A Research Of Extracting Agricultural Growing Areas Information With FY-3 Mersi Data

Timely grasp crop area and condition information is of importance for guiding agricultural production and management for the large area. Therefore, how to retrieve the crop area information with higher accuracy has been one of hot topics in remote sensing field in recent years. This thesis takes the North China Plain as the study area. Considering the corresponding crops phonological information, the thesis firstly built different hierarchical decision trees to extract the bare land and winter wheat growing areas with FY-3 data. Then the thesis built different models to extract plants distribution in different years based on MERSI NDVI ten-day products and corresponding crops feature information.The major study points are as follows:(1)Mapping winter wheat growing areas on the basis of the hierarchical decision tree. Firstly, the people select some MERSI imageries with high quality and use the method of the hierarchical classification. As the different levels, the people choose the most sensitive bands to construct the corresponding decision tree, which will extract winter wheat area and bare land from these images. Secondly, the people compile all classification results into a map of winter wheat and bare land distribution. Finally, the people verify the accuracy with the field survey data and LANDSAT 8 data, compare the result of the hierarchical classification with other classifications and analyze the specific advantages and disadvantages. Once the method is determined, the people mapped the winter wheat since 2010. The results show that the hierarchical decision trees has a higher accuracy than other classification methods, that is, bare land, winter wheat and overall from 2013 to 2014 have the highest precision, 91.80%, 90.19%, 90.90%, respectively. At the same time, the distribution of winter wheat and bare land with MERSI data and LANDSAT 8 images are the same from the counties scale.(2)Mapping different types of corps growing areas based on NDVI time-series. Firstly, The people use maximum synthesis method to generate 250 meters resolution MERSI NDVI ten-day products of the north China plain. Secondly, the people mask the selection of NDVI images with the different years of winter wheat growing areas and bare land to calculate NDVI curve of the spring maize, summer maize and cotton of the north China plain. And combining with the corresponding crop information, the people set up extraction models to get different types of corps. Finally, the people verify the accuracy of the results with the field collection data and LANDSAT 8 imagery interpretation data. The results show that different years’ accuracy of overall results is more than 84%, which meets the needs of agricultural information monitoring from the aspect of remote sensing mapping and monitoring in the resolution of 250 meters. At the same time, the crops distribution with MERSI data and LANDSAT 8 images have a good consistency in the counties scale.(3)Development of the classification system. Base on these classification algorithms and the results of the thesis, the people develop a classification system from the system application requirements and design idea combining with data processing work flow. The system is composed of three modules, including the training sample data format conversion module, the classification method selection module and a data fusion processing module.