Obstacle Detection Of Mobile Robot Based On Binocular Stereo Vision

Research on robot obstacle avoidance is one of the focuses in the field of mobile robot technology,and it is a comprehensive subject that includes multiple disciplines.The detection of obstacles is a necessary guarantee for robots to complete safe obstacle avoidance.Obstacles are mainly divided into static obstacles(immobility obstacles)and dynamic obstacles(motion obstacles).In order to obtain a wealth of obstacle information,this article will use binocular stereo vision to detect obstacles on the path of mobile robots.The main work completed by the dissertation consists of the following parts:1.Camera Calibration: A binocular stereo vision experiment platform was set up to analyze the three calibration algorithms in detail.Combined with the research content of this paper,Zhang's calibration method was selected and the calibration results verified.2.Image correction and stereo matching: The Bouguet image correction algorithm was studied in detail,and stereo image correction was completed.The current mainstream matching algorithm is compared and studied in detail,and a semi-global matching algorithm that combines both accuracy and speed is selected based on the characteristics of obstacle detection.3.Three-dimensional information detection of obstacles: This article will propose an obstacle detection method based on disparity maps.The method mainly includes the following steps: the first step is to perform three-dimensional measurement according to the parallax map,the second step is to remove the region outside the traveling path and the interference from the ground surface according to the measurement result,and the third step uses the specified detection distance to perform binarization and segmentation.The obstacle detection task of mobile robot is accomplished by this method.4.Position prediction of moving obstacles: In addition to the need to know the three-dimensional information of the obstacles,the mobile robot needs to know the movement of the obstacles during the obstacle avoidance process.Static obstacles can start obstacle avoidance directly,but if they are obstacles in motion,they must first predict the position of the obstacle at the next moment to avoid obstacles.This paper proposes to use the historical position information ofobstacles to establish an autoregressive model to predict the position of the next moment of obstacles.The experimental results show that this prediction method is reliable in a certain time range.