Abstract: Controlling the robotic arm to locate the target position is a crucial issue during task execution. This paper proposes a design method for an automatic positioning system for a robotic arm based on binocular stereo vision. The binocular stereo vision system calculates the three-dimensional coordinates of the target object from its two-dimensional image, and then uses these coordinates to control the robotic arm to automatically move to the target position. Experiments show that this system improves the operability of the bomb disposal robot's robotic arm and significantly enhances its performance. Keywords: Binocular stereo vision, Camera calibration, Motion control [b][align=center]The Target Positioning System of a Manipulator Based on Binocular Stereo Vision Wang Wei, Luo Fei, Jiang Liangzhong, Qi Hengnian[/align][/b] Abstract: Target Positioning is very important for a manipulator, which implies controlling the manipulator to move to the position of a target. This paper proposes the design of an automatic target positioning system based on binocular stereo vision technology, where the 3D coordinates of a target are acquired through the binocular stereo vision subsystem, then the manipulator will be controlled to perform automatic target positioning according to the acquired 3D coordinates. The experimental result showed that this system has greatly extended the capabilities of a manipulator. Key Words: Binocular stereo vision, Camera calibration, Motion control Robotics is a comprehensive high-tech field involving mechanics, sensor technology, drive technology, control technology, communication technology, and computer technology. It is both an important foundation of opto-mechatronics and a typical representative of opto-mechatronics technology. It is an inevitable result of a multidisciplinary technological revolution. In recent years, with the continuous development of robotics research, robotics technology has begun to penetrate into all areas of human activity. Combining the application characteristics of these fields, various robots with different functions have been developed and widely used in different application areas. For example, the MR5 and MR7 bomb disposal robots produced by the American company Worstenholme Robotics can be used in indoor and outdoor environments, adapt to various terrains, and perform bomb disposal functions. They have been widely used by the US military. Bomb disposal robots are a type of special robot, mainly used to dispose of explosives and other hazardous materials at incident sites. The multi-functional manipulator of a bomb disposal robot, as the main actuator for completing the grasping task, should be able to complete a series of tasks, including grasping explosives, which is particularly important for bomb disposal robots. Bomb disposal is a complex and ever-changing process. The most crucial step in the process of a bomb disposal robot performing its task is controlling the robot's multi-functional manipulator to grasp the target object, that is, controlling the manipulator to accurately position the target object and complete the grasping action. Currently, all bomb disposal robots in use worldwide require an experienced operator to remotely control the robotic arm for precise positioning. This method places high demands on the operator, and manual control is also difficult to achieve high precision. Therefore, if computer vision technology could be used to achieve automatic and precise positioning of the robotic arm without requiring manual control of its joints, the performance of bomb disposal robots would be greatly improved. This paper details the design of a multi-functional robotic arm automatic positioning system for a bomb disposal robot based on binocular stereo vision. The system first calculates the three-dimensional coordinates of the target object in the visual coordinate system using binocular stereo vision technology, then transforms these coordinates to the robot coordinate system, and finally controls the robotic arm to achieve automatic positioning based on these coordinates. This paper provides a detailed description of the robot's system structure and the binocular stereo vision subsystem. Finally, a practical grasping experiment was conducted, achieving good experimental results. 1. System Structure The main functional components of this bomb disposal robot are a remotely controllable trolley and a multi-jointed multi-functional robotic arm. The assembly diagram of the robot is shown in Figure 1. [align=center]Figure 1 System Assembly Schematic Diagram Figure 2 System Assembly Schematic Diagram[/align] During the bomb disposal robot's mission, the vehicle will be remotely controlled to the environment where the hazardous object is located, placing the hazardous target within the robotic arm's workspace. The robotic arm can then begin its work, completing the task of grasping the target. The robotic arm's design mimics the structure of a human arm, with a total of six degrees of freedom, including waist rotation joint, shoulder rotation joint, elbow rotation joint, wrist rotation joint, gripper rotation joint, and gripper opening and closing joint. This multi-degree-of-freedom design gives the robotic arm great flexibility to adapt to the requirements of complex working environments. Two cameras are mounted on the robotic arm's forearm to act as the binoculars of the stereo vision system. To achieve the robotic arm's automatic target positioning function, in addition to the robot itself working on-site, a back-end computer is also required for support. Two cameras transmit the images of the target object to a backend computer. The stereo vision subsystem on the backend computer calculates the three-dimensional coordinates of the target object in the camera coordinate system using these two two-dimensional images, converts them into three-dimensional coordinates in the robot coordinate system, and transmits them to the embedded computer on the robot body. The embedded computer then controls the arm's motion, guiding the arm to the target object's position, thus achieving automatic target localization. The system structure is shown in Figure 2. 