Research And Application On The Methods Of Crop Monitoring By Remote Sensing Under Cloud Computing Environment

The remote sensing image process can be traced back to the 50s.In 1964 the United States used a computer to process the moon photos from the "travelers seven"spacecraft,they got the unprecedented clear photos.It represented the birth of the remote sensing image processing.Remote sensing image is widely used in the military,meteorology,industry,marine exploration,agricultural monitoring applications and other fields.At present,there are some problems in agriculture monitoring by remote sensing application.On the one hand,the remote sensing image data is increasing rapidly,it's time-consuming for the process of computation;On the other hand,the problems of the processing methods for remote sensing image lead to inaccurate monitoring results.When it used in crop monitoring,a large number of remote sensing image can't get efficient application.At the same time,Crop monitoring methods are varied,we need to build suitable monitoring model according to the characteristics of monitoring area,to achieve the accurate monitoring of crop condition,area in some specific area.From what have been discussed above,in this thesis,the distributed computing technology is used in the crop monitoring under cloud computing environment.The main research work and conclusions are follows:First,the thesis does some relevant research on several kinds of algorithms for remote sensing image classification,and carries on analysis comparison,processes the same remote sensing image with minimum distance classification,parallelepiped classification,maximum likelihood classification and support vector machine classification respectively,then analyzes and compares with the results of the experiment.It is concluded that the advantages and disadvantages of different classification methods and the applicable conditions.The conclusion is that the SVM classification has better accuracy,but slow processing speed.Second,in cloud computing environment,the SVM algorithm's efficiency is improved by combining the Hadoop distributed framework and the particularity of remote sensing image.The special bands are added in SVM classification to improve the classification precision.At the same time,the algorithm is applied to distributed computing,and the speed of image processing for large remote sensing is greatly improved.Third,the improved classification algorithm is applied to the monitoring of crop area and yield.With the research targets and areas,the appropriate crop monitoring model is built.In accordance with the improved classification algorithm of the fourth chapter,a new kind of crop area monitoring and yield estimation model are designed.Fourth,the experiments verify the new monitoring model in efficiency and accuracy.This thesis uses parts of Liao Zhong,Shen Yang,Liao Ning province remote sensing image as research data to implement the monitoring of corn planting area and production in 2013.