基于双目立体视觉的工件识别定位与抓取系统研究

摘  要

随着机器视觉技术的发展，立体视觉尤其是双目立体视觉被广泛应用于物体识别、虚拟现实、工业检测、机器人导航和航空航天等领域。双目立体视觉技术可在多种条件下灵活的获得景物的立体信息，相对单目视觉而言有着不可比拟的优势，是图像处理与机器视觉领域的前沿研究方向。本文采用双目立体视觉系统对目标工件识别定位和机器人抓取试验进行了研究，主要涉及摄像机的标定、目标工件的识别定位匹配和机器人对目标工件的抓取三大部分内容。另外，对目标跟踪进行了初步地研究与试验。

1）采用基于透视变换模型的线性标定方法，在MATLAB软件环境中，利用双目立体视觉系统对摄像机进行标定。该标定方法过程简便，精度基本满足要求。

2）利用HALCON编程环境，对目标工件的图像进行识别匹配，并通过提取边缘特殊点和几何中心，完成工件的定位。该方法在采集图像时可以对图像进行预处理，计算速度快，使用灵活。

3）将工件图像的定位结果传到机器人的控制系统中，通过运动学反解将计算出的世界坐标值转化成机器人各关节需要转过的角度值，然后驱动机器人，完成工件的抓取。

通过工件识别定位和机器人抓取试验，证明该方法计算过程简单，具有较高的控制精度，使用方便，能满足要求。

关键词：双目视觉；标定；HALCON；识别；定位

Abstract

With the development of computer vision technology, the stereo vision, particularly the binocular stereo vision is widely used in object recognition, virtual reality, industrial inspection, robot navigation and aerospace and other fields. The binocular vision can flexibly obtain three-dimensional information of objects in various conditions. It has unparalleled advantages compared with the monocular vision, and it is the forefront of the research directions of image processing and machine vision field. In this paper, using the binocular stereo vision system to study the recognition-positioning of target workpiece and grabbing robot, it mainly relates to the three contents of camera calibration, recognition, loaction and matching of the target workpiece and the workpiece manipulator grabbing. In addition, it also resolves preliminary study and experimental on target tracking.

1) Useing the linear calibration method based on a perspective transformation model to calibrate the binocular stereo vision system in the MATLAB software. This method is simple, and its precision can relatively meet the request.

2) Recognition and matching the images of target workpiece in the programming environment of HALCON, and extraction of the edge of the special points and the geometric center to location the target workpiece. The method can treat the image pre-processing when acquiring the images, it is fasts and flexibility.

3) Sending the location results of workpiece image to the control system of the robot and calculating the world coordinates to get the required turning angle of the robot joints by inverse kinematics solution, then drive the robot to grab the workpiece.

By the experiment of recognition-position of workpiece and robot grabbing,it proved that the method has simple calculation,higher control precition,convenience, and it can meet the requst.

Key words:binocular vision;calibration;HALCON;recognition;locatio