|
We propose a novel method, in this paper, to monitor the change of urban impervious surface area (ISA) using time series high-resolution remote sensing image. The distribution of urban ISA is an important indicator of urban ecological environment. The extraction of urban ISA by remote sensing technology is of great significance for understanding urbanization, hydrological cycle and urban heat island effect. Previous researches on ISA extraction mostly used medium and low-resolution images to study the change of ISA at the macro level. Because the resolution of data source is too low to detect the micro change of meter scale, it cannot meet the needs of sponge city and ecological city planning for fine modeling of urban underlying surface, and it is objectively necessary to carry out high-precision and fast ISA monitoring. For example, using high-resolution remote sensing images combined with machine learning algorithm, introducing time series features to construct a monitoring model. The difficulty of water surface monitoring based on time series is how to find the pseudo change, and how to extract the time-invariant spectral features is the difficulty. At the same time, time series images can solve the spectral confusion of surface features. The purpose of this study is to enhance the spectral difference between ISA and pervious surface by analyzing the time-variant spectral features extracted from time series high-resolution remote sensing satellite images. In this research, we took Hengqin New District, Zhuhai City, Guangdong Province of China as the study area. The area experienced rapid urbanization from 2009 to 2018. According to the ten-year time series of high-resolution remote sensing images, the ten-year ground features classification results of the study area acquired. The contribution of this paper is to use a method to extract time-series spectral features from time-series images to assist in the extraction of ISA. The average overall extraction accuracy of ISA in ten years increased from 87.8% to 91.1%.
|
