| Detecting the changes of the earth surface cover by utilizing the repetitive observational ability of satellites is an important aspect in the research of remote sensing field, which has attracted much attention from the very beginning of the development of remote sensing. The spatio-temporal changes of the environment, the interaction of the four spheres in the geosystem as well as man's activities all contribute to the changes of the landscape on the earth surface and its utilizing forms, which in turn, will have profound influence on the earth's resources and environment. Therefore, how to monitor changes of the earth surface and update relevant geographical information systems timely and efficiently become urgent need for the detection techniques of imagery changes in remote sensing. Currently remote sensing techniques have stepped into a new stage of quick and timely supply of diversiform earth observation data. The huge volume of daily updated data provide rich data source for change detection implementation. The daily enriched and improved foundation and thematic spatial information products provide new contents for change detection research of remote sensing imagery. However, many problems need resolving if we want to find the changes on the earth surface anywhere and anytime. As a result, the change detection of remote sensing imagery has become a focus in the research of remote sensing in recent years. Relevant research work is undoubtedly of great theoretical and practical significances.Based on the analysis and summarizations of researches home and aboard, the dissertation tries to analyze and systematize various ideas and concepts in change detection of remote sensing imagery. Some basic concepts about change detection of remote sensing imagery are given, basic types of change detection are analyzed, and processing models and steps are summarized. Besides, based on the diversity of spatial information, the conventional concepts about change detection of remote sensing imagery are extended to include all types of change detection of spatial information. Mathematical models and general processing steps are summarized andimproved accordingly.In the dissertation, existing detection methods are reviewed, basic principles underlying commonly used methods are expounded, and their virtues and shortcomings are analyzed respectively. From the comparison we can draw the conclusion that there is no one method that fits all circumstances. In practice, which method to choose depends on our purposes and data.Implementing change detection analysis by integrating remote sensing and GIS is one of the developmental trends. Following this trend, a detection method based on the integration of RS & GIS in a lower level is put forward in the dissertation. It adopts a self-adapting determination algorithm for threshold based on polygon area filling rate. The method is tested on the CBERS-2 satellite images of a district in Beijing and the precision of the results is as high as 80%. Under this idea we developed a change detection system for Beijing, which has proved applicable in operation for more than one year. It satisfies all requirements in the project and provides a new choice in change detection methods.Geometry registration errors are one of the main factors affecting the precision of change detection results, therefore studying the influence of geometry registration errors to the results of change detection quantitatively is of importance to the comprehension and improvement of the reliability of detection results. By introducing geostatistics and considering remote sensing images as regionalized variables, we analyze their randomicity and structure. With the tools of semivariogram and semivariogram graph, we propose a set of methods to analyze the influence of geometry registration errors on the change detection results quantitatively. By classifying the types of the pseudo-change caused by the geometric registration errors of imagery, we bring forward a method to differentiate various types of pseudo changes. For our purposes, we choose 5 typical regions to test on the commonly used data source, the images of Landsat, SPOT and QUIICKBIRD, and it proves that the change detection results of Landsat band 5, SPOT band 3, and QUICKBIRD band 4 are more sensitive to geometric registration of imagery; the precision of change detection in rich-textured areas is more susceptive to the precision of geometricregistration of imagery; the precision of geometric registration of imagery should be higher than 0.22 pixel if a result with a precision of change detection as high as 90% is to be obtained.The concept of change detection with the integration of RS & GIS is expatiated, integration methods are classified, and conceptual framework is given. The virtues of different integration methods and their concerning key knots are analyzed, which may pave the way for future studies.Based on the idea of RS-GIS holistic solution of change detection, the dissertation gives the definition, conception, and framework of holistic solution and presents preprocessing methods for remote sensing images and GIS data. The software of the digital raster graphics (DRG) with a high precision is produced. An image preprocessing method based on wavelet transformation is proposed, making it possible for the processed images to smooth homogenous areas and removes random noises while enhancing edges. The preprocessed results are helpful for the holistic solution of change detection. The DRG software developed according to the method is now widely used in production sectors concerning spatial information such as sectors of surveying & mapping, electric design, urban planning, forestry, etc and proved efficient and applicable.The dissertation proposes a RS-GIS holistic iterative method based on areal features and its flow to solve change detection. It studies involved key techniques such as data preprocessing, feature extraction, feature description and searching strategy. It presents a label point determination algorithm for polygon based on mathematical morphology. It defines a set of descriptive metrics for polygon features from coarseness to precise. With the similarity metrics, it proposes an iterative extraction method enlightened by GIS knowledge for areal features. Furthermore, it designs a set of hierarchical searching strategy. With the method of holistic iterative solution, we developed software, which was tested on the data of a district in Shanghai. The method proved to be efficient and applicable as the precision of detection reached 83.33%.When the changes between remote sensing imagery and GIS data are not obvious,the holistic iterative method can calculate a large number of matching control points, which, in combination with the geometry rectification models of remote sensing imagery, can be used in the automatic geometry rectification of remote sensing imagery referred to GIS data. The results show that the accuracy of geometry rectification could reach 0.5 pixel or less, thus the advantages of this geometry rectification method are evident compared with the traditional one.
|
PDF Full Text Request