基于面向对象分类的高分辨率遥感影像变化检测研究

发布时间：2019-01-11 15:07

【摘要】：随着经济社会的发展,城市化进程的加快,人们对土地利用的广度和深度日益增强,土地利用结构发生了深刻的变化,耕地、水域等自然资源不断减少,而建设用地、道路等社会资源不断增加。因此,及时了解土地利用状况,对我国的经济建设和环境保护等方面具有重要意义。随着遥感和信息处理技术的发展,遥感技术正朝着高空间分辨率、高光谱分辨率和高时间分辨率的方向发展,而仍旧采用传统的变化检测方法会暴露出一些的弊端,因此探索新的方法对高空间分辨率遥感影像进行土地利用变化检测是目前研究的热点问题。文中针对变化检测中阈值确定问题进行研究,有效地避免了变化阈值确定的主观性和复杂性,同时又可以清楚地得到各类别的变化情况。采用两期高分一号影像,基于叠加分析的思想,对分类后的矢量图层进行叠加,得到包含变化信息的新对象,并根据叠加结果直接定义变化类型,通过模糊分类的方法对各个新对象进行分类,完成变化检测。在此研究中,首先需要将影像分割成不同的影像对象层;然后分析提取对象特征,建立分类规则,进行多层次的模糊分类;最后采用叠加分析的思想,完成变化检测。利用误差矩阵和各地类的一致性指标评价变化检测的结果,通过分析,本文的变化检测方法检测概率为88.39%,虚警率为11.61%,漏警率为8.04%,总的检测误差为19.65%;各地类的检测精度基本都达80%以上,只有少数类别精度较低,存在误差,有待进一步提高。总体来说,本文的变化检测方法避免了变化阈值确定的主观性和复杂性,有效地反映了区域的变化情况,为土地利用宏观监测提供有利依据。
[Abstract]:With the development of economy and society and the acceleration of urbanization, the extent and depth of land use has been increased day by day. The structure of land use has undergone profound changes, and the natural resources such as cultivated land, water area and so on have been reduced continuously, while the land for construction has been used. Social resources, such as roads, are increasing. Therefore, timely understanding of land use is of great significance to China's economic construction and environmental protection. With the development of remote sensing and information processing technology, remote sensing technology is developing towards high spatial resolution, high spectral resolution and high time resolution. Therefore, exploring new methods for land use change detection in high spatial resolution remote sensing images is a hot issue. In this paper, the problem of threshold determination in change detection is studied, which effectively avoids the subjectivity and complexity of the determination of change threshold, and at the same time, it can clearly obtain the variation of various categories. Based on the idea of superposition analysis, a new object containing change information is obtained by overlaying the classified vector layer with two periods of high score 1 image, and the change type is defined directly according to the superposition result. Each new object is classified by fuzzy classification method, and change detection is completed. In this study, we first need to divide the image into different image object layers; then analyze and extract the features of objects, establish classification rules, and carry out multi-level fuzzy classification; finally, use the idea of superposition analysis to complete change detection. Using error matrix and consistency index of different classes to evaluate the result of change detection, through analysis, the detection probability of change detection method in this paper is 88.399.The false alarm rate is 11.611.The false alarm rate is 8.04. The total detection error was 19.65; The detection accuracy of each class is above 80%, only a few classes have low precision and error, which need to be further improved. In general, the method of change detection in this paper avoids the subjectivity and complexity of determining the threshold of change, effectively reflects the regional changes, and provides a favorable basis for macro-monitoring of land use.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】