Wavelet Analysis and Classification of Urban Environment Using High -Resolution Multispectral Image Data.

Attempts to analyze urban features and classify land use and land cover directly from high-resolution satellite data with traditional computer classification techniques have proven to be inefficient. The fundamental problem usually found in identifying urban land cover types from high-resolution satellite imagery is that urban areas are composed of diverse materials (metal, glass, concrete, asphalt, plastic, trees, soil, etc.). These materials, each of which may have completely different spectral characteristics, are combined in complex ways by human beings. Hence, each urban land cover type may contain several different objects with different reflectance values. Noisy appearance with lots of edges, and the complex nature of these images, inhibit accurate interpretation of urban features. Traditional classifiers employ spectral information based on single pixel value and ignore a great amount of spatial information. Texture features play an important role in image segmentation and object recognition, as well as interpretation of images in a variety of applications ranging from medical imaging to remote sensing. This study analyzed urban texture features in multi-spectral image data. Recent development in the mathematical theory of wavelet transform has received overwhelming attention by the image analysts. An evaluation of the ability of wavelet transform and other texture analysis algorithms in urban feature extraction and classification was performed in this study. Advanced Thermal Land Application Sensor (ATLAS) image data at 2.5 m spatial resolution acquired with 15 channel (0.45 mum--12.2 mum) were used for this research. The data were collected by a NASA Stennis LearJet 23 flying at 6600 feet over Baton Rouge, Louisiana, on May 7, 1999. The algorithms examined were the wavelet transforms, spatial co-occurrence matrix, fractal analysis, and spatial autocorrelation. The performance of the above approaches with the use of different window sizes, different channels, and different feature measures were investigated. Six types of urban land cover features were evaluated. Wavelet transform was found to be far more efficient than other advanced spatial methods. The results of this research indicate that the accuracy of texture analysis in classifying urban features in fine resolution image data could be significantly improved with the use of wavelet transform approach