【作者】 肖迪；

【导师】 王淑云； 邵承会；

【作者基本信息】 吉林大学 ， 计算数学， 2013， 硕士

【摘要】 异物识别与定位是图形图像处理中的一个比较重要的研究内容,本文结合军工项目”机场飞机跑道外来异物识别与警示系统开发研究”,通过对成熟的图像匹配技术的学习和理解,针对给定的机场跑道在不同成像环境下所成的图像中杂物进行识别和定位。以此来确定杂物的位置,排除其对机场正常运行的影响。完成的具体内容如下：1.在实验过程中,根据实时需求,分析图像特征,对所研究图像有一个良好的认知,并从中获知图像的基本特性。2.根据传统的图像匹配技术理论,针对机场跑道中模板图像与实时图像进行配准的问题,提出了一种基于多临近点距离和基于图像纹理相结合的匹配算法,该算法充分利用了两种算法的优势,解决了机场跑道图像因环境等外在因素影响而难以进行配准的问题。同时,对传统的图像匹配算法予以分类并进行简单的描述和对比实验。3.提出一种粗细网格分割的方法进行匹配区域杂物位置的确定：对截取模板和匹配的区域,连续两次对图像进行分割,分别比较每次分割之后相对应小块之间的相似度。其中第一次分割之后相应小块之间的相似性度量采用了标准化的协方差,找出其值满足小于事先设定阈值的小块,对杂物的是否出现进行初步的判定。第二次分割之后则采用了对应灰度差值的办法来精确的判定杂物的出现位置并计算杂物所在区域的面积。4.提出一种自动阈值二值化的方法：利用计算和统计图像直方图的信息,自动选取最佳分割阈值,避免了传统二值化方法阈值难以事先估计的问题。对二值化后的图像进行相减,再做腐蚀和膨胀操作,确定出杂物的位置和面积。实验结果表明,基于多临近点距离和基于图像纹理相结合的匹配算法适用于本文所研究的图像范围,包括对图像的位移、光照变化、以及一定的几何失真都有良好的适应性。更多还原

【Abstract】 Foreign object recognition and positioning is an important content of the graphics pro-cessing.This article by learning and understanding of a mature image matching technologyto identify and locate the image into the imaging environment for a given airport runwaydebris§in order to determine the location of the debris, eliminate its impact on the normaloperation of the airport.The completion of the specific content is as follows:1.During the experiment, based on the real-time requirements, the analysis of imagefeatures,research images have a good understanding of the basic characteristics of the imagefrom informed.2.According to the traditional theory of image matching technology, for the airport run-way in the template image and the real-time image registration problem,propose a basedTemplate-matching method based on image texture and MCD combination matching al-gorithm.This algorithm takes advantage of the advantages of the two algorithms to solvethe problem of the airport runway image registration difcult environment outside influ-ences,meanwhile, the traditional image matching algorithms to be classified and simple de-scription and comparative experiments.3.Propose a thickness of the grid segmentation method for matching determination ofthe position of the region debris:small piece of the similarity between the corresponding in-terception templates and matching area twice the image segmentation, and compare eachsplit.Wherein the first split using a standardized measure of the similarity between the cor-responding small piece covariance,find the values satisfy the small blocks is less than the previously set threshold value,debris whether there preliminary determination.After the sec-ond split take the corresponding grayscale diference approach to precisely determine theemergence of debris and calculate debris Area area.4.Propose an automatic threshold binarized:Improved image binarization method,Co-mputational and statistical image histogram automatically select the best segmentation thresh-old,avoid the problem that threshold is difcult to estimate in advance of,subtraction imageafter binarization,do erosion and dilation operations,determine the location and size of thedebris.更多还原