Computer vision-based multi-objective pose measurement method
|School||Harbin Institute of Technology|
|Keywords||Computer Vision Multi- target recognition Pose measurement Encoding target Image Stitching|
\an important foundation for the environment, for the above ground measurement system to provide micro-satellite simulator system position and attitude of the external measuring instruments. Since computer vision in the non-contact measurement of the target state has obvious advantages, this paper combined with \measurement method, made the following innovative research results. For large-scale, multi-objective, such as the position and orientation precision ground measuring system requirements is presented in the test set aided measurement target surface targets monocular vision measurement system overall program, designed four vision cameras partition imaging and then spliced imaging programs to reduce single-vision camera resolution and field of view requirements; vision measurement system and the key parameters, including the lens field of view FOV, vision camera focal length and image acquisition card, the data transfer rate, etc. computational analysis. Proposed a multi-objective measurement of the position and orientation measurement target design methods, the design of a distribution from a central dot and circle the perimeter of certain coded dots on the target geometry measurements given coded dots binary encoding method gives positive direction measurement target labeling method to the measurement target with tiny satellite simulator for attitude measurement function. Designed for the measurement target is proposed based on a single image encoding target logo recognition and pose determination method and coding design targets, identify, and methods for determining the position and orientation parameters of the simulation, the results show that the proposed method is feasible and effectiveness. Vision measurement system for large-scale high-precision measurement needs, as well as four re-vision camera stitching partition imaging imaging solutions, gives a global multi-camera calibration and multi-camera image mosaic method. In this method, the image of single vision camera and calculating the nonlinear distortion correcting internal parameters of the camera, and the camera determines the camera coordinate system and the reference coordinate system transformation matrix determined by the control point coordinate system of the reference camera with the overall world coordinates the transformation matrix between the Department and finally dividing line treatment using overlapping multi-camera field of view redundant data. Finally, an image sequence based on continuous multi-objective pose Kalman filtering method, effectively improve the moving targets pose measurement accuracy. Analyzes the vision measurement system measurement errors, and vision measurement system performance simulation results show that the proposed overall scheme feasible method is effective.