Classification And Bands Window Division On Reflectance Spectra Of Urban Area

Reflectance spectroscpy plays an important role in remote sensing. Given the high degree of spatial and spectral heterogeneity of and within various artificial and natural land cover categories, the application of remote sensing technology to mapping the urban environment requires specific attention to the spectral characteristics and the spectral dimension of urban surfaces.In this study, the accurate reflectance spectra of urban targets are collected with ground spectroradiometer , covering a wide variety of materials from 400 to 900nm with precise spectral sampling. Investigation of pure urban reflectance spectra is carried to acquire their spectral properties and separability, and to find the most suitable spectral bands window for further urban remote sensing.Using Self-Organizing Competition Neural Network (SCNN), the urban spectrums are classified to sixteen categories. And the spectral angle cosine (SAC) returns a scalar value based on some notion of similarity for any two spectrums. Based on the classified result and SAC value, the paper discussed the spectral properties of each category and summarized the potential types in classified-result. Finally, some valuable conclusions about urban spectra are presented, and several advices for urban remote sensing are also given in the paper.From the point of information, investigation is processed to find the potential fine bands window for remote sensing of urban materials. First spectral derivative, correlation coefficient matrix are used to analyze the similarity and relativity among spectral waves. Then, the method of clustering to obtain potential fine bands is proposed on the foundation of similarity and relativity. Finally, the range from 400nm to 900nm is divided to three kinds of bands, that is 25, SO and 100 bands, The evaluation from information content shows the divisions of 50 and 100 bands are better than the usual high spectral bands. So it proves, to some extent, the bands-division method is rationality. In addition, the classification of new spectra resampled by 25 bands window shows that new spectra holding main spectral characteristics can basically identify urban targets. So 25 bands can be provided as basic bands for later urban remote sensing.As a result, all the works mainly aim to provide a more comprehensive knowledge base for applying remote sensing technology to urban application.