卫星遥感图像地标特征提取与匹配研究

发布时间：2018-03-14 12:44

本文选题：地标　切入点：遥感图像匹配　出处：《国防科学技术大学》2015年硕士论文　论文类型：学位论文

【摘要】：地标匹配是通过图像匹配技术来确定地标模板图像在包含地标区域的遥感图像中的准确位置。匹配的精确性和快速性是图像匹配的两个关键因素,本文主要针对这两个方面展开研究,所做工作如下。1.提出了改进的归一化灰度组合矩阵相关算法(NIC算法)。针对NIC算法计算量大的问题,采用了设置相似门限的方法,通过设置子图像和模板图像灰度总和比值的上下门限,缩小了搜索区域,减少了匹配时间。仿真实验表明,改进NIC算法在确保相同匹配精度的同时匹配速度更快。2.提出了一种改进的基于Canny算子的遥感图像地标边缘检测算法。针对Canny算法高斯滤波参数设置困难、边缘点准确定位和噪声抑制相矛盾以及双阈值选取困难的问题,采用了自适应滤波方法,利用双正交梯度方法计算梯度,结合Otsu算法确定双阈值。仿真实验表明,改进Canny算法提高了抗噪声性能和边缘点定位精度,实现了双阈值选取自动化,能较完整的检测出地标边缘。3.提出了一种改进的基于距离变换的地标边缘匹配算法。针对传统距离变换算法运算量大的问题,采用了边缘距离扩展的方法,通过对二值边缘图像进行形态学膨胀得到最小距离值,简化了计算复杂度,加快了匹配速度。仿真实验表明,改进算法在确保匹配精度的条件下显著加快了匹配速度。
[Abstract]:Landmark matching is to determine the exact location of the landmark template image in remote sensing image contains landmarks in the region of the image matching technology. Accurate and fast, are the two key factors of image matching, this paper focuses on two aspects of work, such as the.1. algorithm proposed normalized intensity combination matrix improved (NIC algorithm). NIC algorithm for the large amount of calculation, by setting a similar threshold method, by setting the sub image and the template image gray sum ratio on the threshold, narrowing the search area, reduce the matching time. Simulation results show that the improved NIC algorithm in the matching accuracy at the same time, to ensure the same faster.2. proposed an improved remote sensing image landmark edge detection algorithm based on Canny operator. The difficulty of Canny algorithm based on Gauss filter parameters, edge point accuracy Localization and noise suppression in conflict and double threshold selection problem, the adaptive filtering method, gradient calculation using bi orthogonal gradient method, combined with the Otsu algorithm to determine the threshold. Simulation results show that the improved Canny algorithm can improve the performance of anti noise and edge positioning accuracy, realize the double threshold selection can be more complete automation. The detected landmark edge.3. presents an improved matching algorithm of landmark edge based on distance transform. In view of the traditional distance transform algorithm computational problems, by using the method of edge distance expansion, through the two image dilation to get the minimum distance value, simplify the computation complexity, speeds up the matching speed. Simulation results show that the improved algorithm ensures the matching accuracy under the condition of significant to speed up the matching.

【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP751

【相似文献】