Study On Land Use Land Cover Change Detection Based On RS And GIS

Land cover and land use (LCLU) data provide important information for environmental management and planning. It is one of the most prominent characteristics in globe environment change, and not only limited by natural factor, but also affected by the factor of social, economics, technique and histories Remote sensing is one of the most important technologies to monitor changes that occur on the Earth surface (eg:landuse and land cover change analysis, nature disaster estimation, etc.). Remote sensing and geographic information systems (GIS) are rapidly developing technologies that offer new opportunities and potentially more effective methods for detecting and monitoring.Satellite remote sensing is an efficient way to obtain land-cover and change data in a timely and consistent manner. However, many concerns exist. First of all, digital land cover change detection relies on the assumption that land cover changes produce changes in the reflectance of the land surface. However, remotely sensed data are not error free. Reflectance of land cover is complicated by many factors such as differences in spatial resolutions, variations in the radiometric response of sensors, variations in solar irradiance and solar angles, variations in atmospheric condition, and differences in phenology (Yuan and Elvidge,1998). To diminish this problem, the remote sensing data used should be obtained from a sensor system that acquired data at approximately the same time of day to eliminate diurnal sun angle effects that can cause anomalous differences in the reflectance properties of the remotely sensed data, and the anniversary dates should be used to ensure general, seasonal agreement between dates of imagery. Secondly, selection of critical dates in the phonological cycle of the plants and a keen awareness of how their reflectance properties change through time are required (Jensen et al.,1997). Also, land-cover changes are not discrete since they lie on a scale that ranges from no alteration through major modifications. Studies of land-cover change using satellite remote sensing are often constrained to depict land-cover conversions only. Therefore, to help measure land cover modification or subpixel scale conversions, change detection algorithms of the future should incorporate some "fuzzy logic" (Jensen,1996; Foody,2001). Furthermore, change detection results are scale dependent because changes occur at all scales, and changes at local scales can have dramatic, cumulative impacts at broader scales (Walsh et al,1999). Finally, it is extremely difficult to implement a consistent, comprehensive, quantitative accuracy assessment, especially for a large area change database, due to the difficulties in acquiring an adequate database of historical reference data (Foody,2001). Unfortunately, there is no good solution for this issue, and it might be a major limitation of the digital change detection.Change detection is a process that analyzes a pair of remote sensing images acquired on the same geographical area at different times in order to identify changes that may have occurred between the considered acquisition dates.Automated land cover/landuse change detection from multitemporal satellite data is one of the most important challenges facing the remote sensing community. Its identification is difficult. Human expertise is often called upon to identify the areas where changes occurred. It is even more difficult in an urban environment because of its complexity. It is characterized by a great diversity in limited areas' land use. Furthermore, the elements that constitute this environment (for example, roads, buildings) differ from rural and natural environments due to higher repeatability and smaller sizes. To take all these characteristics into account and thus improve the change detection results, it is imperative to develop an approach that uses all the information available on the objects studied such as color, texture and shape, as well as contextual parameters such as the relation with the neighbourhood. Image matching is the primary work in the process of change detection between multi-temporal remotely sensed images. Many methods have developed both from manual and automatic view.In our research, we got spot image data in 2002 and 2005 of Huangpi city (Hubei, China) for first case study. We use Scale Invariant Feature Transformation (SIFT)algorithm to process our tast for change detection on land use land cover and it shows many advantages in the automatic methods. On the orther hand we could not identifying the type of change, quantifying the amount of change.And another way in our case study we propose to apply of Support vector machine (SVM) to remote sensing classification. we use the method of SVM to do the training on the map of land-use to identify the categories of land mass in remote sensing images. In the experiment, the basic Gray Level co-occurrence matrix (GLCM) is used in the SVM training,and the result is less effective. Secondly, Gabor Wavelets features are adopted, and the result is also less effective.Finally SIFT features are used, and the result is also poor. By analyzing these experiments it is found that our experiments are implemented based on single layer of remote sensing images and found that remote sensing images in 2002 and 2007 are flatten data, and the brightness of each mass are different, so there are many error in the data training,thus resulting in less effective.For second case of our study, we got Aerophoto image 1999 and Quickbird image data 2009 of Vientiane (Laos). Based on the analysis, aerophoto image and the remote sensing images data (quickbird) of Vientiane (Laos) in 1999 and 2009 are used as test data in the SIFT registration, change detection on land cuse land cover and also apply of SVM to remote sensing classification. Gray Level co-occurrence matrix, Gabor and SIFT feature are also adoted in these experiments. The result of second case study is better achieved. Overall,it is better to use multiple layer information then single layer information in the classification of images. While the gray level co-occurrence matrix, GABOR wavelet, SIFT characteristics have their own advantages in the classification,so it is hard to say which one is better than others. The whole experiment was done in C++ programming language, on the OPENCV platform, and with the online source code of SIFT registration algorithm.