Research On Obstacle Recognition Technology Of Modern Tram Based On Radar And Machine Vision

With the rapid development of economy,people's life is getting richer and richer,and the use of automobiles is becoming more and more popular.However,the environmental problems and traffic congestion caused by these are becoming more and more serious.In order to solve the problems of energy crisis,environmental pollution and traffic congestion,modern trams have come into people's vision.With its large transportation volume,beautiful appearance,green energy saving and so on,modern tram has been affirmed by the public and the government.Modern trams are becoming the focus of urban public transport development.The construction of modern trams in China is mostly based on the reconstruction of existing roads.In this mode,trams share the right of way with social vehicles or pedestrians,and the lane is greatly disturbed by other vehicles or pedestrians,which is easy to cause traffic accidents.Therefore,it is necessary to develop a barrier recognition technology to ensure the safe running of trams.In this thesis,multi-sensor fusion based on millimeter-wave radar and machine vision is used to recognize and detect obstacles.Firstly,for the radar part,the radar sensor is mainly used,which is fast and safe.The radar sensor can handle the contradiction between measurement performance and high security,and has good stability.It also can detect the distance of the target in real time,and judge whether there is collision risk and reliability according to the current speed.In this thesis,according to the communication protocol of obstacle detection system,the program control interface of millimeter wave radar data acquisition program is written with VC++.Then,the effective target determination is carried out for the obstacle information in front of the modern tram acquired by millimeter wave radar.In practical operation environment,the front objects detected by millimeter wave radar may come from effective obstacles,or from invalid targets,such as wire poles,fences,etc.Moreover,millimeter wave radar may not find signals in the process of bumping up and down,so it is necessary to screen the information detected by millimeter wave radar effectively.Based on the principle of the closest distance between the same lane,the initial selection of obstacle information such as pedestrians and vehicles acquired by millimeter wave radar is made.In order to judge whether the measured target information is consistent with the predicted target information,the Kalman filtering method is used to predict target obstacles.If they are inconsistent,the life cycle of the effective target is used to make the final decision for the target.Secondly,in the part of machine vision,the target is converted into image signal by using Charge Coupled Device(CCD)camera.After consistency checking of millimeter wave radar obstacle information,different effective target information becomes the region of interest.In this thesis,a large number of vehicle and pedestrian samples are collected to builda picture library.Haar-like is used to extract the relevant features in the picture.Then the concept of "integral graph" is introduced.Then,Adaboost and cascade classifier are combined to generate recognition classifier to recognize the obstacles such as vehicles and pedestrians in front.Thirdly,for the multi-sensor fusion part,the obstacle information in the millimeter-wave radar coordinate system is converted to the three-dimensional space coordinate system,then the model is transformed to the camera coordinate system,and then it is projected to the image coordinate system.Combined with Zhang Zhengyou's calibration principle,the internal and external parameters and distortion parameters of the CCD camera needed to build the fusion model are obtained.The millimeter wave radar data information and machine vision detection information are fused in space.Multithread method is used to integrate millimeter wave radar and machine vision in time.Finally,the software of the model is designed,and the driving experiment of the fusion system in real road environment is carried out.The experimental results show that the method can detect and recognize the obstacles in front efficiently and accurately.