Research On Typical Crop Classification Methods For Agricultural Insurance In Henan Province

The rapid development of agriculture is of great significance to the implementation of the rural revitalization strategy and poverty alleviation.It can effectively guarantee food security and the stable development of social economy.As an effective risk protection mechanism,policy-based agricultural insurance plays a role on the basis of safeguarding the development of modern agriculture and protecting the interests of farmers.The scientific and effective supervision is crucial to the promotion of policybased agricultural insurance in order to improve the government’s guiding nature of agricultural insurance and enhance farmers’ agricultural insurance awareness,and expand agricultural insurance coverage.Therefore,it has been an important part of scientific issues of agricultural remote sensing to know how to use the rapidity of remote sensing technology to obtain timely and accurate crop planting area and spatial location and effectively guarantee the authenticity and validity of the underwriting and claims data in the agricultural insurance process to provide data support for the government’s financial decision-making.The author takes Tongxu County,Kaifeng City in Henan Province as the research area,and constructs time-series Sentinel-2 images as the data source.The winter wheat,garlic,vegetable fields and other main features in the area are the research objects,based on object-oriented thinking Combining analysis of multi-temporal data to realize the extraction of information on the types,locations and areas of crops in the study area.Firstly,the author extracts the spectrum,texture and vegetation index characteristics of typical features in the time series image data of the study area based on the random forest feature importance evaluation method,the feature set is optimized.Twenty most important characteristic bands are determined by the feature selection method to participate in the classification during the crop growth period,which not only guarantees the classification accuracy but also reduces the data redundancy.Secondly,it is multi-temporal feature combination segmentation.Combined with the idea of object-oriented classification,multi-scale segmentation is carried out by using the optimal features of multi temporal remote sensing image,and the optimal segmentation scale is determined to obtain the object more matching with the actual features as the minimum analysis unit,so as to improve the accuracy of crop classification and recognition.Finally,the object and pixel are the basic analysis units and the support vector machine classifier is used to identify and evaluate the accuracy of crop classification in the target area.The results show that the accuracy of object-based classification is 93.74%,which is 9.71% higher than that of pixel based classification.