Texture Feature Research Of Multi-spectral Remote Sensing Image

Remote sensing, based on aerial photography technology, is a newtechnology developed from Tthe sixties of the 20th centuryT. It is widely used inTmeteorology, mapping, earth resources and environmental science, agriculture,forestry, geology and other major areas.Remote sensing image interpretation is the process of making remotesensing image data into land cover information, which can be divided into threeways: manual visual interpretation, automatic interpretation of computer andhuman-computer interactive interpretation. Only through the interpretation,remote sensing images can be applied to different areas. At present, thetraditional manual visual interpretation has been unable to fit the increasinglylarge amount of remote sensing data. In the future, human-computer interactiveinterpretation or automatic interpretation is an inevitable trend.Spectral features and texture features are often used in remote sensing imageinterpretation, but traditional texture analysis methods and the texture definitionof images, used for the single-band image, are unsuitable for multi-spectralimage. Therefore, it becomes a very important issue that how to analyse theimage texture according to the characteristics of the multi-spectral images. Thereare lack of the theories and methodologies for the processing and analysis ofremote sensed images. Starting from texture analysis of multi-spectral remotesensing images, we proposed a new texture analysis method, gray leveldifference associated probability matrix(GLDAP), which is more suitable formulti-spectral image, by studying the commonly used gray level co-occurrencematrix texture analysis method. TWith the extracted features used to the multispectralremote sensing image classification, we sTelected water and land, water,urban areas and mountains as Tsubjects of classificationT experiments, and usedsupport vector machine technology developed from the ninetiesT of the 20thcenturyT as a classifier in order to verify the effectiveness in the field of remotesensing classification using GLDAP. Experimental results show advantage ofGLDAP, in multi-spectral remote sensing classification, over the traditional GLCM method. In addition, some experiment and results were given in the direction of three-dimensional information restoration from two-dimensional image.