Extraction of fall crop types based on spectral analysis

For meeting the demand for large-scale agricultural monitoring system with remote sensing technology, extracting crop information on the remote sensing image must be rapidly, precisely and reliably conducted. In this paper, the fall crop identification with Terra/MODIS was taken as an example in Beijing of China. Applying spectral analysis and time series characteristics, the decision tree algorithm was put forward, which can extract the main fall crops effectively and easily. Firstly, according to the spectral and biological characteristics of the fall crops, the spectral reflectances of MODIS were analyzed. One of red, blue, NIR and ESWIR band was selected as working band. Secondly, land surface water index(LSWI), which is defined by NIR and ESWIR, enhanced vegetation index(EVI), which is defined by Red, NIR and Blue bands were used as characteristic parameters to improve the precision. Finally, the decision tree algorithm was used for the fall crop identification. To verify the result, the extracting results were compared with the statistical result of State Statistics Bureau. The precision reaches 86%. This shows that it can obviously improve the crop identification accuracy with the decision tree algorithm and can be good enough to meet the operational method for agricultural condition monitoring with remote sensing and information service system at national-level.