Land Information Extraction Based on Multi-temporal Remote Sensing Spectrum Character in Saline Alkali land Area of Jilin-Qian'an County

In this paper, the typical saline-alkali soil in the western Songnen Plain is taken as the study area, and the spectral information of multi-temporal remote sensing images is extracted the spectral characteristics of various features in different months, and the classification is conducted according to the spectral curve characteristics of various features in different periods. The method is that the decision tree classification algorithm based on multi-temporal spectrum and phenological characteristics can effectively integrate multi-temporal and multi-spectral information, so as to overcome the defect of single-temporal image classification, and judge dry and paddy fields, light, moderate, severe saline alkali soil, alkali lake and other ground objects. The overall classification accuracy of target features is 76%, and the Kappa coefficient is 0.82. Among them, the classification effect is better for those with heavy saline-alkali soil, light to moderate saline- alkali soil, farmland (dry and paddy fields) and lakes. In the case of limited information in research area, we can get better classification results using spectrum character in multi-time image.