2. Binocular Stereo Vision Subsystem Binocular stereo vision refers to using two CCD cameras of comparable performance and fixed positions to acquire two images of the same scene. The three-dimensional information of the scene is calculated using the two-dimensional images acquired by the two cameras. In principle, it is similar to human binocular vision. Building a complete binocular stereo vision system generally requires steps such as camera calibration, image matching, and depth calculation. 2.1 Camera Calibration During camera imaging, the position of the object's image on the screen is related to the geometric position of the corresponding point on the object's surface in space. The relationship between these positions is determined by the camera's imaging geometric model, and the parameters of this geometric model are called camera parameters. These parameters must be determined through experimentation and calculation; this process is called camera calibration. Generally, the imaging model of a camera can be described by the following functional relationship: where I_left and I_right are the left and right image coordinates of the target, respectively; C_left and C_right are the three-dimensional coordinates of the target in the left and right camera coordinate systems, respectively; and the functions H_left and H_right represent the imaging models of the left and right cameras, respectively. H_left and H_right can be conveniently estimated using the Zhang plane method. After calibrating the left and right cameras separately, a stereo calibration of the binocular stereo vision system is also required. The goal is to obtain the relative positional relationship between the two cameras, described by the rotation matrix R and the translation matrix T. If we obtain the C[sub]left[/sub] and C[sub]right[/sub] data for multiple spatial points, we can obtain the estimated values ​​of the rotation matrix R and translation matrix T using the least squares method. The calibration results of this binocular stereo vision system are shown in Figure 3. [align=center] Figure 3 Calibration Results[/align] 2.2 Depth Calculation After single-camera calibration and stereo calibration, the model of this binocular stereo vision system is established. The following relationship exists between the two-dimensional image coordinates of the two cameras and the three-dimensional spatial coordinates: If we obtain the image coordinates I[sub]left[/sub] and I[sub]right[/sub] of any point in space in the images of the two cameras respectively, we can obtain the three-dimensional coordinates C[sub]left[/sub] of the spatial point in the left camera coordinate system by solving the above equations. Similarly, we can also obtain its three-dimensional coordinates C[sub]right[/sub] in the right camera coordinate system. The two image coordinates of the same point are obtained by the image matching process. 3. Embedded Control Subsystem The embedded computer located on the robot body is used to control the manipulator. This includes the overall motion planning of the arm and the position control of the DC motors of each joint. After the embedded computer receives the 3D coordinates of the target object from the backend computer, it plans the motion of each joint of the arm based on these coordinates and implements low-level motor position control. See Figure 4. [align=center]Figure 4 Embedded Control Subsystem[/align] The embedded control system uses a PC104 embedded computer in its hardware and the xPC kernel from The Mathworks in its software. 4. Experimental Results A target object was placed within the working range of the robotic arm, and an automatic grasping experiment was conducted using this system. During the experiment, the joints of the robotic arm were manually controlled first, allowing the target object to be observed through two cameras. Then, the binocular stereo vision system was activated, and the 3D coordinates of the target object were calculated using the images captured by the two cameras. These 3D coordinates were then transmitted to the embedded computer on the robot body, which controlled the arm to automatically position itself to the target object and perform the grasping. The experimental data is shown in the table on the right. The experiment shows that the system can achieve high positioning accuracy. 5. Conclusion Currently deployed bomb disposal robots require manual control of each joint of the robotic arm during actual bomb disposal operations to locate and grasp the target. This method places high demands on the operator and struggles to achieve high positioning accuracy. Automated positioning control of the robotic arm based on binocular stereo vision technology overcomes the shortcomings of traditional manual operation, improving the performance and ease of operation of bomb disposal robots. The innovation of this paper lies in achieving automated positioning control of the robotic arm through binocular stereo vision technology. This involves using two cameras to capture the two-dimensional image coordinates of the target object, calculating its three-dimensional coordinates, and then controlling each joint of the robotic arm based on these coordinates to achieve automatic positioning. This method improves the ease of operation and accuracy of bomb disposal robots, enhancing their performance and promising applications in real-world scenarios. References: [1] Xu Yongxi, et al. Control system for bomb disposal robot based on MatlabRTW. Microcomputer Information, 2006, 2-2: 218-220. [2] Liu Shui, Zhao Qunfei, et al. Design of control system for anti-riot robot based on PLC. Computer Engineering and Application, 2002, 31(1), 22-24. [3] Liu Yanyu, Li Deliang, Zhang Feilong, Wang Changlong. Application of ADS7852 in binocular ranging. Microcomputer Information, 2006, 4-2: 200-202. [4] Chen Xiai, Huang Xiaoming, Xu Fang. Robot positioning system based on eye fixed installation method. Microcomputer Information, 2006, 3-2: 182-184. [5] Liu Jianghua, Chen Jiapin, Cheng Junshi. Research on binocular vision platform [J]. Robotics Technology and Application, 2002, (01).