Zonal mapping of urban areas using MEIS-II airborne digital images

The aim of this study is to evaluate the contribution of texture information to high spatial and spectral resolution for zonal mapping in urban areas. The airborne data used were acquired over the Montreal Urban Community (Montreal, Canada) by the Multi-detector Electro-optical Imaging Scanner (MEIS-II) of the Canada Centre for Remote Sensing. Spectral analysis of the divergence between classes has shown that bands 3 (847-901 nm), 4 (622-659 nm); and 7 (433-463 mm) offer the optimal combination for discriminating between the urban classes we have defined. Textural features derived from three different order histograms were calculated from band 4 and evaluated in terms of their ability to discriminate between urban classes. We then extracted the best feature from each of the histograms. The integration of these three textural features with the three spectral bands has shown that textural information permits an improvement in class separability resulting in an increase in the rate of correct classification in the order of 12%. This increase is however dependent on the type of class and varies from 4.5% for forest and parks to 16.8% for urban areas with low vegetation density. Classification by maximum likelihood of these spectral-textural data and the visual analysis of results also show that textural information, in conjunction with high spatial and spectral resolution, provides appropriate zonal mapping of urban areas.