Abnormal Extraction Of Geochemical Data Based On Kalman Filter And SVM

Exploration geochemical prospecting is a kind of important means of mineralresources exploration, the data processing method of exploration geochemistry is oneof the key points to accurate determine the anomaly delineation. Previous explorationgeochemistry data processing method is always based on normal statistical method,this kind of method has been widely applied and obtained a large number ofachievements. But the conventional data processing methods have its inevitable flaws,if the geochemical data does not meet the linear features, the conventional dataprocessing methods will inevitably lead to the information suppression and lost.Mineralization is a complex process, the complexity of the primary environmentand secondary evolution of geochemical dispersion pattern show a complicatedfeatures. Geochemical spatial distribution of single element for the location of oredeposit will be partial, limited. Thus, elements combination features and thecomprehensive characteristics has become an important indicator of mineralprediction, the new technology and method of the abnormal exaction is particularlyimportant. In the past, the way to determine the elements or elements integrated isbased on the further study of metallogenic mechanism, and adopting the conventionalstatistical methods, such as factor analysis, principal component analysis. In recentyears, nonlinear characteristics of geochemical data has been taken seriously by theresearchers, nonlinear methods has been widely used, such as chaos method, fractalmethod and intelligent neural network method. Pay more attention to the role ofgeological metallogenic regularities, and accurately determine the comprehensiveanomalies can correctly identify mineralization information related to the ore-forming.Therefore, this article aims to provide a new effective method for geochemicalexploration.Data fusion method is one of research hotspots in the field of data analysis, inrecent years, its application in signal and image processing has achieved great success.Information fusion is a processing method of multi-order, multi-aspects andmulti-level for multi-sensor information, and getting new efficient information, thisinformation cannot obtained by any single sensor. Kalman filtering is one of thetypical representatives. It is a kind of Optimal Estimation Algorithm take for linear,unbiased and the minimum variance as the criterion. The method of data fusionmethod based on Kalman filtering for the geochemical data processing and thecomprehensive abnormal extracting has theoretical basis and feasibility.Support vector machine (SVM) are based on statistical learning theory, supportvector machine (SVM) take for structural risk minimization principle as the criterion. The method of support vector machine can effectively solve the over learningproblem of machine learning, and it has generalization. Besides, there must be aglobal optimal solution in theory, and cannot get into the local optimal. Theintroduction of kernel function reduce the dimension in high-dimensional spacewithout increasing the complexity of the algorithm, based on the theory of supportvector machine (SVM), decomposition the background and anomaly with the mind ofclassification.In brief, this article combining analysis the Kalman filter data fusion and theadvantage of support vector machine (SVM). The Kalman filtering data fusiontechnology is utilized to extract integrated geochemical anomaly information, and themethod of support vector machine (SVM) is used to distort segregating the fusion data,and delineating anomaly area. After the geochemical data analysis of Tibetautonomous region male village of copper and gold mine area, the results objectivelyand comprehensive reflect the actual mineralization conditions, comprehensiveanomalies are in conformity with the actual. The selection of threshold is reasonable,the delineated area is consistent with actual mining area. These results showed thatcombined the method of Kalman filter data fusion and support vector machine (SVM)to processing geochemical data and extracting the integrated anomaly is reliable andeffective.