A Comparison of Shadow Detection methods for High spatial resolution Remote Sensing Images

Shadow detection is one of major research problems in processing high spatial resolution remote sensing images. Developing effective shadow detection methods is one of the essential topics in remote sensing image processing, particularly for urban regions and mountainous forest. Accurate detection of shadow areas in remote sensing images is vital for subsequent image classification and analysis. In this paper, the current shadow detection algorithms are reviewed and classified into 4 types: geometric model-based methods, physical model-based methods, color spacebased model methods and threshold. The research progress, advantages and disadvantages of these methods are compared, analyzed and discussed. According to the comparison, the potential promising research topics includes:(1) making the shadow detection process more robust and accurate, (2) solving the problem of automatic threshold selection. (3) utilizing machine learning algorithms, especially deep learning methods