鸡腿菇采摘机器人的视觉系统研究

发布时间：2018-04-02 12:33

本文选题：采摘机器人　切入点：双目立体视觉　出处：《兰州理工大学》2017年硕士论文

【摘要】：在采摘机器人领域,双目立体视觉的应用非常广泛。双目立体视觉基于双目摄像机与人的双眼在感知空间物体方面的相似性,具有系统结构简单、定位精度高和设备费用低廉等优点,一直是机器视觉的一个非常活跃的分支。本文是鸡腿菇采摘机器人的视觉系统研究,以双目立体视觉为研究对象。采用改进的Canny边缘检测算法识别鸡腿菇果实,使用张正友标定法做了双目摄像机的标定,并使用SURF法做了立体图像的特征提取和匹配,为获取鸡腿菇三维坐标信息打下了基础。本文所做主要工作包括:(1)在阅读大量国内外文献的基础上,搭建了完整的双目立体视觉系统,分析了选择双目视觉的原因,说明了双目视觉的测距原理。其中该系统所用的主要硬件有:两台同型号高清摄像机、一台计算机和简易云台。(2)为了达到识别鸡腿菇果实目的,本文采用改进的Canny边缘检测算子的方法来提取鸡腿菇菌盖边缘。由于双边滤波具有比高斯滤波更好的保边能力,后期再使用小波变换对低频系数进行压缩,同时拉伸高频系数,使得边缘信息得到增强。通过实验证实了该改进算法相比原来算法的优越性,该改进算法是本文主要创新点。(3)为了获取摄像机的内部参数和外部参数,本文通过使用经典的张正友标定法,对双目立体视觉系统的左、右摄像机进行了标定。通过调用计算机开元视觉库Open CV2.11来验证张正友标定法,实验表明该标定方法具有操作简单,对设备要求低,标定精度高的优点。(4)为了找到目标物体在左、右摄像机所拍摄的两幅图像中的对应点,计算出视差,为深度恢复和三维重建打好基础,本文研究了立体匹配的基元选择、匹配准则和方法,并基于SURF图像匹配算法做了匹配实验。实验结果表明该方法对于一般图像匹配效果不错,匹配正确率能达到95%;但该方法应用于立体图像匹配,匹配正确率在80%以下,尚需改进算法或者使用其他匹配算法来做匹配。
[Abstract]:In the field of picking robot, binocular stereo vision is widely used. Binocular stereo vision is based on the similarity between binocular camera and human eyes in sensing space objects. The advantages of high positioning accuracy and low cost of equipment have been a very active branch of machine vision. This paper is the visual system research of the picking robot of Coprinus comatus. The improved Canny edge detection algorithm was used to identify the fruit of Coprinus comatus, the calibration method of Zhang Zhengyou was used to calibrate the binocular camera, and the SURF method was used to extract and match the features of the stereo image. This paper has laid a foundation for obtaining three-dimensional coordinate information of Coprinus comatus. The main work in this paper includes: 1) on the basis of reading a large number of literatures at home and abroad, a complete binocular stereo vision system is built, and the reasons for choosing binocular vision are analyzed. The principle of binocular vision ranging is explained. The main hardware used in this system is: two high-definition cameras of the same type, a computer and a simple cloud head. 2) in order to recognize the fruit of Coprinus comatus, In this paper, the improved Canny edge detection operator is used to extract the edge of the mushroom cap of Coprinus comatus. Because the bilateral filter has better edge preserving ability than Gao Si filter, wavelet transform is used to compress the low frequency coefficient and stretch the high frequency coefficient. The edge information is enhanced. Experimental results show that the improved algorithm is superior to the original algorithm, which is the main innovation of this paper) in order to obtain the internal and external parameters of the camera, In this paper, the left and right cameras of binocular stereo vision system are calibrated by using the classical calibration method of Zhang Zhengyou, and the calibration method of Zhang Zhengyou is verified by using Open CV2.11. Experiments show that the calibration method has the advantages of simple operation, low requirement for equipment and high calibration accuracy. In order to find the corresponding points of the target object in the left and right images, the parallax is calculated. In order to lay a good foundation for depth restoration and 3D reconstruction, this paper studies the basic selection, matching criteria and methods of stereo matching, and makes a matching experiment based on SURF image matching algorithm. The experimental results show that this method is effective for general image matching. The accuracy rate of matching can reach 95%, but this method is applied to stereo image matching, and the accuracy of matching is below 80%, so we still need to improve the algorithm or use other matching algorithms to match.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】