Application Of Change Detection Based On Remote Sensing Data On Extracting The Slip Mass

In the Worldwide, especially in mountainous areas, the landslide which is one of the geological hazards that is only next to the earthquake can cause heavy casualties and economic losses. According to statistics, during the last 20 years in the twentieth century, the countries which are affected by the landslide hazard most severely, such as Italy, Japan, the United States and Russia, reached an average annual economic loss of 15-20 billion U.S. dollars. China is also a landslide-prone countries, only have 5·12 earthquake in Wenchuan triggering landslides had more than 50 thousand sites, including for the towns and villages they brought direct and indirect threats which are more than 4000 sites. In addition to casualties and economic losses, landslide often brings a huge ecological disaster. Given the grim situation of landslide and the resulting huge losses, the effective investigation and monitoring of landslide distribution and status, and prediction and evaluation of the slip mass can provide a scientific basis for the disaster mitigation and relief.In this paper, "5·12" Wenchuan Beichuan earthquake triggering landslide and the landslide of Wulong caused by improper mining are used as study objects. The CCD data from CBERS-02B satellite with 19.6 m spatial resolution, Radarsat-1 SAR images and the HJ-1B/CCD image are respectively used. Two kinds of change detection methods are used to extract the landslide. The main contents are as follows:(1) In this paper, multi-source remote sensing data are employed to investigate a fast and effective way to detect landslides, which block the rivers, and quickly locate those disasters. As a result, this detection is able to provide critical information to people who in charge of rescuing sufferers and mitigating the disasters. First, adopting CCD data acquired from CBERS-02B as pre-disaster image, this research extracts river channel information by applying NDWI method, which is based on the difference between green and NIR bands. Radarsat-1SAR data is hired as post-disaster image for the reason that optic images are unable to work in the condition of bad weather. Due to its particular characteristic, the post-disaster river channel range extracting is conducted by using threshold method after several trails. Second, it operates intersection of these two extracting results to detect the changing sections of the channel. The river channel information obtained by generating the above procedure is vague and discontinuity. And these unexpected phenomena are not all brought by the landslides blockage. Meanwhile it results from speckles, which come from the data themselves. Subsequently, this research generates expansion algorithm to the river channel intersection, so that to weaken the speckle effect, and to manifest the characteristic of landslides. The last step is to identify the exact range of landslides via human-computer interacting interpretation. Landslides disaster happened in Beichuan county of Sichuan Province are used in this research to validate the effectiveness and reliability of the above methodology. The experiment takes Beichuan in Sichuan Province as the example, proves the effectiveness of this method and provides a feasible method for extracting the landslide after the disaster.(2) In this part the large-scale landslide happening in Wulong, Chongqing is regarded as study object. Applying the pre-disaster and post-disaster multispectral data from HJ-1B satellite, the difference method is used for change detection of landslide. In difference image, the distinction between change and no-change in the area can be equivalent to a type of the division of background and objectives, whose difficulties focus on the selection of the threshold. The four categories of automatic selection threshold methods for change and no-change area are examined:cyclic segmentation algorithm, class variance automatic threshold method invariable moment automatic threshold method, optimum entropy. According to test results, pros and cons of various methods are analyzed for landslide extraction, and consequently, a kind of automatic change detection threshold is selected.(3) Comparing with other natural disasters such as floods, droughts, snowstorms, the scale of landslides is small, so remote sensing data based assessment of landslide disaster has its unique characteristics. In this paper, it employs some simple and convenience space analyzing model to assess the infrastructure damage, agricultural damage and suffered population according the characteristics of landslides reflecting from the remote sensing images.