Texture analysis for urban spatial pattern study using SPOT imagery

Abstract

SPOT panchromatic imagery of Beijing has been studied to capture the unique spatial pattern of the city. Texture analysis, which reveals spatial variations, was adopted to interpret the urban spatial patterns. Statistical and structural texture features were extracted from the SPOT image and evaluated for their capability of detailed mapping of urban structures. Supervised image classifications were performed on combinations of different texture features. The imagery was classified into: low-rise, old multi-storey, newer multi-storey and most-recent multi-storey, high-rise, roads, construction sites, open water/vegetated and agricultural areas. The study shows that the classification accuracy of original SPOT imagery is only 46%. By using six textural channels, the accuracy has increased to 72%. The best classification shows the urban spatial pattern well.