Extraction Method For Paddy Rice Planting Region Based On Sample Knowledge Mining

Food security is the important basis of national security.Since rice is the main agricultural crops in China,it means a lot to monitor rice planting area so as to ensure national food security.Obtaining the rice planting area information accurately and timely offers significant reference for the government and helps them to make the scientific agricultural development policy.As a province with less cultivated land,monitoring the rice planting area plays an important role to make argicultral decisions for the government of Zhejiang province.Remote sensing techniques is widely used in agricultural statistics for the reason that it has excellences of updated operation,high precision and wide coverage.This paper analyzes the rice cultivation and distribution features in our study area.A specific spatially measurement called Map-spot is defined in this paper,according to the characteristic of high resolution remote sensing image.In addition,a sampling knowledge mining method is brought up,which extracts new sampling features on the basis of traditional sampling feature estimation methods.An appropriate rice planting area estimation strategy is applied in our study area combined with the new sampling features and geographical law,which is effective.The study content is as below:(1)Taking advantage of statistics which are based on the rice cultivation and distribution features in our study area to show problems and benefits by using high resolution remote sensing image on the rice planting area estimation.Defined Map-spot to be the spatial object unit in this study,which is constituted by a group of pixels with similar spectrum.(2)Based on the Analysis of the traditional classifying feature for rice estimation.In addition,this study further developed the classifying feature by sampling knowledge mining and brought up new features for classification.The sampling knowledge mining is carried out with respect to the characteristics of high resolution remote sensing image and Map-spot.The new features are described with three indexes:the area of Map-spot index,the area ratio of Map-spot index and the concentration of Map-spot index.(3)A rice planting area estimation strategy is brought up with the new sampling features and geographical law.The strategy mainly includes three aspects:first step,judge the main Map-spot from the area ratio of Map-spot index.Then,use spectrum feature and spatial feature,also neighboring correlation specifically,to extract rice planting area;Finally,making use of concentration of Map-spot index and EMD algorithm to get rid of the mistake.(4)Rice planting area estimation with our proposed strategy experiments is performed in different areas with different types of planting field.The result indicates that it is effective to make use of the sampling knowledge mining method to study the rice planting area estimation.