Multi-level monitoring of subtle urban changes for the megacities of China using high-resolution multi-view satellite imagery

Abstract

Megacities have evolved at an unprecedented rate under the pressure of urban population growth and economic development, particularly in the developing countries. For instance, many cities in China are experiencing a large number of rapid but subtle changes resulting from urban infrastructure construction. Information concerning such small changes is imperative for understanding the local environment and human activity, and can also provide key insights for urban planners. Undoubtedly, high-resolution remotely sensed data should play an essential role for the monitoring of such subtle changes, due to the improved observation capacity of the spatial details. However, few studies have investigated high-resolution data for change detection at a large geographic scale, due to the multi-temporal heterogeneity of the data, e.g., spatial mis-registration, parallax distortion for high architectures, and the different viewing angles. In this study, we attempted to fill this gap and resolve these problems by the use of multi-view ZY-3 satellite data, which are used to generate multi-temporal orthographic images through photogrammetric derivation. In the meantime, we present a general framework for precise urban change analysis in a multi-level (pixel, grid, and city block) approach. Two typical Chinese megacities—Beijing and Wuhan—are chosen in the experiments. The results confirm the accuracy of the proposed multi-level method for monitoring subtle urban changes, achieving Kappa coefficients of ~0.8 at the pixel level and a correctness of 93–95% at the grid level. The landscape analysis further indicates that the rapid urban construction led to greater fragmentation and spatial heterogeneity of buildings and decreased minimum distance between building patches (by ~1.0m between 2012 and 2013). Moreover, the performances of ZY-3 and Landsat for the monitoring of subtle urban changes are compared, revealing that the high-resolution sensor—ZY-3 (2.5m)—is essential for precisely detecting subtle urban changes, whereas the Landsat data (30m) are not sensitive to most of the subtle changes that occur in the urban areas.