一种单目视觉位姿测量系统的误差分析方法

针对单目视觉位姿测量传统误差分析方法中只考虑单一误差因素,分析结果与工程实际有较大差异的问题,提出一种考虑多误差因素共同作用的误差分析方法。根据单目视觉系统的测量范围和相机参数建立其位姿测量模型;综合考虑相机内参、镜头畸变、图像点、靶标三维点等误差因素,将其同时引入位姿测量模型进行分析;分析抑制不同参数误差及提高相机分辨率等在多误差因素共同作用下对位姿测量结果的影响,找到最有效的精度优化方法。采用小型机械臂手眼定位中的单目视觉位姿测量系统进行实验,结果表明通过抑制相机径向畸变误差和提高相机分辨率,能够有效地提升该系统的位姿测量精度,位姿精度分别提升6.57%、4.21%和5.88%、5.54%。

Error analysis method for monocular vision pose measurement system

The traditional error analysis of monocular vision pose measurement only considers a single error factor, which causing the problem that there is a large difference from the actual project. An error analysis method considering multiple error factors was proposed. Firstly, a pose estimation model for monocular vision was established based on the measurement range and camera parameter. Secondly, the camera internal error, lens distortion error, image points detection error and target error were introduced into the measurement model simultaneously based on actual error levels. Finally, the influence of suppressing parameter errors, improving camera resolution on the pose measurement results were obtained, and the most effective precision optimization method was found. In addition, a monocular vision pose measurement system for small manipulator hand-eye location was adopted for experiment. The analysis results indicate that reducing the radial distortion error and improving the camera resolution can effectively improve both the position and attitude measurement accuracy of this system. The pose accuracy is increased by 6.57%, 4.21% and 5.88%, 5.54%, respectively.