligence technology has accelerated the launch and commercialization of intelligent transportation and vehicle,which is irreversible.It is expected that in the future the realization of freeing hands off the steering wheel and the safety experience of intelligent driving.Combined with a pre-research project,this paper is located in the field of target recognition for intelligent vehicle-assisted driving system.In view of the problem of missed detection and false detection in target recognition of intelligent vehicle-assisted driving system,research including the vehicle object and the pedestrian goal based on the SVM classifier are conducted.Besides,the theory and experiment of the front vehicle's distance based on monocular vision are studied in the paper.The concrete contents and conclusions are as follows:Firstly,the research status,development trend,research result,policy and so on of intelligent vehicle are expounded in the paper.Then,it is discussed on the necessity and extensiveness of machine vision application in target recognition.What's more,the paper summarizes the background and significance of the research.Therefore,the major technology and research content is determined and then the chapter writing arrangement is organized.Secondly,the theory of image target recognition technology is studied.In this paper,two kinds of typical machine vision sensors are analyzed and the CCD camera is used as the research object.The image preprocessing techniques such as image gray scale,image gray scale enhancement,image filtering and image morphology are studied.Meanwhile,Machine learning target recognition technology is studied,and the support vector machine algorithm which is suitable for a large number of sample data sets is determined to be used in this paper.Next,the research on the feature extraction technology of vehicle and pedestrian is carried out.This paper summarizes the concept of feature and its function in the process oftarget recognition.It expatiates the typical image features such as edge feature,Hough feature,Surf characteristic,color feature and texture feature,and focuses on the research and analysis of HOG characteristics.The principle and basic algorithm of SVM support vector machine are studied,and the vehicle target recognition method based on sample HOG feature information and pedestrian target recognition method based on DPM feature information of sample are determined.Afterwards,the vehicle distance measurement based on machine vision is studied.This paper analyzes the characteristics of commonly used ranging sensors,and puts forward the monocular vision distance measurement scheme based on machine learning.At the same time,the calibration of the camera is carried out.The self-calibration algorithm is put forward on the basis of the principle of Zhang 's calibration.The monocular vision distance measurement equation is deduced on the basis of the image centroid pixel coordinates and the camera internal and external parameters,and the vehicle distance measurement model based on machine vision is constructed.Finally,combined with the previous theoretical research on the subject,the experimental verification study was conducted.On the one hand,the vehicle target identification was conducted into test validation study.The validity and accuracy of the SVM classifier based on the multi-core kernel function optimization are verified by comparing with the real road environment target test.On the other hand,the pedestrian target recognition is carried out,the validity and accuracy of the DMP classifier based on the improved HOG feature are verified by comparing with the real road environment target test.In addition,the vehicle distance measurement based on machine vision is tested and tested,which verified the validity and accuracy of the monocular vision distance measurement model.This paper has made a useful attempt to study the key technologies of intelligent vehicles in intelligent vehicles.The research results have certain practical significance for improving the identification of road moving targets and improving the identification of missing and false alarms.For intelligent vehicle environment perception technology,the application and popularization of visual technology also have some theoretical and reference value.