Research On Information Extraction Methods For Mineral Remote Sensing Image

Information extraction of mineral remote sensing image is an important application of remote sensing technology in geological exploration.Researching on information extraction methods based on mineral remote sensing image can provide the basis for the study of regional metallogenic prognosis and speed up the evaluation of mineral resources exploration,which is beneficial to promote the healthy and stable development of local mining economy.Based on the previous work,several key technologies involved in the information extraction based on mineral remote sensing image are deeply studied,which include image enhancement,edge detection,matching,clustering and mineralization alteration information extraction.The main work is as follows:Firstly,a method of mineral remote sensing image enhancement in non-subsampled shearlet transform(NSST)domain based on multi-stages particle swarm optimization(MSPSO)is proposed.The image to be enhanced is decomposed into a low-frequency sub-band and several high-frequency sub-bands through NSST.The coefficients of high-frequency sub-bands are enhanced according to adaptive Bayesian threshold method and nonlinear transform,while that of the low-frequency sub-band is processed by using the fuzzy enhancement method with its fuzzy parameters optimized by MSPSO algorithm.Experimental results show that compared with such four enhancement methods as bidirectional histogram equalization method,stationary wavelet transform method,non-subsampled contourlet transform(NSCT)adaptive threshold method and artificial bee colony(ABC)optimization method in NSCT domain,the proposed method can effectively improve the contrast and definition of remote sensing image and enhance edge details with the best visual effect.Then,a mineral remote sensing image edge detection method based on L0 gradient minimization model and NSST is studied.L0 gradient minimization model which has edge preservation ability is applied to smooth the input image and thus highlight the edge information in the image.Then,the Canny operator is applied to obtain the preliminary detection result of the image.The non-subsampled shearlet transform(NSST)is applied for image decomposition to obtain the high-frequency component of the image while the abundant detail information in every direction is extracted.The preliminary result and the high-frequency detection result are combined to obtain the final edge.Experimental results show that compared with Canny method,NSCT modulus maximum method and NSST scale product method,the studied method is able to detect more complete and clear edges with accurate edge localization.And then,a mineral remote sensing image matching method based on improved features from accelerated segment test(FAST)and speed up robust features(SURF)is discussed.Firstly,the scale space of images is constructed and the Gaussian pyramids are constructed in the standard image and the image to be matched,respectively.Then the FAST algorithm is applied to extract the preliminary feature points in each layer of the pyramids.And a feature point set is obtained through sifting from the preliminary feature points by Harris corner detection.Next,the SURF descriptor of each feature point is calculated,and the coarse matching is carried on based on the nearest neighbor search strategy.Finally,the mismatching point pairs are eliminated by using random sample consensus(RANSAC)algorithm to get the final matching result of images.A large number of experimental results show that,compared with SURF method,the recently present methods based on Harris and SURF or based on FAST and SURF,the discussed method can improve the correct matching rate as well as the matching accuracy,and exhibits better adaptability to brightness changes,scale changes and noise.Subsequently,a mineral remote sensing image clustering method based on improved iterative selforganizing data analysis(ISODATA)and fast search and find of density peaks is given.The fast search and find of density peaks algorithm is applied to find the initial clustering centers of the image to be clustered.Subsequently,the image to be clustered is transformed from RGB color space to Lab color space.Meanwhile,the neighborhood information of pixels is introduced to improve the clustering criterion of ISODATA method.The improved ISODATA clustering criterion is applied in the Lab color space to obtain the final clustering result.Experimental results show that compared with such three clustering methods as ISODATA method,improved fuzzy ISODATA method and the method based on ISODATA and particle swarm optimization(PSO),the given method is able to accomplish accurate and fast clustering for remote sensing image while exhibiting better performance in terms of anti-noise ability.Finally,a remote sensing mineralization alteration information extraction method based on principal component analysis(PCA)and support vector machine(SVM)optimized by cuckoo algorithm is proposed.The mineralization alteration information in the remote sensing image of study area is enhanced by band ratio method and the ratio images are obtained.Then PCA is applied to the ratio images and the hydroxyl and iron staining principal components are selected and the training samples are extracted.Subsequently,the the training samples are trained by SVM while cuckoo algorithm is used to find the optimal kernel parameter and penalty factor of SVM and thus the optimal SVM model is determined.The optimal SVM model is used to accomplish the extraction of remote sensing mineralization alteration information.Wulonggou area in Qinghai province is selected as the study area while the hydroxyl alteration information and iron alteration information are extracted.Experimental results show that compared with three methods proposed recently,the proposed method can obtain the highest matching degree with the best extraction effect.