Construction And Application Of A New Class Of Image Texture Feature

With the development of computer technology and the wide application of digitalmultimedia technology and intelligent information processing technology, imageprocessing attracts more people’s attention and image classification is an importantcontent of image processing. Important prerequisite of image classification is toaccurately extract the image feature. The texture features can exhibit the nature of theimage. Texture is an important visual features of image recognition. Texture featurecontains important information arranged in the surface structure of the organization.Texture analysis technology has been active research area in computer vision, imageprocessing and image retrieval. The appropriate image texture features to describe anobject or image of particular concern is the important foundation of the imagerecognition and image understanding.1. By analyzing the content of the Gray-level Co-occurrence Matrix(GLCM)element, constant or gradient image can make the main diagonal elementsand sub diagonal elements of GLCM lager. hop-hop image can make the away frommain diagonal elements and sub diagonal elements of GLCM lager. Therefore, thispaper constructs jump/flat features, which reflecting differences in the two typeimage.2. AP clutter can interfere with strong convection echo’s identification of weatherradar image. This paper start with AP clutter and strong convection echo’s distributioncharacteristics in the radar reflectivity image. On the basis of region division, it willgenerate GLCM. GLCM elements are divided into two subsets, which build two newgentle/jumping feature. These features show significant differences between APclutter and strong convection echo. Taking into account without loss strongconvection, this paper form classification decision tree under using radial velocitycharacters and gentle/jumping feature. After filtering AP, for the short-term heavyrainfall and hail echo in strong convection, reflectivity shown slow degeneration, butthe reflectivity of hail more compact.3. This article will make six characteristic variables of hail to reduce the numberof dimensions through the use of factor analysis, and with the use of support vectormachines. Large hail and small hail achieve the effect of the initial classification.4. Jump feature can reflect mist and the basis of quantitative estimates ofvisibility in fog.The algorithms involved have been realized by program which presents strong stability and satisfactory results on a large number of sample test.