Abstract: In order to quickly obtain the regional distribution information of rice and lay a technical foundation for crop sensing monitoring,promoting the development of monitoring and management of rice growth areas in the middle and high latitudes,Ulan Hot City of Inner Mongolia Autonomous Region was selected for analysis of the regional application potential of Sentinel-2 data.In this study,the single-temporal remote sensing image on September 9,2018 was used as the best observation phase,and the method of the support vector machine classification was used.The method of maximum likelihood classification and object-oriented method of approaching -K,combined with visual interpretation results were used to classify and recognize the paddy in the whole region.The accuracy of classification was verified by the confusion matrix and ground sample data.In the high latitude growth area of single-cropping rice,the accuracy using the method of maximum likelihood classification（89.35%） was higher than that using support vector machine classification and object-oriented method.It was 4.6% and 12.45% higher,respectively,than the other two ways of classification accuracy.In term of the accuracy of evaluation of rice area monitoring,the average accuracy（85.91%） of maximum likelihood classification was higher than the other methods by 8.90% and 12.61%,respectively.In term of the rice harvest,the methods of rice area extraction based on the data of Sentinel-2,the method of maximum likelihood is better than the method of supporting vector machine and the method of the object-oriented classification.

Key words: Rice, Support vector machine classification, Maximum likelihood classification, Object-oriented method, Classific-ation accuracy