Research Of Robot Positioning Technology Based On Binocular Vision

With the development of science and technology, robot as a product integrating all subjects, is blossoming rapidly. Intelligence has become the primary goal to develop robotics, and machine vision is one of the key factors to improve intelligent level. As a key member of machine vision, binocular vision has been widely used in many fields of production such as industrial detection, medical examination, robot navigation, etc. A further research on binocular vision technique can not only accelerate the development of theory, but also have practical significance. This paper built a robotic vision system using the self-developed industrial robot JNPF-6R, and studied on the robot location technique based on binocular vision.Firstly, researched the imaging model of camera and the reference coordinate systems involved, specifically illustrated the most common models in camera, the pinhole imaging model and the internal and external parameter model. Then, researched the two models of binocular vision, and chose an ordinary model according. Finally, after completing the hardware selection, designed and built the hardware system of robot vision.Specific to the camera calibration of internal parameter, after comparing three methods, adopted Zhengyou Zhang calibration for the system, also optimized the experiment parameters. Then, researched the mathematical model of hand-eye calibration and some common calibration means. On this basis, combining the traditional calibration with active vision calibration, proposed a hand-eye blended calibration scheme which used the shift of camera and the obtain of external parameter by target board to achieve the final hand-eye parameters and took advantage of this hybrid method to write Matlab programs, testing the hand-eye calibration and verifying the experiment results by the traditional equation AX=XB. Finally, proved the accuracy and effectiveness of this algorithm.After calibration, specific to the noise in the image, studied on the principles of various filtering methods, conducted an experiment and discussed the results. According to the analysis conclusion, the median filter algorithm was adopted. Then, researched the segmentation means based on threshold, namely, image binarization at length. In addition, conducted an experiment, and chose self-defining threshold method for further processing after comparing the experiment results. In allusion to the handled images on the left and right, solved and matched the centroid point as well as the pixel coordinate, also focused on researching and deducing the 3D localization algorithm of target point. On the basis of this algorithm, solved the 3D position information of target point by the two pixels' coordinate values and calibration data.Based on the study on the principle and algorithm of robot visual location technique, developed robotic visual positioning control software, designed the robot visual positioning instance. Analyzed primary functions of the software, proposed the overall scheme and general framework of the software design. Finally completed the instance for robot visual localization, verifying that both the proposed scheme and the developed software could be realized.