基于单应性矩阵和特征点筛选机制的线结构光平面标定  

The linear structured light plane calibration based on homography matrix and feature point screening mechanism

作　　者：霍鹏飞 洪磊 梁越霄 Huo Pengfei;Hong Lei;Liang Yuexiao(School of Automotive and Rail Transit,Nanjing Institute of Technology,Jiangsu Nanjing,211167,China)

机构地区：[1]南京工程学院汽车与轨道交通学院,江苏南京211167

出　　处：《机械设计与制造工程》2025年第1期79-84,共6页Machine Design and Manufacturing Engineering

基　　金：江苏省高等学校基础科学(自然科学)研究重大项目(23KJA460007)。

摘　　要：针对交比不变方法在靶标平面获取特征点数量少、初始特征点提取受光照条件和图像噪声畸变影响存在较大误差等问题,基于单应性矩阵的映射关系和特征点筛选机制,提出了线结构光平面标定的优化方法。利用OpenCV提取光条纹并结合单应性矩阵计算特征点在相机坐标系下的坐标,将其拟合得到初始光平面,以特征点到光平面距离的平均值作为阈值构建特征点筛选机制,通过筛选点拟合得到优化后的光平面。对标准量块进行测量,平均误差为0.06484 mm,实验表明该优化方法显著降低了初始特征点的拟合误差,具有较高的精度。Aiming at the small number of feature points in target plane and large error in initial feature point extraction due to illumination conditions and image noise distortion,an optimization method of linear structured light plane calibration is proposed by combining the mapping relationship of homography matrix and feature point screening mechanism.OpenCV is used to extract the optical fringe and calculate the coordinates of the feature points in the camera coordinate system combined with the homography matrix,and the initial light plane is obtained by fitting it.The average distance between the feature points and the light plane is used as the threshold to construct the feature point screening mechanism,and the optimized light plane is obtained by fitting the screening points.The average error of the standard measure block is 0.06484 mm,and the experiment shows that the optimization method significantly reduces the fitting error of the initial feature points,and has high accuracy.

关 键 词：线结构光传感器 单应性矩阵 光平面标定 三维测量 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]