Research on Obstacle Avoidance and Image Registration Technology Based on Binocular Vision

To enable robots to navigate and avoid obstacles in uncertain environments, a 3D coordinate system was constructed based on visual information decision-making and camera calibration results. A region-based stereo matching algorithm was used to extract depth information from the binocular stereo images to improve the reliability of the disparity map and to enable the robot to achieve local obstacle avoidance. The algorithm works by first selecting a point that does not belong to any growing region, then using an energy function to determine its disparity and setting it as the root disparity, and calculating the error of adjacent points based on the root disparity to determine the region correlation. This process is repeated until all points in the image have been processed. Building a 3D coordinate system based on visual information decision-making and camera calibration results enables the robot to autonomously navigate and avoid obstacles. It allows the robot to accurately perceive its location and direction in uncertain environments, and adjust its trajectory in real-time based on environmental changes. Simulation analysis shows that using binocular recognition and Bug2 algorithm can improve the reliability of the disparity map and enable the robot to achieve local obstacle avoidance.