Abstract: Aiming to the removal slab burr methods in traditional steel industry cannot meet the needs of automatic production, the method of robot removal slab burr based on binocular vision positioning was proposed. The simulation model of the required workstation was configured under the framework of the robot operating system (ROS). The scale-invariant feature transform algorithm was improved through the features from accelerated segment test corner detection algorithm. The nearest neighbor and next nearest neighbor ratio method was used to obtain the corners which had the highest matching reliability, and then the circular region centered on the corners was used to screen the effective corners required for trajectory planning. According to the effective corner 3D information obtained by binocular vision, the trajectory planning of robot cutting operation was carried out. The results show that the method is convenient and effective, corner locating is accurate, and the measurement error is within a reasonable range. The possibility for guidance of robot removal slab burrs is verified by simulating the process of robot cutting slab burr integrated with binocular vision.