基于阵列相机的图像测量技术研究

发布时间：2018-05-20 05:58

本文选题：阵列相机 + 相机标定　；参考：《北方工业大学》2017年硕士论文

【摘要】：图像测量技术是一种新兴高效的测量技术,它在许多的领域都有重要的应用,比如计算机视觉,智能检测,人脸识别和模式识别等。图像测量的方法就是把目标的信息量传入测量系统并进行测量的一门技术。目前的生产科学和工业技术正在快速发展,图像测量技术也越来越多的应用到了生产和制造等方面。因此,图像测量技术的研究是一个非常有价值的问题,它可以降低工业成本,提高工艺精度,提高工作效率。在21世纪,阵列相机开始逐步走进人们的视野,与传统的相机相比,它有许多的优势:先拍照后聚焦、获得深度信息、重建的图像视角更大等。基于双目的图像测量已有大量的研究成果,但是在基于阵列相机的图像测量方面,国内的研究还很少。基于此,本文选用3×3的阵列相机作为研究平台,研究基于阵列相机的图像测量,研究内容主要包括相机标定、立体匹配、三维重建、图像的长度测量四个主要部分,具体的工作内容分为:1.简单叙述了相机成像原理、张正友的摄像机标定原理、常用的立体匹配方法以及视差测距原理;2.本文使用3×3的阵列相机拍摄目标图像,并且对9台相机分别标定获,进而获取它们的内外参数;3.通过NCC立体匹配的方式得到深度信息,并通过卷积优化运算速度,提高效率,然后对于目标图像进行超分辨率重建,获得完整的场景信息;4.在图像中通过鼠标操作可以获得图像中点的三维坐标,完成两点的欧氏距离测量。本文算法针对距离较近的目标物体测量结果精度较高,这是因为阵列式相机获得的图像视角更广,而且还可以从重建的图像中提取目标点的深度信息。阵列相机相关方面的研究在国内目前还很不成熟,但是和传统相机相比其具备明显的优势,所以阵列相机应用前景非常广阔。
[Abstract]:Image measurement technology is a new and efficient measurement technology. It has important applications in many fields, such as computer vision, intelligent detection, face recognition and pattern recognition. The method of image measurement is a technique to transmit the information of the target into the measurement system and measure it. At present, production science and industrial technology are developing rapidly, and image measurement technology is more and more applied to production and manufacturing. Therefore, the research of image measurement technology is a very valuable problem, which can reduce the industrial cost, improve the process accuracy and improve the working efficiency. In the 21st century, array cameras begin to come into the field of vision gradually. Compared with traditional cameras, array cameras have many advantages, such as taking pictures first and then focusing, obtaining depth information, and so on. There have been a lot of research achievements in binocular image measurement, but there are few researches on array camera based image measurement in China. Based on this, this paper chooses the 3 脳 3 array camera as the research platform to study the image measurement based on the array camera. The research contents mainly include camera calibration, stereo matching, 3D reconstruction and image length measurement. The specific content of the work is divided into 1: 1. The principle of camera imaging, the camera calibration principle of Zhang Zhengyou, the commonly used stereo matching method and the principle of parallax ranging are briefly described. In this paper, 3 脳 3 array camera is used to capture the target image, and 9 cameras are calibrated respectively, and their internal and external parameters are obtained. The depth information is obtained by NCC stereo matching, and the operation speed is optimized by convolution to improve the efficiency. Then the target image is reconstructed by super-resolution, and the complete scene information is obtained. The 3D coordinates of the middle point of the image can be obtained by mouse operation in the image, and the Euclidean distance between the two points can be measured. The accuracy of the proposed algorithm is higher for the objects with a close distance because the array camera can obtain a wider view and can extract the depth information of the target point from the reconstructed image. The research on array camera is not mature at present in China, but it has obvious advantages compared with traditional camera, so the application prospect of array camera is very broad.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【引证文献】