Research On Obstacle Avoidance Of Mobile Robot Based On Binocular Stereo Vision

With the rapid development of artificial intelligence and intelligent manufacturing technology,the demand and the functional requirements for mobile robots are gradually improving.However,due to various factors,the obstacle avoidance problem of mobile robots is still the main reason that hinders its extensive promotion.Therefore,this paper uses binocular stereo vision technology to study obstacle avoidance of mobile robot.Firstly,this paper elaborates the development and application of mobile robot technology in detail,and focuses on the research progress of binocular stereo vision technology,obstacle detection and recognition,mobile robot obstacle avoidance.At the same time,the implementation scheme of autonomous driving is also studied.On this basis,the technical scheme of this paper is established,and the scheme is decomposed by using the idea of system engineering,which is studied and realized step by step.Secondly,this paper builds the experimental development platform needed by the research,and will use the development platform to complete a series of research work.The binocular stereo vision coordinate system is established and transformed between each other.For the problem of binocular lens distortion,a nonlinear distortion imaging model is established.Then the principle of Zhang ' s calibration method is deduced.The fixed binocular camera is used to shoot the self-made checkerboard for binocular calibration,and the error is optimized to obtain accurate calibration parameters.The images captured by the binocular camera contain abundant information.In order to make the images better used,it is necessary to process the images.Therefore,this paper studies a variety of image processing techniques,including gray conversion,gray value equalization,filtering denoising,image sharpening and image segmentation techniques.The binocular image is used for processing experiments,which provides reference value for improving image features,accelerating processing efficiency and studying obstacle segmentation and recognition.After that,the stereo rectification of the binocular image is completed,and binocular images with epipolar constraint are obtained,which is ready for stereo matching.Thirdly,for the random roadblock driving scene set in this paper,BM algorithm and SGBM algorithm are selected according to the components of stereo matching,and the improvement principle of SGBM algorithm is studied.Based on Open CV,the stereo matching of the two algorithms is realized,the depth disparity map is obtained,and the target ranging is completed.Then,the SAD matching window experiment is carried out,and the best window in the applicable scene is selected according to the matching effect of different windows.In order to verify the reliability of ranging,this paper conducts multiple sets of ranging experiments under different scenarios,and proves the effectiveness of ranging.Finally,in this paper,the binocular ranging system is used to collect the obstacle information of three groups of randomly set roadblock driving environment,and the twodimensional map in the robot coordinate system is established according to the depth and width of the obstacle.The obstacle avoidance test is completed by using path planning and obstacle avoidance strategy.The feasibility of the scheme is verified by three groups of experiments and simulations.