Research Of Lithological Discrimination Methods To Geological Remote Sensing Based On Grid Technology And Intelligent Algorithms

The rapid process of mass data and requirement of orient applications need the support of powerful computation resource during the lithological discrimination of remote sensing, when the study area expands from the experimental points to the region, the existing PC computing power is far less than the requirement of time efficiency.Currently, most of lithological discrimination algorithms are based on statistical characteristics of multi-band intensity, these algorithms need to statistic bands before classification. With the increasing of the bands, algorithm processing speed will be slow. The defect induces that it is often applied to multi-spectral remote sensing, instead of high-resolution spectral remote sensing. The lithological characteristics of wave-based discrimination algorithms, in the discrimination are also inefficient due to slow calculation speed.At present, how to deal with the problem of slow processing speed and efficiency caused by the massive data or algorithm itself, most studies have focused on the optimization algorithm instead of optimization methods of high-performance calculation. This paper puts forward a distributed parallel computing environment by grid technique, and intelligent algorithms, and proposes a new discrimination method of lithology. Finally, a new service-based grid lithological discrimination model of remote sensing is proposed. The main contributions of this paper are as follows:(1) In order to resolve the problems of the remote sensing lithological discrimination, such as the massive calculation, long processing time and low efficiency, a grid-based lithological discrimination application model of remote sensing is proposed. The model supports distributed parallel computing and integration, supports service-oriented model in response to user application requirements and supports decomposition of calculation-intensive in the course of large-scale remote sensing lithological discrimination by step, process and data. The model makes up the deficiencies of the massive computing power lithological discrimination processing, and can meet the high demand of computing performance on each individual application.(2)In the traditional statistical discrimination of remote sensing lithology, limited samples are difficult to obtain good discrimination accuracy, so, the remote sensing lithology discrimination method is presented for the interest region based on support vector machines binary decision tree algorithm. Referring to the existing regional geological data, a certain number of universal, representative of the different types of rock samples are selected in the interest regions area. and then, various types of samples in the corresponding band Spectral reflectance of various rock types generated training samples, Participate in training and discrimination. The method eliminates the impact of discrimination accuracy because of insufficient sample.(3)Typical computational problems, such as consuming large amounts of memory due to high sample dimensions during large-scale training sets SVM discrimination, are overcome for the first time with the strategy of reducing large-scale SVM training samples based on PSO. In order to decrease the dimension of the discrimination computation, a training model is constructed based on the resulting samples. A fast, efficient and accurate discrimination is established in combination of using K-fold cross validation as the fitness evaluation mechanism. That gives an effective optimization method for extracting lithological information from large areas by using large-scale training sets.(4) To overcome the problem of the lithological discrimination of traditional over-reliance on training samples, and decreased effectiveness of discrimination results in human error or bias, a feasibility analysis on the parallel computation realization of the large-scale lithological discrimination based on PSO is performed. A discussion referring to realizing the multi-PSO lithological discrimination is made. The flow, result and evaluating method of discrimination are also given in this paper. The discrimination of rocks, combined with the computational efficiency of analysis and visual interpretation of the accuracy compared to verify the model using lithological discrimination, this discrimination method proposed based on multi-PSO and grid is an effective supplement of statistical discrimination methods, it can reduce the computation time and reduce man-made factors on the quality of discrimination.(5) In order to make use of idle PC resources, available remote sensing of geological data and lithological discrimination documents, avoid waste of resources and duplication, a service-based distributed parallel lithological discrimination framework is proposed, the grid computational-resource sharing mode based on the syntax and service quality combined extension is also presented. It achieves the efficient usage of idle resources, the extension of the grid service makes it easy to be found and be used by the grid users.In this paper, remote sensing lithological discrimination workflow is established based on the grid computation and the above-mentioned algorithms. Preprocessing remote sensing data, sample reduction and lithological discrimination are applied to several typical targeted areas. Through applying the united parallel processing to the Arkin metallogenetic zone, experiments results demonstrate the new method can save processing time, improve efficiency and discrimination accuracy. This method also provides the theoretical and practical base for the united parallel processing of lithological discrimination in the whole district.