遥感交叉验证农业保险数据的路径与指标研究  

A Study on the Technology Path and Indicators of Cross Validation of Agricultural Insurance Data with Remote Sensing

作　　者：赵思健[1] 赵坚 陈爱莲 赖积保 李宏伟 ZHAO Si-jian;ZHAO Jian;CHEN Ai-lian;LAI Ji-bao;LI Hong-wei

机构地区：[1]中国农业科学院农业信息研究所 [2]国家航天局对地观测与数据中心 [3]国家航天局对地观测与数据中心成果转化部 [4]国家航天局对地观测与数据中心、中国航天科工集团第三研究院

出　　处：《保险研究》2024年第1期49-58,共10页Insurance Studies

基　　金：国家国防科技工业局重大专项工程中心课题“高分遥感在农业保险真实性交叉校验中试点应用”(2X2X-CGZH-40-202238);国家自然科学基金面上项目“自然灾害风险的时空尺度效应分析与推绎技术研究——以农业旱灾风险为例”(41471426)。

摘　　要：有效识别、及时遏制虚假和不规范承保,提升农业保险数据的真实性是农业保险高质量发展过程中亟待解决的问题。本文以S省M县小麦完成成本保险为例,开展卫星遥感交叉验证农业保险数据的技术路径与验证指标研究。在技术路径上,采用基于深度学习的耕地自动提取算法和基于决策树的冬小麦分类提取算法,提取全县高精度的耕地地块和冬小麦种植分布数据。在验证指标上,提出区域交叉验证指标、地块交叉验证指标和保险机构交叉验证综合评价指标体系。从区域交叉验证结果来看,M县15个乡镇中有2个乡镇存在疑似超面积承保的问题,1个乡镇存在保险覆盖率不足的问题。在1个超面积承保较严重的乡镇中,存在11个村庄疑似超面积承保的问题,2个村庄保险覆盖率不足的问题。从地块交叉验证结果来看,保险机构的标的地块采集规范性和真实性普遍较差,甚至出现非耕地承保、非保险作物承保等严重问题。从综合评价来看,保险机构在承保真实性和采集规范性上存在较大差异。建议各地政府管理部门开展常态化的遥感交叉验证工作确保农业保险数据的真实性,并制作标准化的地块数据共享给保险机构;建议保险机构以地块数据为基底开展保险标的采集,确保保险标的的规范与真实。Effectively identifying and timely curbing false and non-standard underwriting,and improving the authenticity of agricultural insurance data is one of urgent issues to be addressed in the process of high-quality development of agricultural insurance.Taking the full-cost insurance of winter wheat in County M,Province S as a case,this article uses high-resolution satellite remote sensing to explore the path and indicators of validation to agricultural insurance data.In terms of technical path,an automatic extraction algorithm for cultivated land based on deep learning and a winter wheat classification algorithm based on decision tree are used to extract high-precision cultivated land data and winter wheat planting distribution throughout the county.In terms of validation indicators,regional cross validation indicators,land mass cross validation indicators,as well as a comprehensive evaluation index system for cross validation for insurance institutions are proposed.From the perspective of regional cross validation results,2 out of 15 towns in County M are suspected to be over-insured in terms of land area and 1 town has insufficient insurance coverage.In a town with an excessive area insurance coverage,there are 11 villages suspected of excessive insurance,and 2 villages have insufficient insurance coverage.From the land mass cross validation results,it can be seen that the data collection regularization and authenticity of the subject land parcels by insurance institutions are generally poor,and there are even serious problems with non-cultivated land or non-insurance crop being covered.From a comprehensive evaluation perspective,there are significant differences in the authenticity and regularization of underwriting services among insurance institutions.The research suggests that relevant local government organs carry out regular cross validation work with remote sensing to ensure the authenticity of agricultural insurance data,and produce standardized land parcel data to share with insurance institutions.It is a

关 键 词：农业保险 承保 真实性 遥感 交叉验证 道德风险 

分 类 号：F840.66[经济管理—保险]