【作者】 胡平波；

【导师】 张春森；

【作者基本信息】 西安科技大学 ， 摄影测量与遥感， 2011， 硕士

【摘要】 摄影测量的主要目标之一是从非接触式影像中获取精确可靠的量测信息,计算机视觉其研究的主要目标是使计算机系统具有类似于人类的视觉能力,从而获取三维场景的几何信息。无论是在摄影测量领域,还是在计算机视觉领域,特征提取及匹配都有广泛地应用。特征提取及匹配是三维重建、运动跟踪、序列影像分析、信息融合、模式识别等研究课题的必要环节,对特征提取及匹配进行研究具有重要的现实意义。随着工业制造技术的不断提高和加工工艺的不断改进,在大批量工业生产中,用人工视觉检查产品质量效率低且精度不高,用计算机视觉技术可大大提高生产效率和自动化程度,而摄影测量中丰富的平差理论有助于获取极高的精度。近年来,高分辨率、高信噪比的数字成像CCD器件迅速发展,计算机图象处理和模式识别技术的快速进步,使得摄影测量计算机视觉技术成为目前具有广泛前景的一种精密测量技术。本论文针对工业生产中的精密测量要求,结合工业规则物体中拥有的丰富线特征信息,采用摄影测量中严密的数据处理理论及计算机视觉前沿理论和技术,重点研究了点、线特征提取及匹配问题。提出了直线元离散化获取点特征的方法及线特征的高精度提取新算法;改进了相关系数影像匹配(加权相关系数);将计算机视觉中的核线约束引入到最小二乘影像匹配中,并给出了一种遮挡直线匹配的新方法。通过实验验证了特征提取及匹配的有效性,得到了较好的结果。更多还原

【Abstract】 The dominant purpose of photogrammety is to use CCD images to obtain the accurate and reliable measurements while computer vision is to make computer have visible ability which resembles that of human’s eyes and to regain the geometries of three-dimensional space.Feature extracting and matching,which is an essential component of such research as three-dimensional reconstruction, motion tracking, image sequence analysis, information fusion, pattern recognition, are widely used in both computer vision and photogrammetry. There exits important practical significance in such studying.As the improving of industrial manufacturing technology and processing technic,we usually get a poor efficiency and short precision by human check in large quantities production although the production efficiency and automation can be greatly enhanced by computer vision and a high accuracy by the extensive adjustment theory. It makes a great progress in CCD devices with a high resolution and high signal to noise ratio and in image processing and pattern recognition, which enables photogrammetry-comuter vision to be a wide application way for technology of precise measurements.In order to accommodate the demands of industrial precision measurement, This paper which combines with a wealth of line features in organized objects and adopts rigorous data processing theory of photogrammetry espically the advanced technology in computer vision,focuses on the feature extraction and matching. We mainly undertook the following findings that demonstrated by our experiments with a good results: proposing a new method for points extraction by line discretization and a high precision line extraction algorithm; improving the image matching based on correlation coefficient called weighted correlation coefficient; bringing epipolar constraint into the least squares matching and creating a new way for line matching especially blocked.更多还原