Research On Radar-vision Integrated Controller For Intelligent Vehicle

With the rapid development of intelligent vehicle automatic driving technology,the requirements for environmental awareness technology and design is also rising.In the smart cars,environment-aware devices are clearly indispensable,among which millimeter-wave radars and visual sensors play an important role.And the full use of the electromagnetic information collected by millimeter-wave radar and visual sensors collected by the image information,especially the two information effective integration,is one of important research of the current intelligent vehicle environment awareness technology.Millimeter wave radar sensor is an active detection type sensor,which can transmit electromagnetic waves and receive electromagnetic echo signal.With the corresponding signal processing technology,Millimeter can access to the movement of the target location,speed and other state parameter information in the real-time detection,but relatively the information is not enough.The vision sensor is a camera-based image sensor,which can obtain a clear image of the surrounding environment of the car,providing rich features such as color,edge,texture,size and so on,and has a unique advantage in the field of target inspection and recognition.But because the image information is too rich,real-time in the analysis is not well.Therefore,how to integrate the real-time information provided by the millimeterwave radar and the rich image information provided by the visual sensor to realize the continuous estimation of the moving target state in the surrounding environment of the automobile driving has gradually become a hot and key problem in the intelligent vehicle environment perception technology.In this paper,by analyzing the characteristics of intelligent vehicle driving in urban environment,selecting the tracking filter,designing the data correlation method,selecting the fusion structure for the target tracking,designing the information fusion algorithm for the millimeter-wave radar and the visual sensor,and then realizing the complex environment target tracking.This paper focuses on the radar-vision integrated controller for the relevant research and algorithm design,the main research content includes the following aspects:Firstly,this paper studies the basic theory of target tracking.On the one hand,we study the typical filtering techniques such as Bayesian,Kalman,Extended Kalman and Unsensitive Kalman.On the other hand,the adaptive target tracking method and the maneuvering target tracking method are compared and analyzed.In addition,data correlation technology has been discussed.In this paper,we use extended Kalman filter technology,interactive multi-model target tracking method and global nearest neighbor data association algorithm to provide the theory for radar-vision integrated controller tracking algorithm design in view of the large number of moving targets in the urban environment and the high mobility accordingly.Next,the basic theory of multi-sensor fusion is studied in this paper.Through the data fusion framework of American JDL,understanding the processing flow and main contents of data fusion,comparing and analyzing the typical data fusion method and typical data fusion structure,aiming at the characteristics of radar information and visual information,a sequential fusion method is proposed.Compared with the target tracking using a single sensor,the method can be more accurate target state estimation,so as to lay the premise of the algorithm of radar-visual integrated control software design.Then,this paper has carried on the software design of the radar-vision integrated controller.According to the characteristics of the vehicle in the urban environment,the state equation of the stationary motion model,the uniform linear motion model and the acceleration / steering motion model,as well as the observation equations of the radar and the visual sensor,are established by using the interactive multi-model tracking method.The model of the radar-vision integrated controller including the interactive fusion module,the conditional filter module,the model probability module,the output fusion module and the measurement and forecasting module are determined,and the tracking filter module,the radar quantity The data association module and the visual data association module are used to design the algorithm,and the tracking and integration of the moving target is realized by using radar information and visual information.Finally,this paper has carried on the simulation research to the radar-vision integrated controller software.Based on the MATLAB simulation platform,the relevant simulation model is constructed,and the simulation conditions are determined according to the moving target characteristics in the urban environment.Then,the target of the variable speed maneuvering target,the steering maneuvering target and the multi-target under the clutter environment The simulation and analysis of the radar are carried out to verify the feasibility and effectiveness of the radar-vision integrated controller software algorithm.Through the research of this paper,the radar-visual information fusion and the tracking of the moving target state are realized.The necessary target motion state information is provided for the intelligent positioning,path planning and driving control of the intelligent vehicle,It also has done a basic job for the research of the radar-visual integrated controller and has certain theoretical and reference value for the intelligent vehicle environment perception technology.