Classification of optical high resolution images in urban environment using spectral and textural information

Conventional multispectral classification methods show poor performance in the detection of urban object, due to the high within-class spectral variance of classes corresponding to complex urban areas. In this paper, to improve the classification accuracy, we propose a data fusion approach based on the joint use of spectral and spatial information provided by the texture features extracted from the Gray Level Co-occurrence Matrix (GLCM). Specifically, a three-stage process characterizes our approach. The first stage concerns texture feature extraction considering several combinations of the three GLCM parameters: window size, step and angle. In the second stage a feature selection algorithm is applied to reduce the redundancy of the feature vector composed of both spectral and texture features. The third stage is a supervised classification. Finally, we propose an adaptive approach to extract the GLCM features which exploits the spatial information provided by a conventional segmentation algorithm. The proposed approach has been tested by using IKONOS data at 4 m resolution.