Spectral knowledge database for sugarcane identification with multi-temporal remote sensing images

Multi-temporal remote sensing data can provide much more information that could be used to improve the accuracy of classification of vegetation types. However, it is always required to manually select a set of training samples by using conventional supervised classification methods, which is a time-consuming and costly task. In this paper, a new classification method based on spectral knowledge database (SKD) has been proposed. The spectral knowledge database is composed by a serious of spectral datasets, prior knowledge, and remote sensing models. For a specific remote sensing image, the data of object class can be simulated based on the knowledge database, which can be used to generate the classification rules. With this approach, an experiment of sugarcane identification was conducted. And the results show that the accuracy of the results by using proposed method was comparable to that by using supervised classification.