【作者】 姜淑华；

【导师】 王文生；

【作者基本信息】 长春理工大学 ， 光学工程， 2009， 博士

【摘要】 本文将现代光学测试技术、计算机技术和光电子技术相结合,研究了基于面阵CCD、成像物镜和微机的微小尺寸检测系统。检测系统将被测工件图像通过CCD相机成像,该模拟信号经过图像采集卡的A／D转换,转换为数字信号后,送入微机存储在内存中,再通过编写的程序对工件图像进行图像处理。本文解决了非线性平滑、阈值确定和边缘提取等关键技术,利用边缘轮廓跟踪的方法获取工件的边缘轮廓图像,系统可精确的确定工件的边缘轮廓曲线。本系统进行了大量的实验,以微米级工件、圆形工件和矩形工件等工件测量为例,说明了编写工件测量程序的思想,方法简单,实现了二维尺寸的高精度、非接触、实时自动检测。系统结构简单,易于操作,对环境要求宽松。在二维尺寸非接触自动测试的基础上,基于双目立体视觉技术本文研究了三维尺寸的非接触自动检测,对双目立体视觉技术的摄像机标定、特征提取和特征立体匹配等内容做了详尽的研究。在摄像机标定部分,本文提出了对Tsai两步法的改进算法,完成了对双目视觉摄像机的标定。在介绍立体视觉三维测量的基本原理的基础上,研究摄像机标定算法,运用RAC"两步法”标定技术,利用Levenberg-Marquardt非线性优化算法,并对CCD摄像机进行了标定实验。在特征提取部分,对Harris角点提取算法、SUSAN角点提取算法和SIFT特征提取算法进行了分析。通过不同图片对上述三种特征点提取算法进行比较和分析,实验表明SIFT特征提取算法提取的特征点数目较多,有利于作为匹配基元进行下一步的立体匹配,并且图片发生旋转时,SIFT特征提取算法检测出来的特征点在位置和数目上几乎没有变化。在立体匹配部分,对完全可见的物体表面和部分可见的物体表面进行特征匹配分别进行了研究,得到对应的匹配公式。分别实现了Harris角点匹配和SIFT匹配算法,并对这两种方法的匹配结果进行比较和分析。实验结果表明基于SIFT特征点的匹配算法要优于基于Harris角点匹配算法。本系统以VisualC++为开发平台结合图像处理的OpenCV库函数合理地构建了一个逻辑清晰、工作稳定的软件框架,满足了系统测试要求。在所开发的实验系统上,分别以书包和花瓶图片为例,开展了三维测量实验,完成了测量任务。更多还原

【Abstract】 In this paper modern optical test technique, computer technique and photoelectric technique are combined and micro-size detection based on the CCD matrix, imaging objective and computer is researched. The detection system images the measured workpiece by the CCD camera, the analog signal is transferred to the digital signal by the A/D converter on the image collector, the digital signal is stored in the memory and the workpiece image is processed by the compiled program. The paper solves the key technologies of non-linearity smooth, the threshold value selection and edge abstraction. the edge contour curve of workpiece can be precisely determined with the method of the edge contour tracking to obtain the edge contour image of workpiece. The system carried out experiments and explained the program ideas with the example of the nanoscale workpiece, circular workpiece and rectangle workpiece. It’s easy to realize 2-D high accuracy, non-contact, real time and automatic measurement. The system has a simple structure and can be easily operated without the strict requests for environment.In this paper non-contact automatic measurement of three dimension is researched based on non-contact automatic measurement of two-dimension and binocular stereo vision technology. The paper studies the camera calibration, feature extraction and feature stereo matching of binocular stereo vision technology in detail.In the part of camera calibration, improved algorithm of Tsai-two-step method is proposed to achieve the binocular vision camera calibration. The paper researched the camera calibration algorithm based on the stereovision 3-D measurement principles and carried out the calibration experiments by employing the RAC two-step calibration technology and Levenberg-Marquardt nonlinear optimal algorithm.In the part of feature extraction, the paper analyses the Harris corner extraction algorithm and SUSAN corner extraction algorithm and SIFT feature extraction algorithm. The three feature extraction algorithms of different pictures are compared and analyzed. Experimental results show that the extracted feature points of SIFT feature extraction algorithm are more than other methods and it’s beneficial to realize the further stereo matching of the matched base unit. When the picture is revolved, the position and number of the detected feature points are hardly changed.In the part of stereo matching, the objective surfaces of full-visible and part-visible are feature matched respectively and the matching formulas are obtained. The Harris corner matching and SIFT matching algorithms are implemented and the matching results of the two methods are compared and analyzed. Experiment results show that the performance of the SIFT matching algorithm is better than that of the Harris algorithm.A software framework with clear logic and stable performance based on the image processing library function OpenCV in the Visual C++ development platform is constructed. The system measurement demand is satisfied. In the developed experiments system,3-D measurement experiments are carried out with the examples of schoolbag and vase and performed the measurement tasks.更多还原