Search

We combine k-means cluster algorithm with the statistical model of SAR(synthetic aperture radar) images and develop the probabilistic water body mapping algorithm based on the priori probability estimation. Firstly, we make the statistical model assumption about backscatter values based on Bayesian theory. Then, we classify the images based on cluster algorithm, calculate the prior probability of the water body mapping and estimate the parameters of the statistical model of water distribution.Thewater body probabilistic maps based on GF-3 images in Luquan and Xianning are calculated and then validated with GF-1 images. The algorithm is effective on high-precision probabilistic water body mapping of SAR images.

With the development of remote sensing technology, especially the improvement of sensor resolution, the amount of image data is increasing. This puts forward higher requirements to manage huge amount of data efficiently and intelligently. And how to access massive remote sensing data with efficiency and smartness becomes an increasingly popular topic. In this paper, against current development status of Spatial Data Management System, we proposed a self-adaptive strategy for image blocking and a method for LoD(level of detail)model construction that adapts, with the combination of database storage, network transmission and the hardware of the client. Confirmed by experiments, this imagery management mechanism can achieve intelligent and efficient storage and access in a variety of different conditions of database, network and client. This study provides a feasible idea and method for efficient image data management, contributing to the efficient access and management for remote sensing image data which are based on database technology under network environment of C/S architecture.

A topology adaptive snake (T-Snake) model based on orthogonal grids is introduced and improved in this paper, and a proper energy function is designed. A detection and handling mechanism for topological conflict that caused by island shaped hollow is proposed in the model, and therefore accurate extraction for complex boundary of river containing river islands is achieved. For the disadvantage of the need to manually construct the initial contour in the orthogonal T-Snake model, using the minimum fractal dimension to obtain one area of the water and automatically generate an initial contour. The experiment shows that the algorithm of this paper can accurately extract the boundary of the complex water which contains deeply concave regions or river islands, and it has higher accuracy and less time cost than classic Snake model and GVF Snake model.

SOLAP (Spatial On-Line Analytical Processing) has been applied to multi-dimensional analysis of remote sensing data recently. However, its computation performance faces a considerable challenge from the large-scale dataset. A geo-raster cube model extended by Map-Reduce is proposed, which refers to the application of Map-Reduce (a data-intensive computing paradigm) in the OLAP field. In this model, the existing methods are modified to adapt to distributed environment based on the multi-level raster tiles. Then the multi-dimensional map algebra is introduced to decompose the SOLAP computation into multiple distributed parallel map algebra functions on tiles under the support of Map-Reduce. The drought monitoring by remote sensing data is employed as a case study to illustrate the model construction and application. The prototype is also implemented, and the performance testing shows the efficiency and scalability of this model.