Multi-scale remote sensing to support insurance policies in agriculture: from mid-term to instantaneous deductions     

Sarvia F  De Petris S  Borgogno-Mondino E 《地理信息系统科学与遥感》2020,57(6):770-784

ABSTRACT

Climate change is today one of the biggest issues for farmers. The increasing number of natural disasters and change of seasonal trends is making insurance companies more interested in new technologies that can somehow support them in quantifying and mapping risks. Remotely sensed data, with special focus on free ones, can certainly provide the most of information they need, making possible to better calibrate insurance fees in space and time. In this work, a prototype of service based on free remotely sensed data is proposed with the aim of supporting insurance companies’ strategies. The service is thought to calibrate annual insurance rates, longing for their reduction at such level that new customers could be attracted. The study moves from the entire Piemonte region (NW Italy), to specifically focus onto the Cuneo province (Southern Piemonte), which is mainly devoted to agriculture. MODIS MOD13Q1-v6 and Sentinel-2 L2A image time series were jointly used. NDVI maps from MODIS data were useful to describe the midterm phenological trends of main crops at regional level in the period 2000–2018; differently, Sentinel-2 data permitted to map local crop differences at field level in 2016 and 2017 years. With reference to MODIS data, the average phenological behavior of main crop classes in the area, obtained from the CORINE Land Cover map Level 3, was considered using a time series decomposition approach. Trend analyses showed that the most of the crop classes alternated three phases (about 7 years) suggesting that, presently, this is probably the time horizon to be considered to tune mid-term algorithms for risk estimates in the agricultural context. Crop classes trends were consequently split into three phases and each of them modeled by a first-order polynomial function used to update correspondent insurance risk rate. Sentinel-2 data were used to map phenological anomalies at field level for the 2016 and 2017 growing seasons; shifts from class average behavior were considered to locally and temporarily tune insurance premium around its average trend as described at the previous step. Synthesizing, one can say that this approach, integrating MODIS and Sentnel-2 data, makes possible to locally and temporarily calibrate premiums of indexed insurance policies by describing the average trends of crop performance (NDVI) at regional level by MODIS data and refining it at field and specific crop level by Sentinel-2 data.  相似文献   

高应力区破裂岩体回采巷道一次强化支护技术及其应用研究   总被引：2，自引：1，他引：1  

朱建明  高谦  林新平 《工程地质学报》2002,10(4):400-403

基于对高应力区破裂岩体回采巷道支护 10多年的经验和教训 ,本文首次提出了相对于主次承载区协调作用关系的关键支护技术 ,即一次强化支护技术。该支护技术不仅强调了对于高应力区破裂岩巷道支护时间的及时性 ,支护强度一次到位 ,更重要的是区别于先让后抗、二次支护的指导思想 ;同时 ,一次强化支护技术还详细明确了支护类型、支护结构以及施工工艺。该支护技术在小官庄铁矿高应力区的应用成功 ,表明该支护关键技术是正确的。  相似文献   

基于支撑向量回归的高光谱混合像元非线性分解   总被引：10，自引：1，他引：10  

吴波  张良培  李平湘 《遥感学报》2006,10(3):312-318

提出了基于支撑向量回归的高光谱混合像元自动分解.首先利用投影迭代的方法自动寻找到影像的典型地物光谱,然后利用Hapke近似函数模拟出非线性的训练和测试数据.支撑向量回归的混合像元分解方法与基于基函数分解方法的不同点是不需要预先确定非线性的映射形式,它通过核函数,把像元矢量从低维空间映射到高维特征空间,使得在特征空间中构造的线性光谱组合对应着原始空间（像元空间）的非线性组合特性,从而揭示了典型地物光谱之间的高阶性质,提高了混合像元的分解精度.实验结果证明,这种方法具有很高的混合像元的分解精度.利用模拟数据作分解精度的评价,表明97%以上的像元分解绝对误差不大于10%,而各类总体平均平方根误差均小于3.5%.  相似文献   

基于支撑向量机概率输出的高光谱影像混合像元分解   总被引：5，自引：0，他引：5  

吴波  张良培  李平湘 《武汉大学学报(信息科学版)》2006,31(1):51-54

提出利用支撑向量机（SVM）后验概率来分解高光谱影像的混合像元，通过支撑向量机的输出值转化为两两配对的后验概率，再由两两配对的概率值求得多类后验概率，并以像元所属类别的后验概率作为地物的组分信息。实验结果表明，该方法能较好地估计出混合像元的组分比。  相似文献   

