基于TM遥感影像的河套灌区土地利用变化分析与实现

发布时间：2018-05-09 16:59

本文选题：遥感 + 土地利用变化　；参考：《西北农林科技大学》2017年硕士论文

【摘要】：河套灌区作为我国三个特大型灌区之一,地处黄河中游,同时也是亚洲最大的自流引水灌区,其作用和地位举足轻重。为了准确掌握河套灌区在1990年至2010年间的土地利用变化情况,本文以TM遥感影像为数据源对1990年、2000年和2010年河套灌区的土地利用情况进行分类和统计分析,从而为河套灌区的土地资源管理和灌区用水合理分配提供科学依据。主要研究内容如下:(1)研究河套灌区TM遥感图像的拼接和裁剪方法,并对拼接后的全景图像进行目视解译并建立灌区土地分类系统。三期遥感影像均分别包含三景图像,每一景图像分别包括7个波段,根据TM遥感图像所含信息量较大的特点和河套灌区面积较大的实际情况,故本文选用具有旋转、尺度缩放不变性的SURF算法对图像进行特征点提取,采用基于Shearlet变换算法对配准之后的图像进行融合,拼接后获得完整的河套灌区全貌,借助河套灌区的矢量文件采用多边形不规则裁剪算法进行图像裁剪提取出研究区域。最佳指数法简单高效、方便计算,故本文采用该方法分别获得每期影像的最佳波段组合。(2)分别采用三种监督分类方法包括最大似然法、神经网络法和支持向量机法进行土地分类,后将分层分类和支持向量机法相结合改进分类方法。支持向量机分类是用以研究小样本情况下机器分类的方法,在实际的遥感影像分类应用中表现出较好的适用性。同时传统的单分类器对不同的地物类别识别准确率不尽相同没有针对性,而分层分类正是针对不同地物采用不同的分类策略。故本文采用支持向量机和分层分类相结合的策略进行土地分类方法改进。结果证明支持向量机相比于前两者分类效果较好,同时,将支持向量机和分层分类相结合,充分发挥各自优势,分类效果显著提高。1990年分层分类得到的总体分类精度是86.10%,Kappa系数是0.81。2000年分层分类得到的总体分类精度是93.66%,Kappa系数是0.91。2010年分层分类得到的总体精度是92.80%,Kappa系数是0.90。(3)对河套灌区土地分类结果数据借助土地利用动态度进行统计分析。为了获得定量的河套灌区从1990年至2000年的土地利用变化统计信息,土地利用动态度定量地描述了土地利用的变化速度,对预测未来土地利用变化趋势有积极的作用。故本文采用该方法获得了河套灌区从1990年至2010年土地利用变化统计结果,1990年至2000年的总体土地利用动态度是2.08%,2000年至2010年的总体土地利用动态度是2.83%,而1990年至2010年的总体土地利用动态度是1.13%。综合这二十年间的变化情况,变化速率最快的是居民地,达13.80%,而变化速率最慢的是耕地,达0.97%。
[Abstract]:As one of the three super large irrigation areas in China, Hetao Irrigation District is located in the middle reaches of the Yellow River and is also the largest self-flowing irrigation area in Asia. In order to accurately understand the land use change in Hetao Irrigation area from 1990 to 2010, the land use of Hetao Irrigation area in 1990, 2000 and 2010 was classified and statistically analyzed by using TM remote sensing image as the data source. It provides scientific basis for land resource management and rational allocation of water in Hetao Irrigation District. The main contents of this paper are as follows: (1) the paper studies the mosaic and tailoring of TM remote sensing images in Hetao Irrigation area, and sets up a land classification system for irrigation areas by visual interpretation of the stitched panoramic images. The three stages of remote sensing images are composed of three images, each of which includes seven bands. According to the characteristics of the large amount of information in TM remote sensing images and the large area of Hetao irrigation area, this paper chooses rotation. The scale scaling invariant SURF algorithm is used to extract the feature points of the image, and the image after registration is fused based on the Shearlet transform algorithm, and the complete panorama of Hetao irrigation area is obtained after stitching. With the help of vector files in Hetao Irrigation area, the irregular polygon clipping algorithm is used to extract the research area. The best exponent method is simple and efficient, so this paper uses this method to obtain the best band combination of each image, respectively) and uses three supervised classification methods, including the maximum likelihood method, respectively. The neural network method and the support vector machine method are used to classify the land, and then the hierarchical classification method and the support vector machine method are combined to improve the classification method. Support vector machine (SVM) classification is a method used to study machine classification in the case of small samples, which has shown good applicability in the application of remote sensing image classification. At the same time, the traditional single classifier does not have the same accuracy for different ground objects classification, and the hierarchical classification adopts different classification strategies for different ground objects. Therefore, the strategy of combining support vector machine and hierarchical classification is used to improve the land classification method. The results show that the support vector machine is more effective than the former two classification methods. At the same time, support vector machine and hierarchical classification are combined to give full play to their respective advantages. The overall classification accuracy obtained by stratified classification in 1990 is 86.10kappa coefficient is 0.81.The overall classification accuracy obtained by stratified classification in 2000 is 93.666kappa coefficient is 0.91.The overall accuracy of stratified classification in 2010 is 92.80kappa coefficient 0.90.0.The Kappa coefficient of stratified classification is 0.90.0.The Kappa coefficient of stratified classification in 1990 is 0.90.The Kappa coefficient of stratified classification is 0.90.93. Land classification data of irrigation district were analyzed by land use dynamic degree. In order to obtain quantitative statistical information of land use change from 1990 to 2000 in Hetao Irrigation area, the land use dynamic degree describes quantitatively the change rate of land use, which plays an active role in predicting the trend of land use change in the future. Therefore, the statistical results of land use change from 1990 to 2010 in Hetao Irrigation District were obtained by using this method. The total land use dynamic attitude from 1990 to 2000 was 2.08, the total land use dynamic attitude from 2000 to 2010 was 2.83, and that from 1990 to 2010 was 2.83. The overall attitude of land use is 1.13%. Synthesizing the change of these two decades, the fastest change rate is the resident land (13.80%), while the slowest change rate is cultivated land (0.97%).
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S127;F301.2

【参考文献】