Monitoring urban impervious surface growth in Xuzhou using CBERS and BJ-1 Remote Sensing images

As the study area, Xuzhou City was chosen, located in the northwestern of Jiangsu Province, China. And two CBERS images and one BJ-1 small satellite image were employed to impervious surface extraction. Using multi-layer perception (MLP) neural network, all pixels were decomposed to the four fraction images representing the abundance of four endmembers: vegetation, high-albedo objects, low-albedo objects and soil. Then, the impervious surface was derived by the combination of high-albedo and low-albedo fraction images after removing the influence of water body. Furthermore, ALOS Pan high resolution image, that covering the city center of study area were selected to validate the impervious surface estimation results and evaluate the accuracy of impervious surface extraction. By comparing the urban impervious surface abundance from three remote sensing images, the change pattern of impervious surface was studied. The past years saw the impervious surface had been grown rapidly in Xuzhou City, especially in the north-east region, and south-east (Tongshan New District, Nanhu Campus of China University of Mining and Technology (CUMT)).