Three-Dimensional Target Location System for Large Scene

Abstract

For the research of non-contact target location in large scene, a complete set of 3D positioning system is built. The system considers the basic principles of using binocular vision. Firstly, for binocular calibration problems in large scenes, such as the determination of large-scale calibrators, the distortion of industrial cameras and the parameter calibration of binocular vision system, the 2/3 principle of calibrators is adopted to construct the distortion model and binocular vision model respectively. The second is stereoscopic correction part. BM algorithm is used to stereo match the same points in the left and right camera images to get the parallax map, which speeds up the response speed of the system to get a fast perception of the change of position information. Finally, according to the parallax information in the parallax diagram, the world three-dimensional coordinate of the target object is solved. In this paper, the size binocular vision system was used as a control experiment. Large-scale calibration and distortion correction are concentrated in the large system, while stereoscopic correction and coordinate acquisition are completed in real time by the small system. It was verified in a small model experiment. The experimental results illustrate the situation of the target positioning system for large scene constructed in this paper. When the distance between the target object and the camera is no more than 1350mm, the positioning error can be kept at 10% to meet the requirements of system accuracy.