立体视觉系统及其在静态尺寸与动态轨迹测量中的应用

发布时间：2018-01-09 15:05

本文关键词：立体视觉系统及其在静态尺寸与动态轨迹测量中的应用　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：现代制造业的快速发展对测量技术的自动化和智能化程度的要求不断提高。机器视觉具有非接触、精度高、速度快、灵活、信息丰富等优点,在产品的质量检测、机械加工、目标追踪等领域具有重要的研究价值。本文以立体视觉测量系统为研究对象,分析了相机标定、图像采集和预处理、特征提取与匹配、几何测量算法等理论问题,构建了立体视觉测量系统软硬件平台,实现了立体视觉系统在静态工件尺寸测量和动态目标空间轨迹信息获取中的有效应用。首先阐述了机器视觉测量技术国内外现状和发展趋势,重点分析了视觉系统的硬件性能指标和选型依据,基于Basler工业相机、Kowa镜头、STM32等硬件设备搭建了立体视觉测量平台,使用STM32开发板对工业相机组进行同步触发控制实现了立体视觉图像采集。分析了相机成像的原理,建立了理想成像和实际畸变模型。深入研究张正友标定法,明确了单目标定、双目标定、畸变矫正的原理和实现流程,得到由图像坐标向实际坐标转换的理论依据和计算方法。研究了视觉测量技术相关的图像滤波、二值化、几何特征提取算法,对比Hough算法和最小二乘算法效果,结合圆度误差和最小二乘法对圆形特征进行检测,结合最小外接矩形和Harris算法对矩形特征顶点亚像素坐标进行提取,提出了基于Harris-NCC算法进行角点特征立体匹配优化的计算依据。将研制的立体视觉测量系统应用于工件几何尺寸的机器视觉测试实验和羽毛球三维空间运动轨迹测量实验,分析了实验误差,明晰了视觉测量系统的性能参数,验证了所提算法的有效性,同时也为一些先进的立体视觉算法在工业静态测量和动态目标监测领域的实用化提供了一种解决方案。
[Abstract]:The rapid development of modern manufacturing industry requires the automation and intelligence of measurement technology. Machine vision has the advantages of non-contact, high precision, fast speed, flexibility, rich information and so on. It has important research value in the field of product quality testing, machining, target tracking and so on. In this paper, the camera calibration, image acquisition and preprocessing are analyzed with stereoscopic vision measurement system as the research object. The software and hardware platform of stereo vision measurement system is constructed based on the theory of feature extraction and matching, geometric measurement algorithm and so on. It realizes the effective application of stereo vision system in static workpiece dimension measurement and dynamic target space track information acquisition. Firstly, the present situation and development trend of machine vision measurement technology at home and abroad are described. The hardware performance index and the selection basis of the vision system are analyzed emphatically, and the stereo vision measurement platform is built based on the hardware equipment such as Basler industrial camera, Kowa lens and STM32. Using STM32 development board to realize the synchronous trigger control of industrial camera group, the stereo vision image acquisition is realized, and the principle of camera imaging is analyzed. The ideal imaging and actual distortion models are established. The calibration method of Zhang Zhengyou is studied in depth. The principle and realization process of single target setting, double target setting and distortion correction are clarified. The theoretical basis and calculation method of the transformation from image coordinates to actual coordinates are obtained, and the image filtering, binarization and geometric feature extraction algorithms related to visual measurement technology are studied. Compared the effect of Hough algorithm and least square algorithm, combined with roundness error and least square method to detect the circular features. The sub-pixel coordinates of the rectangular feature vertices are extracted with the minimum external rectangle and Harris algorithm. The calculation basis of corner feature stereo matching optimization based on Harris-NCC algorithm is put forward. The developed stereo vision measurement system is applied to machine vision testing experiment of geometric dimension of workpiece and badminton. The experiment of motion trajectory measurement in dimensional space. The experimental error is analyzed, the performance parameters of the vision measurement system are clarified, and the validity of the proposed algorithm is verified. It also provides a solution for the application of some advanced stereo vision algorithms in the field of industrial static measurement and dynamic target monitoring.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】