使用SVM进行短期气候预测的应用试验     

高宇  燕东渭  杨艳  赵国令  刘志镜 《沙漠与绿洲气象(新疆气象)》2006,29(5):15-16

从气象角度介绍了利用SVM方法的预报原理,对西安短期气候预测做了实验,并分析预报结果,得出SVM方法在天气预报领域有一定的使用空间的结论。  相似文献   

基于小波分解和最小二乘支持向量机的西太平洋副高预测模型   总被引：1，自引：1，他引：1       下载免费PDF全文

刘科峰  张韧  于鹏  王彦磊  余丹丹 《热带气象学报》2007,23(5):491-496

用小波分解（WT）和最小二乘支持向量机（LS-SVM）相结合的方法,建立西太平洋副热带高压面积指数的预报模型。该方法首先将西太平洋副热带面积指数（SI）分解为相对简单的带通分量信号,利用LS-SVM建立各分量信号的独立预报模型,然后对预报结果进行集成。为了评估和比较该方法的预报效果和技术优势,最后比较了在同等条件下WT～LS-SVM模型和神经网络、线性回归模型的独立检验预报效果。试验结果表明,该方法具有泛化能力强、预报精度高、训练速度快、稳定性好、便于建模等优点,具有良好的应用前景。  相似文献   

Deformation Mechanism and Stability of a Rocky Slope   总被引：1，自引：1，他引：1  

黄润秋肖华波  巨能攀赵建军 《中国地质大学学报(英文版)》2007,18(1):77-84

A high slope is located on the side of the spillway at a hydropower station in Southwest China, which has some weak inter-layers inclining outwards. Parts of the slope show heavy weathering and unloading. There appeared deformation and tensile crack either on the surface or on the afteredge of the slope during excavation, and under a platform （elev. 488 m）, two levels of slopes collapsed on the downriver side. Based on the investigation in situ and the analysis of the geological structure, the conceptual model of deformation and failure mechanism was erected for this slope. Furthermore, the deformation characteristics were studied with FLAC^3D numerical simulation. Comprehensive analysis shows that the whole deformation of the slope is unloading rebound in certain depth scope and the whole body does not slide along any weak interlayer. In addition, two parts with prominent local deformation in the shallow layer of the slope show the models of ＂creep sfiding-tensile cracking＂ and ＂slidlng-tensile cracking＂, respectively. Based on the above analysis, the corresponding project of support and reinforcement is proposed to make the slope more stable.  相似文献   

加权支持向量回归机及其在水质预测中的应用   总被引：1，自引：0，他引：1  

徐红敏  王继广 《世界地质》2007,26(1):58-61

支持向量机是一种基于结构风险最小化原理的学习技术，也是一种新的具有很好泛化性能的回归方法。本文对用于回归估计的标准支持向量机加以改进，提出了一种新的用于回归估计的支持向量机学习算法，针对各样本重要性的差异，给各个样本的惩罚系数和误差要求赋予不同权重，并利用加权支持向量回归机的理论及其算法构建水质预测模型。实验结果表明，该方法对水质具有较好的预测效果。  相似文献   

采用锚固和支撑与新型防护网SNS联合治理黄山风景区云汤线危岩崩塌     

马作明 《安徽地质》2007,17(1)

将传统的支撑锚固与新型SNS边坡防护技术应用于危岩的治理工程,不但解决了大块危岩治理问题,还解决了松动而无法清除危岩的崩落、掉块问题,且对自然景观破坏小,对旅游公路运营影响小,解决了场地狭窄与工程安全的矛盾,施工方便迅速,工程安全可靠。  相似文献   

土地覆盖制图：基于最优化遥感数据的支撑向量机分类   总被引：1，自引：0，他引：1  

Gidudu Anthony  Heinz Ruther 《遥感学报》2007,11(5):694-701

遥感数据具有在不同空间、光谱和时间尺度上获取地表测量信息的能力,使其成为获取土地覆盖信息的一个主要数据源。影像分类即把卫星影像上的相关像元划分给某类已知的土地覆盖类型的过程。支撑向量机（SVMs）是一种土地覆盖分类的新技术。三种常用的SVMs是：基于线性和多项式的SVM以及具有高斯核函数的SVM分类器,分类能否成功地应用有赖于其各自选择的最佳参数。但是海量的遥感数据使得这些参数的确定速度十分缓慢。本文研究了一种新的基于最优化遥感数据压缩技术的SVM分类方法。研究显示用于获取SVM参数的数据量能够在不影响土地覆盖的分类精度的前提下进行压缩。数据压缩成功的应用于多项式和高斯核函数的SVM分类,而线性SVM的分类精度却非常低。  相似文献