Jiakui Tang

Aerosol Optical Thickness determination by exploiting the synergy of TERRA and AQUA MODIS (SYNTAM)

Aerosol retrieval over land remains a difficult task because the solar light reflected by the Earth-Atmospheric system mainly comes from the ground surface. Dark Dense Vegetation (DDV) for MODIS data has showed excellent competence at the aerosol distribution and properties retrieval, which is, however, restrictedly used for lower reflectance ground surface such as water body and dense vegetation. In this paper, we attempt to derive Aerosol Optical Thickness (AOT) by exploiting the synergy of TERRA and AQUA MODIS data (SYNTAM), which can be used for various ground surfaces, including high reflective surface. Preliminary validation result compared with AERONET data shows good accuracy and a promising potential

Preliminary Study on Unsupervised Classification of Remotely Sensed Images on the Grid

Grid is a new technology. With corresponding middleware it can give strong computing power. In this paper we mainly discuss the middleware technology and architecture used in remote sensing image classification algorithm. Because unsupervised classification middleware is the key of the classification middleware algorithms, we study the alternant-unsupervised middleware and put forward a non-alternant unsupervised middleware scheme. Based on this scheme, main factors which effect the performance of non-alternant unsupervised classification are analyzed.

Experience of Remote Sensing Information Modelling with Grid Computing

In this paper, we focused on the remote sensing information modeling and determination using Grid computing platform. We have underdone the experiments using remotely sensed images for thermal inertial modeling in Condor system that is one of the Grid Projects existed nowadays worldwide. We divided remote sensing data into several parts and run them on Condor pool and on one single machine. From these tests, the relationship among the work efficiency of image processing in Condor system and the number of the separated parts of image and the number of machines in this system is presented. Given a certain number of machines, a most efficient image size is existed among varies sized images. Besides, the possible causes of the longer put-off in this process are given, and some possible methods to resolve this problem are also presented. It is feasible to use Grid computing system such as Condor to process remote sensing data. And if the postpone problem can be resolved, the work efficiency of Grid systems will be high. Even with so many problems, it is a good thing that Grid systems do many things for you during all of us are in sleep. Our next major task will concentrate on realizing an arithmetic that can read and divide remote sensing images based on image size and number of machines in Grid system automatically, and transfer results back to the submitted machine as a whole data file.

Reconstruction of 3D Curvilinear Wireframe Model from 2D Orthographic Views

An approach for reconstructing wireframe models of curvilinear objects from three orthographic views is discussed. Our main stress is on the method of generating three-dimensional (3D) conic edges from two-dimensional (2D) projection conic curves, which is the pivotal work for reconstructing curvilinear objects from three orthographic views. In order to generate 3D conic edges, a five-point method is firstly utilized to obtain the algebraic representations of all 2D projection curves in each view, and then all algebraic forms are converted to the corresponding geometric forms analytically. Thus the locus of a 3D conic edge can be derived from the geometric forms of the relevant conic curves in three views. Finally, the wireframe model is created after eliminating all redundant elements generated in previous reconstruction process. The approach extends the range of objects to be reconstructed and imposes no restriction on the axis of the quadric surface.

Data-Parallel method for georeferencing of MODIS level 1b data using grid computing

Georeference is a basic function of remote sensing data processing. Geo-corrected remote sensing data is an important source data for Geographic Information Systems (GIS) and other location services. Large quantity remote sensing data were produced daily by satellites and other sensors. Georeferenceing of these data is time consumable and computationally intensive. To improve efficiency of processing, Grid technologies are applied. This paper focuses on the parallelization of the remote sensing data on a grid platform. According to the features of the algorithm, backwards-decomposition technique is applied to partition MODIS level 1B data. Firstly, partition the output array into evenly sized blocks using regular domain decomposition. Secondly, compute the geographical range of every block. Thirdly, find the GCPs triangulations contained in or intersect with the geographic range. Then extract block from original data in accordance with these triangulations. The extracted block is the data distributed to producer on Grid pool.

