首页|基于蜂窝立体结构靶标的多目相机标定方法

基于蜂窝立体结构靶标的多目相机标定方法

扫码查看
为解决多目相机快速标定以及构建不同坐标系之间位姿关系的问题,提出一种基于新型蜂窝立体结构靶标的多目相机标定方法,能够快速准确完成多目相机标定。通过仿生学原理设计的蜂窝立体结构具有5个附着自识别阵列标记的不同空间位姿平面,提升了标定稳定性以及鲁棒性;使用多目相机拍摄一幅图像即可完成相机标定以及坐标系间关系构建,提升了标定效率;最后,将所提方法与传统的相机标定方法进行了实验对比。结果表明:相比于传统标定方法,所提方法的相机标定平均重投影误差降低了 18。4%,多目相机间外参平均重投影误差降低了 8。5%,遮挡干扰下平均重投影误差的相对误差仅为0。01 pixel,进一步验证了所提标定方法的可行性。
Calibration Method for Multi-camera Based on Honeycomb Stereo Target
In order to address the issue of rapid calibration of multi-camera and the construction of pose relationships between dif-ferent coordinate systems,a multi-camera calibration method based on a novel honeycomb stereo target was proposed to enable swift and precise multi-camera calibration.The honeycomb structure designed by bionics principle has five different spatial pose planes attached with self-identification array tags,which could improve the calibration stability and robustness.Using a multi-camera to capture one im-age could complete camera calibration and build relationships between coordinate systems,which improved calibration efficiency.Final-ly,the proposed method was compared with traditional camera calibration methods.The results indicate that compared to traditional cali-bration methods,the proposed method reduces the mean reprojection error of camera calibration by 18.4%and the mean reprojection er-ror of external parameters between multi-camera systems by 8.5%.The relative error of the mean reprojection error under occlusion in-terference is only 0.01 pixel,the feasibility of the new calibration method is further validated.

machine visionstereo targetself-identification array tagsmulti-camera calibrationvision measurement

秦鹏举、蒋周翔、苏瑞、马紫怡、宋鹏成

展开 >

北京信息科技大学机电工程学院,北京 100192

机器视觉 立体结构靶标 自识别编码图案 多目相机标定 视觉测量

国家自然科学基金面上项目国家自然科学基金青年科学基金项目

5217545252005046

2024

机床与液压
中国机械工程学会 广州机械科学研究院有限公司

机床与液压

CSTPCD北大核心
影响因子:0.32
ISSN:1001-3881
年,卷(期):2024.52(12)
  • 10