From Spectrum to Spectrotemporal: Research on Time Series Change Detection of Remote Sensing

Spectral information of ground object is an important feature of remote sensing data. The utilization of remote sensing spectral information has experienced the development from black and white panchromatic image to multi-spectral, hyperspectral and time series images. In recent years, with the development of satellite remote sensing and the accumulation of historical data, a large number of revisit observation data have been obtained. The long time series remote sensing data integrate spectral, temporal, and spatial information and can avoid the phenomenon of different object with the same spectra characteristics and same spectrum with different objects to some extent. But there is still no unified concept to describe it. This paper proposed the concept of spectrotemporal based on the concept of spectrum, constructed the theoretical system of spectrotemporal, and discussed the theoretical methods and key technologies of spectrotemporal in detail. First, this paper reviewed the traditional time series change detection methods of remote sensing and analyzed the advantages and disadvantages of the traditional change detection methods. Then, the theory and methods of spectrotemporal were expounded, and the key technologies of spectrotemporal in data organization and storage, data reconstruction, feature depth mining and data source were introduced. The methods of classification and change detection for typical crop based on spectrotemporal information were also proposed. Finally, the research of spectrotemporal change detection is prospected.