遥感图像中建筑物提取与分析研究

【作者】 徐锋；

【导师】 夏德深；

【作者基本信息】 南京理工大学 ， 计算机应用， 2004， 硕士

【摘要】 建筑物识别技术是计算机模式识别领域逐渐兴起的研究课题，在军事、城市规划等领域有着广泛的应用前景。建筑物识别问题也是模式识别领域的一个相当困难的问题，要使这一技术成为完全成熟的技术还有许多工作需要去做。本文结合基于线状特征的建筑物识别算法和典型建筑物变化检测方法，对其中的部分问题分别进行了探讨，并给出了相应的解决方案。 本文工作包括： (1)、基于线状特征和感知组织理论的建筑物提取和识别算法 本文分析了建筑物成像的几何模型和空间关系，在基于感知组织原理和线状特征的建筑物提取识别算法的基础上，提出改进的基于线状特征分析的建筑物提取算法。本文通过改进线段预处理过程、线段搜索依赖规则和建筑物验证方法来改善原算法，最后进行实验证明。 (2)、基于纹理特征和基于图像匹配的建筑物变化检测算法比较 本文介绍基于纹理特征的建筑物变化检测方法及实现过程，并分析该方法存在纹理特征选择困难，算法复杂等缺陷，因而提出采用基于时序图像相关性匹配的变化检测方法。通过实验结果相比较，得出结论：由于基于相关性的匹配方法有效地利用了同一地区地物间的相关信息，降低了识别错误率，提高了定位精度，取得优于基于纹理变化方法的效果。更多还原

【Abstract】 The technology of building recognition is an active subject in the area of pattern recognition. In this paper, one of building recognition technology is probed based on linear features of the targets, and then change detection of the buildings is also discussed. The corresponding solutions are given. The work including:(1) The algorithm of building recognition based on linear features and the Theory of Perceptual OrganizationBased on linear features of the images, this paper first introduced the Perceptual Organization -Based building recognition algorithm. We emphasized the construction of the hypotheses of the buildings, which are comprised of edge lines. Considering the recognition performance and the computation time, this paper proposed an improved method using to pre-treat disorder edge lines, to search hypotheses of targets based on knowledge and to validate the hypotheses on regions of shadow.(2) Comparison of Change Detection algorithms of Building: based on image matching and based on texture.This paper introduced two kinds of change detection algorithm of building: based on image matching and based on texture. Firstly, this paper introduced the method based on texture, and then pointed out lack of precision of location and too much computation time. Using image matching base on relativity, the result proved this algorithm is better than the one based on texture.更多还原