Research On Dynamic Obstacle Avoidance System Based On Binocular Stereo Vision

In the development of intelligent manufacturing,the research of the dynamic obstacle avoidance system for binocular stereo vision has always been the main direction of machine vision,which is widely used in agriculture,medicine,industry,transportation,remote sensing and other fields.Therefore,the research in this paper has a strong Practical significance.This paper studies the path planning of static obstacles in a binocular stereo vision system in an indoor environment.Analyze the problems of the traditional stereo matching algorithm,such as time-consuming feature matching,more false matches,and path planning failure during dynamic obstacle avoidance.In order to improve the matching precision and accuracy of the binocular stereo vision system,the robot can avoid obstacles smoothly.This paper conducts the following research:In terms of camera calibration,considering the influence of tangential distortion on calibration accuracy,an optimized Zhang Zhengyou calibration method was proposed,and the binocular camera was successfully calibrated.The internal and external parameters and distortion coefficient of the left and right cameras were obtained,and the calibration results with good robustness and high accuracy were obtained,which could meet the requirements of this system.In binocular stereo matching,the existing image stereo matching algorithm does not have image scale invariance.Based on the Harris operator,the scale invariance is enhanced by adding image transformation scale parameters,and the new Harris operator is used to replace SIFT The operator performs image corner detection,calculates corner sub-pixel coordinates to improve corner positioning accuracy,and completes image matching after obtaining SIFT feature vectors.The algorithm takes less time,improves calculation accuracy and robustness.The disparity map and depth map obtained by the algorithm of this paper have short time consumption,which shows that the algorithm of this paper can enhance the processing ability of image feature points and can be applied to obstacle recognition and path planning of binocular cameras.In terms of path planning,combined with the characteristics of binocular visual obstacle recognition,the defects existing in the traditional artificial potential field method are improved,and the repulsion function is modified.The problem of unreachable target point.Modify thedirection of repulsive force to control the angle between repulsive force and gravity less than or equal to 90 degrees,which solves the problem of falling into the local minimum.An improved artificial potential field method combined with edge detection is further proposed.When the robot encounters a large obstacle during driving,it will cause edge movement to ensure that the robot reaches the target point smoothly.A cubic B-spline curve is introduced to the robot's driving trajectory in the improved algorithm for path optimization.Simulation and data analysis in the Matlab development environment verify the feasibility of the algorithm in the binocular visual obstacle avoidance system.