基于形态学的遥感影像房屋提取研究

发布时间：2018-03-21 08:35

本文选题：遥感影像　切入点：形态学边缘检测　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着社会的发展,遥感技术正渐渐步入我们的视野,这其中对于遥感影像中房屋建筑物的提取被视为遥感的重要课题,而伴随着遥感技术的成熟与进步,高分辨率的遥感影像为我们进一步的研究提供了很大的帮助。房屋作为城市的主体,它的提取在城市规划、土地探测等多方面领域具有重要的应用,如何高效、准确地将影像中房屋部分提取出来,对于城市规划、精确农业以及军事探测方面均有很大的影响与意义。本文基于上述研究背景以及实际应用,经过深入的研究与实验,提出了一种新的提取房屋地物的方法,主要进行了如下工作:(1)遥感影像的预处理。针对遥感影像包含的数据量大,受各方面因素影响带来的辐射失真、几何畸变以及多噪声等特点,对初始的遥感影像进行了包括图像融合、图像去噪以及图像增强等预处理工作,得到了一幅在辐射和几何上尽可能真实的遥感影像。(2)基于边缘检测的图像分割。将传统的边缘检测方法应用于我们的实验中,但由于其检测精度以及抗噪能力方面的不足,不能满足我们的实验需求。经过多次的研究与实验,选择数学形态学的方法对遥感影像的边缘进行检测,并将传统的形态学边缘检测方法予以改进,构造了一种结合多方位、多尺度结构元素的边缘检测算子,这种方法可以在提取边缘的同时更好的保留图像的原始信息,经过与多种边缘检测方法的实验结果进行对比,得到了明显优于其它边缘提取方法的实验结果。(3)基于多特征的房屋提取与分类。首先对遥感影像中房屋部分进行了特征分析,包括其几何特征、光谱特征、纹理特征等,针对这些特征,提出了一种将基于最大相关最小冗余(MRMR)的特征选择算法与模糊c均值(FCM)聚类相结合的方法,即根据这些特征采用MRMR特征选择方法进行相应的特征选择,并根据优选特征子集进行FCM聚类,最后对我们得到的分类结果图作进一步的优化处理。(4)设计对比分析实验。选取样本数据,采用不同的特征选择与分类方法对选取的遥感房屋影像进行提取实验,并利用混淆矩阵对不同方法的分类结果进行精度评价分析。分析结果表明本文所采用的方法可以更加有效的将房屋信息快速、精确的提取出来,可以作为遥感影像提取房屋的一种有效方法。
[Abstract]:With the development of society, remote sensing technology is gradually stepping into our field of vision, in which the extraction of buildings from remote sensing images is regarded as an important subject of remote sensing, and along with the maturity and progress of remote sensing technology, The high resolution remote sensing image provides a great help for our further research. As the main body of the city, the extraction of the house has important applications in many fields, such as urban planning, land exploration and so on. The accurate extraction of the house part of the image has great influence and significance on urban planning, precision agriculture and military exploration. Based on the above research background and practical application, this paper has been deeply studied and experimented. In this paper, a new method of extracting house features is proposed. The main work is to preprocess the remote sensing image as follows: 1) aiming at the large amount of data contained in the remote sensing image and the radiation distortion caused by various factors, The original remote sensing image is preprocessed by image fusion, image denoising and image enhancement. An image segmentation based on edge detection is obtained, which is as real as possible in radiation and geometry. The traditional edge detection method is applied to our experiments. However, due to the shortcomings of its detection accuracy and anti-noise ability, the image segmentation is based on edge detection. After many studies and experiments, we choose the mathematical morphology method to detect the edge of remote sensing image, and improve the traditional morphological edge detection method. Multi-scale structure element edge detection operator, this method can extract edges while better preserving the original information of the image, and compared with the experimental results of various edge detection methods. The experimental results, which are obviously better than other edge extraction methods, are obtained. Firstly, the feature analysis of houses in remote sensing image is carried out, including geometric features, spectral features, texture features and so on. Aiming at these features, a new feature selection algorithm based on maximum correlation minimum redundancy MRM (MRMRM) and fuzzy c-means (FCM) clustering is proposed. According to these features, the MRMR feature selection method is used to select the corresponding features. According to the optimal feature subset, the FCM clustering is carried out. Finally, a comparative analysis experiment is designed for the classification result map obtained by us, and the sample data are selected. Different feature selection and classification methods are used to extract the selected remote sensing house image. The results show that the method used in this paper can extract the house information more quickly and accurately. It can be used as an effective method to extract house from remote sensing image.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP751

【相似文献】