The Extraction Methods Research Of Remote Sensing Image Linear Information

Linear information is on the remote sensing image from the form of informatioonline, a straight line or arc. Linear information for government decision-making andcity planning, feature recognition and geological structure interpretation is of greatsignificance.Traditional linear information extraction methods are mainlyconcentrated in the transformation of image spatial domain, such as the houghtransform, this thesis mainly from frequency domain to study the linear informationextraction of remote sensing image.Linear texture information is on the remote sensing image texture information, ina straight line or arc shape is one of linear remote sensing information.There are kindsof methoeds to extract the texture information, such as statistics, institution and model.In this paper by wavelet method to extract the different decomposition levels,differentresolutions, the directional texture information of the frequency domain.At the sametime, considering the texture segmentation image gray-level threshold segmentation,set different gray level in the program of texture segmentation threshold egmentation,get a different threshold extraction of texture image, and analysis the result.The causes of the linear edge information is changed step change or roof ofpixels, Mainly through the classical edge detection operator template and convolutionoperation to extract the edge of image,because the limitation of the algorithmitself,and the set of gradient single threshold segmentation, the edge of theinformation there will be a signal-to-noise ratio of contradictions.If better to eliminatethe noise, extract the edge of less information,and if possible to extract the edge of theaccurate information, extract the false edge contain noise.Base on the reason, Thispaper aimed at the median filter for denoising is not ideal,using wavelet denoising,and then with different resolution, different levels of image decomposition gradientdouble threshold segmentation to extract the edge information.At last to extract theedge of the image post-processing mathematical morphology, by the corrosion andopen operation.In this paper, research the algorithm and extract method of linear information ofremote sensing image, main achieved the following results:1.By using the method of wavelet transform and different set of gray-levelthreshold segmentation, extract different decomposition levels, different resolution,different threshold segmentation of texture information of shabaosi, The study ofdifferent regions, different application purpose and the conditions of different data, should set up different gray-level threshold segmentation.2.Compare and analysis extract the edge information of several classic operatorextracted from algorithm to the actual results. It is concluded that the traditionaloperator contradictions of signal-to-noise ratio.In order to extract the edgeinformation better SNR, get rid of the noise by wavelet, improving the gradient ofsingle threshold segmentation for double threshold, at the same time of try toeliminate noise, improve the continuity of edge detection and accuracy,the resultsshow that, compared with the traditional algorithm,the effect of edge extraction.3.In order to improve the visual effect of the linear edge and continuity,furtherreduce the dots and cavernous noise, use introduction of mathematical morphologymethod with the result of edge image for corrosion and open operation, the resultsshow that, The method of mathematical morphology method can effectively eliminatenoise point, To improve the continuity of edge,and enhance the edge of the visualeffect,is a good methoed to postprocessing aftercure.