Spatiotemporal Fuzzy Clustering Strategy for Urban Expansion Monitoring Based on Time Series of Pixel-Level Optical and SAR Images

Abstract

Monitoring urban expansion dynamically using remote sensing technology is an essential method for obtaining and understanding urban spatial structure. However, the quality of traditional optical images in some areas is poor due to clouds and fog. Compared to optical images, synthetic aperture radar (SAR) can achieve earth observations without the limits of sunlight and weather conditions, but its speckle is too obvious. This paper combined the advantages of pixel-level optical image and SAR image time series and proposed a spatiotemporal fuzzy clustering (STFC) strategy for urban expansion monitoring. This strategy includes three parts: 1) the construction of optical-SAR image mixed time series; 2) a time-series fuzzy information granulation method to ascertain change nodes; and 3) STFC to determine the change types and range. In our study, 13 TM images and 25 SAR scenes taken from 2005 to 2011 were selected as raw data. We used the proposed method to monitor the urban expansion of Chengdu, China, and then, analyzed its main causes according to the monitoring results. The results suggested that: 1) the proposed methods could effectively extract the change nodes and change pixels, with the correctness of 85.20% and the completeness of 86.06%, outperforming the time series only (nonspatial) fuzzy clustering method, as well as traditional classification methods; and 2) the urban expansion of Chengdu is most apparent from 2005 to 2011, with the expansion direction shifting from the traditional ring structure expansion to point-axis expansion following the priority given to construction of new urban areas.