面向对象的高分辨率遥感影像植被信息提取研究

发布时间：2018-01-02 08:07

本文关键词：面向对象的高分辨率遥感影像植被信息提取研究　出处：《吉林大学》2015年硕士论文　论文类型：学位论文

【摘要】：近些年来了解和掌握植被覆盖变化趋势已经成为了人们日益关注的焦点，不少研究人员做了大量与植被覆盖相关问题的研究，，从而对植被资源进行合理、有效、高效的管理，以便促进资源环境与社会经济协调发展。本论文主要针对于对植被信息提取的研究，目前对植被分类的研究比较薄弱，分类精度不高，本文选取朝鲜某地区作为研究区，采用面向对象的方法来研究植被分类的方法，从而提高植被分类的精度。本文利用GeoEye 1的高分辨率遥感影像，基于面向对象的影像分类方法，对研究区的植被信息进行提取，主要成果如下：（1）研究中首先根据研究区的地质特征和所采用的数据源，对数据进行几何校正和影像融合等预处理。研究中还针对影像融合技术进行了深入分析，提出了小波和IHS相结合的方式，提高了融合的影像的质量和精度，为后期植被信息提取打下基础。（2）本文通过多尺度分割实验，探讨了影像进行分割的尺度问题及其参数的选择问题，得到了基于影像不同特征的最优尺度。在确定了不同地物的最优尺度的基础上，利用面向对象的分类方法，分别针对不同地物的最优尺度，将高分辨率遥感影像大致分为林地，草地，建筑物，裸地，河流5类，并对其结果利用混淆矩阵进行精度评价，精度为90.14%，kappa系数为0.8514. （3）利用基于像元的遥感影像分类方法对同一区域的影像进行监督分类，非监督分类和决策树分类，进行精度评价；结果表明，面向对象的分类方法分类结果精度高于基于像元的分类方法，最后研究中利用最邻近分类法和模糊分类的方法分别对林地和草地进行了细分，实现了高分辨率遥感影像的植被信息提取。
[Abstract]:In recent years, understanding and mastering the trend of vegetation cover change has become the focus of increasing attention. Many researchers have done a lot of research on vegetation cover related issues, so that the vegetation resources are reasonable and effective. Efficient management, in order to promote the coordinated development of resources, environment and social economy. This paper mainly focuses on the study of vegetation information extraction, the current research on vegetation classification is relatively weak, classification accuracy is not high. In this paper, a certain area of North Korea is selected as the research area, and the method of vegetation classification is studied by using object-oriented method, so as to improve the accuracy of vegetation classification. In this paper, the high resolution remote sensing image of GeoEye 1 is used to extract the vegetation information from the study area based on the object oriented image classification method. The main results are as follows: Firstly, according to the geological characteristics of the study area and the data sources used, the data are preprocessed by geometric correction and image fusion. In the study, the image fusion technology is also deeply analyzed. The method of combining wavelet with IHS is proposed to improve the quality and accuracy of the fusion image and to lay a foundation for the extraction of vegetation information in the later stage. In this paper, the scale problem of image segmentation and the selection of its parameters are discussed through multi-scale segmentation experiments. The optimal scale based on different features of the image is obtained. On the basis of determining the optimal scale of different ground objects, the object oriented classification method is used to target the optimal scale of different ground objects. The high-resolution remote sensing images are divided into five categories: woodland, grassland, buildings, bare land and rivers. The accuracy of the results is evaluated by using the confusion matrix, and the accuracy is 90.14%. The kappa coefficient is 0.8514. Thirdly, the method of remote sensing image classification based on pixel is used to evaluate the accuracy of supervised classification, unsupervised classification and decision tree classification in the same region. The results show that the classification accuracy of the object-oriented classification method is higher than that of the pixel based classification method. In the end, the nearest neighbor classification method and fuzzy classification method are used to subdivide the forest land and grassland, respectively. The vegetation information extraction from high resolution remote sensing image is realized.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P237

【参考文献】