Hexagon-Shaped Screw Recognition and Positioning System Based on Binocular Vision

Traditional artificial live work asks for a high standard with operator's mental state and technical level in high-voltage and strong-magnetic environment, which not only reduces the efficiency of modern production and living but also causes potential threats to the safety of operators. Therefore, developing robot live work has great significance for both academic research and industrial applications. A key problem of the live work for an industrial robot is recognition and positioning of the hexagon-shaped screw in substation with high accuracy. To solve this problem, this article designs a hexagon-shaped screw recognition and positioning system based on binocular vision in transformer substation. A method for stereo matching including 3 steps is proposed: edge noise of the target is reduced by median filter at the beginning; then the target can be obtained through the image segmentation method of adjusting threshold; further moments are used to express the centers of hexagons in 2 images which work as stereo matching points. Finally depth information can be obtained due to visual geometric relationships based on stereo model and camera calibration parameters based on camera model. The experiment results show that the binocular system designed in this article can accurately identify the target with good stability, and the measurement error is controlled in 4%.