Java-Based grid service spread and implementation in remote sensing applications

Remote sensing applications often concern very large volumes of spatio-temporal data, the emerging Grid computing technologies bring an effective solution to this problem. The Open Grid Services Architecture (OGSA) treats Grid as the aggregate of Grid service, which is extension of Web Service. It defines standard mechanisms for creating, naming, and discovering transient Grid service instances; provides location transparency and multiple protocol bindings for service instances; and supports integration with underlying native platform facilities. It is not effective used in data-intensive computing such as remote sensing applications because its foundation, Web Service, is not efficient in scientific computing. How to increase the efficiency of the grid services for a scientific computing? This paper proposes a mechanism Grid service spread (GSS), which dynamically replant a Grid service from a Grid node to the others. We have more computers to provide the same function, so less time can be spent completing a problem than original Grid system. This paper also provides the solution how to adept the service duplicate for the destination node’s Grid environment; how each service duplicate communicates with each other; how to manage the lifecycle of services spread etc. The efficiency of this solution through a remote sensing application of NDVI computing is demonstrated. It shows that this method is more efficient for processing huge amount of remotely sensed data.

High throughput computing for spatial information processing (HIT-SIP) system on grid platform

For many remote sensing application projects, the quality of the research or the product is heavily dependent upon the quantity of computing cycles available. Middleware is software that connects two or more otherwise separate applications across the Internet or local area networks. In this paper, we present the High Throughput Computing Spatial Information Processing (HIT-SIP) System (Prototype), which is developed in Institute of Remote Sensing Applications, Chinese Academy of Sciences, China. Several middleware packages developed in the HIT-SIP system are demonstrated. Our experience shows that it is feasible that our grid computing testbed can be used to do remote sensing information analysis.

Feasibility Study of Geo-spatial Analysis Using Grid Computing

Spatial applications will gain high complexity as the volume of spatial data increases rapidly. A suitable data processing and computing infrastructure for spatial applications needs to be established. Over the past decade, grid has become a powerful computing environment for data intensive and computing intensive applications. In this paper, we tested and analyzed the feasibility of using Grid platform for spatial analysis functionalities in Geographic Information System (GIS). We found that spatial interpolation, buffers, and spatial query can be easily migrated to Grid platform. Polygon overlay and transformation could achieve better results on Grid platform. To do network analysis and spatial statistical analysis on Grid platform could be no significant improvement of performance. The most un-suitable spatial analysis on Grid platform is the spatial measurement.

Soil moisture retrieval from MODIS data in northern china plain using thermal inertia model (SoA-TI)

Soil moisture plays an important role in monitoring of drought and waterlogging. In this paper, a thermal inertia map was obtained from MODIS data on April 16, 2004 using SoA-TI model in Northern China Plain covering Latitude from 38.5°N to 40.5°N and Longitude from 115°E to 117.5°E. A soil moisture map was obtained based on the relationship between thermal inertia and soil moisture. And a distributive map of drought and waterlogging was obtained based on the relationship between ranks of drought and soil moisture. The drought map derived from MODIS data shows that most of the studied area was in a drought condition except the eastern littoral area where the soil moisture was in order. Generally speaking, the Northern China Plain was generally in a drought condition on April 16, 2004. The actual drought information obtained from the official website of Beijing Meteorological Bureau indicates that it was in a drought condition in Northern China Plain in the middle ten days of April. The result derived from MODIS data using SoA-TI model is in consistent with the actual situation. It indicates that it is an effective way to monitor regional drought and waterlogging by means of deriving soil moisture using SoA-TI model from MODIS data.

Data-oriented composite kernel-based support vector machine for image classification

One novel composite kernel based support vector machine (SVM), which is called DOCKSVM (Data Oriented Composite Kernel based Support Vector Machine) is proposed in the paper. SVM have been proved good potential in various studies, and tried to application for pattern classification problems such as text categorization, image classification, objects detection etc. Recently, more and more researches show that SVM is promising in remote sensing image classification. Unlike traditional SVM method, DOCKSVM could integrate the bio-geophysical character into final classification through the composite kernels, which lead to the accuracy improvement of classification results. Firstly method of DOCKSVM is described in detail, then the novel method according to information entropy of training data to evaluate the weighted value of kernels is proposed, finally, preliminary results of application to remote sensing image classification is given which show that its good potential tool for remote sensing image classification.