面向视觉标定的图像特征点检测算法研究

发布时间：2018-05-08 06:22

本文选题：机器人视觉 + 特征点提取　；参考：《昆明理工大学》2017年硕士论文

【摘要】：视觉的引入是机器人智能化的一个巨大飞跃,机器人视觉标定技术是实现机器人视觉伺服控制、引导机器人运动和测量的前提,由此可见机器人视觉标定对机器人完成相关动作至关重要。本文主要就摄像机视觉标定技术在机器人视觉引导方面进行研究,机器人视觉标定精度直接影响到视觉引导的准确性,而图像特征点检查精度又直接影响摄像机标定精度,因此重点研究了摄像机标定过程中图像特征点检查技术,本文的主要研究工作如下:1、从应用角度出发,提出了一种改进的Harris方格板角点检测算法。该算法综合运用了 Harris角点检测技术,并在保留了 Harris角点检测算法良好的可重复性和相对较高的检测效率的情况下使其精度和可重复性更高,同时很好的解决了相机内参数计算过程中:方格板图像的角点坐标和空间点相匹配这一难点问题,并有效的解决了原有Harris角点检测算法阈值的选取过度依赖的问题。2、提出了一种基于几何对称性并应用于椭圆阵列图像的圆心检测算法,该算法通过设定约束条件来限定搜索范围,避免出现漏检或相互混淆而影响检测的精度;同时有效的解决了标定板的标记点和图像特征点难以严格匹配的问题。3、寻求一种完全自动的圆心特征提取算法,达到降低机器人标定对工人的文化水平需求,从而降低企业成本的目的;并针对现有相关技术中同时对多个椭圆进行检测时出现精度不高和难以准确排序的问题,提出一种将最小二乘法应于椭圆阵列图像的椭圆拟合算法,该算法无需人机交互协同完成检测工作,全部自动完成提取,提取到的圆心特征点精度和匹配度相对较高,可重复性好。4、利用上述的三种特征点提取算法利用平面标定法进行实验,计算出摄像机的内外参数,并进行对比分析;将标定数据写入机器人控制器,进行视觉抓取实验进一步验证标定算法的正确性。
[Abstract]:The introduction of vision is a great leap of robot intelligence. Robot vision calibration technology is the premise to realize robot vision servo control, guide robot motion and measurement. It can be seen that robot vision calibration is very important for robot to complete related actions. In this paper, the camera vision calibration technology in robot vision guidance is studied. Robot vision calibration accuracy directly affects the accuracy of vision guidance, and image feature point inspection accuracy directly affects camera calibration accuracy. Therefore, this paper focuses on the image feature point detection in camera calibration. The main work of this paper is as follows: 1. From the point of view of application, an improved Harris grid corner detection algorithm is proposed. In this algorithm, the Harris corner detection technique is used synthetically, and the accuracy and repeatability of the Harris corner detection algorithm are improved by keeping the good repeatability and relatively high detection efficiency of the Harris corner detection algorithm. At the same time, it solves the difficult problem of matching corner coordinates and space points in the calculation process of camera inner parameters. The problem of over-dependence of threshold selection of the original Harris corner detection algorithm is effectively solved, and a circle center detection algorithm based on geometric symmetry and applied to elliptical array images is proposed. The algorithm limits the search range by setting the constraint conditions to avoid the error detection or confusion and affects the accuracy of the detection. At the same time, it effectively solves the problem that the mark points and image feature points of the calibration board are difficult to match strictly, and seeks a completely automatic center feature extraction algorithm to reduce the requirement of robot calibration for workers' cultural level. In order to reduce the cost of the enterprise, aiming at the problems of low precision and difficult to sort the ellipses in the existing related technology, an ellipse fitting algorithm is proposed, which applies the least square method to the elliptic array image. This algorithm does not need man-machine interaction cooperation to complete the detection work, all of which are automatically extracted, and the accuracy and matching degree of the extracted centroid feature points are relatively high. The method of plane calibration is used to carry out experiments, to calculate the internal and external parameters of the camera, to compare and analyze, to write the calibration data into the robot controller, and to write the calibration data into the robot controller. The calibration algorithm is verified by visual capture experiment.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】