Monitoring And Automatic Extraction Of Land Use Change Information With Remote Sensing Approach

The remote sensing technology has been applied in land use monitoring for nearly twenty years in our country. The purpose of such monitoring method is twofold. One is to detect the dynamic variation of land use with the remotely sensed imagery. The other is to analyze the influencing factors of land-use changes, the dynamic process and its development trend as well. The various automatic detecting methods have the different sensitivity to the land-use change information. Besides, the monitoring precision depends on the signal noise ratio in the imagery and the method of image segmentation used. Therefore, the current monitoring methods is semi-automatic operation with human intervention, operator-dependent detecting results, and inefficient data processing. These are often unacceptable for investigating a very large region. In addition, since the final results derived are usually represented in some formats such as PSD data (Photoshop format), TIF data, Excel data and various documents, it is difficult to manage these data even with a geographic information system (GIS). On the other hand, the regional dynamic monitoring is of a long-term process, meaning that the derived data are typically characterized by the temporal feature. Therefore, such multi-temporal results should be represented with the appropriate formats that facilitate data management and information query with a GIS.This study aims to optimize the data processing flow, to develop data fusion technique, to improve monitoring precision, to decrease human intervention. The experimental site chosen is around Mianyang City (in Sichuan of China). The testing datasets used over this site include the multi-spectral images of Landsat TM, the panchromatic image of SPOT2, and both the multi-spectral and the panchromatic images of SPOT 5. The thesis thus concentrates on proposing and developing the methodologies and algorithms of automatically extracting land-use change information and of automatically segmenting the detected results. Furthermore, the archiving method of such derived change data is presented on the basis of a GIS environment. The final part of the thesis proposes a newmethod, based on a three dimensional virtual environment, to analyze the influencing factors of land use change and predict the development trend of land use. The contributions of the thesis are summarized as follows.Two different algorithms based on the two-dimensional Wavelet Transform (WT) are proposed to automatically detect the land-use change parts using two images acquired at different times. They are called "image fusion" and "Wavelet-coefficient (WC) differencing". The former retrieve change information by using the weighted average of the corresponding WCs of two images, whereas the latter retrieve change information by differencing the corresponding WCs of two images. Both methods are novel in the field of land-use change detection. The results derived with the two algorithms are evaluated by comparing with those derived with multi-band Principal Component Transform. The ground truth data is also used for the further assessment. These show that both the two algorithms can meet the requirement of such applications.Proposed are the two different methods for the specific-target segmentation on the basis of analyzing the features of the change image derived with the above methods. One is first to determine the dual thresholds by using the histograms of multiple small areas, and then to segment specific objects into different areas using the thresholds, the region-growing algorithm and the limiting conditions of gray differences around the object's boundary. The other is that the original gray image is first processed in Photoshop, i.e., rough segmentation, to minimize the disturbance of background noises. By combining gray information of objects with WCs, the edges of specific object such as change pattern, can be automatically extracted.Since various automatic detecting methods have different sensitivity to the land use change information, and the resulting precision is gover...