| Since the beginning of the21st century, a series of large scale natural disasters have broken out in many countries around the world, which have caused huge casualties and economic losses. At the same time, with the rapid development of today’s spatial information technology, especially the rapid development of remote sensing sensor technology with high temporal and spatial resolution, the role of spatial information technology in disaster prevention and mitigation has become more and more prominent.Significant elevation changes are likely to occur in the process of geological disasters growth and outbreak. Even though experts and scholars have made numbers of important achievements in geological disaster monitoring, prevention and evaluation, there has been no a completely convenient and effective ways for real-time3D deformation detection of geological disasters so far. Therefore, such detection is extremely urgent for geological disaster prevention and disaster management.On the basis of the existing change detection theories and methods, this paper explores theoretical methods for the detection of3D changes, and uses these methods to test the results of3D deformation about geological disasters in Yingxiu region. The main contents are as follows:1) An algorithm of automatic matching of high-precision images. Using strong robustness of the SIFT algorithm for image matching ensures the effective matching on the conditions that zooming, rotation, translation, perspective transform or light transform happen between images. RANSAC algorithm is used to gradually eliminate the mismatching points. Finally Least Squares Method is used to improve the image accuracy of image matching from pixels to sub-pixels. To test the method, satellite images of Wenchuan county are used. Two images are used to create many eigenvectors of the corresponding image points. After removing the mismatching points and Least Squares Image Matching, automatic matching of high-precision image is successfully achieved.2) The integration of remote sensing imaging models on condition of the generalized stereopair. This part explores the generalized tridimensional imaging mathematical model on the condition that the image point, Perspective Center and object point are collinear. In order to implement the Image Location on the condition of the generalized stereopair, there integrating the Rational Function Model which is more extensive and more applied on the basis of strict space intersection of collinearity equations, an integrated model is build which can take advantage of both line array and planar array image to positioning of the spatial3D information. This generalized model that can adapt to the positioning requirements of multi-source data and fully embodies the connotation of the"generalized".3) The synergy of3D image registration and3D change detection supported by GIS data. Using the GIS data (mainly include digital orthophoto and DEM) which come from old period as a control,This part builds a mapping equation between the gray value function g (x, y) in new image and Corresponding Point gray value function G (X, Y) in old orthophoto and interpolating Z value function from DEM information as the elevation change, combined with3D information fom the generalized stereopair, so3D alignment of the two periods can be achieved. In the process of registration, when the coordinate information of the feature points does not meet the registration model,they will be found and given the appropriate weights and the deformational region will be highlighted gradually in iterative process. The coordinate information of undeformed area will meet the registration model and be given the weights of0. By this way we can fulfill synergy of3D image registration and3D change detection. At the same time, using the Selecting Weight Iteration Method to detect gross errors, gross Errors which impact the test results greatly will be eliminated during the change detection process.4) The detection on3D deformation of Yingxiu landslide disasters and then management of disasters. From the3D detection in Yingxiu, we can find that the deformation zones are mainly located on the sides of the downstream valleys and on mountains with higher elevation. Spatial analysis of landslide disasters based on GIS:To be specific, landslide deformation which concentrated on the typical "V" shaped valley, they are widely distributed and in large-scale, The reasons for its formation is possibly related to valley landforms. The other landslide deformation happen on the higher mountain,some of them are on the southeast and west upland in the region and there is a landslide in the northeast corner,two large parts with cyclic deformation exist in east and west region,the overall distribution of this deformation is more scattered, the reasons for its formation is possibly related to large slope gradient and altitude difference in mountainous region. Based on the result of detection,some suggestions regarding disaster management are advanced according to the different situation of landslide in different regions.The significance of this research is as follows: On one hand, extracting and detecting three dimensional changes from remote sensing data and auxiliary GIS data, which can help analyze changes before and after disasters more accurately, is an important method of disaster monitoring and assessment and therefore has been a heated topic of remote sensing information science and geographical information science. On the other hand, the speed of extracting spatial information currently still falls behind that of ground object change; three dimensional change information is necessary to update GIS data and improve the timeliness of geographical information, which can remedy the current deficiency in three dimensional information extraction and therefore also very important in applicable study in other related subjects.The main characteristics and innovations of this article has two aspects: First, the integration for strict geometric image model under the conditions of the generalized stereopair of the linear array and planar array sensor of multi-source; the second is to explore the synergy of3D image registration and3D change detection of high-precision、fully automatic level. |
PDF Full Text Request