Research On Crop Remote Sensing Classification Method Based On Integrated Learning

Efficient and accurate understanding of the distribution and dynamic changes of regional spatial planting of crops,is an important basis for a fixed range of crop yield estimation and adjustment,at the same time,it is also an important reference significance to put forward appropriate food plans,economic policies and managements of our country.The classification of remote sensing images can be used to monitor the planting area of crops,but most of the current single classification methods have limitations.The focus of this study is how can we combine the advantages of several individual classifiers and improve the classification accuracy of remote sensing images.Therefore,in this paper,the maximum likelihood method,Mahalanobis distance method and support vector machine(SVM)in supervised classification are combined by a based on output decision ensemble learning method.The Landsat 8 remote sensing image in Helen City,Heilongjiang province was studied by this method.Through the classification of remote sensing images in this project completed six sets of data obtained the accuracy of comparison,found that the precision of the ensemble learning method based on output decision higher than three single classification methods,not only reflected in the overall accuracy and Kappa coefficient,but also various types of precision and overall robustness is improved.This paper focuses on the research of crop remote sensing image classification.The main research contents and innovations are as follows:(1)This paper Summarizes the research status of current classification methods in remote sensing images.First,analyses and studies the characteristics and algorithms of several classic classifiers,studies their,and considers the characteristics of crops remote sensing images,then finishes the crop remote sensing image classification based on classical classification.(2)This paper summarizes the application of the existing ensemble learning methods in the classification of remote sensing images,then analyzes and verifies them by experiments.Based on the detailed analysis of multiple classifier performance and characteristics,selects of good performance of three kinds of classic classification methods,and then using a based on output decision ensemble learning method to combine this three classic methods,and analyzes the relationship between the diversity of classifiers and classification accuracy.Research carried out for multiple sets of experiments to determine the classification rules and weights,the confusion matrix by comparing several groups of results,which can prove this based on output decision ensemble learning method can effectively improve the classification accuracy.