基于冠层反射率模型的作物参数多阶段反演方法研究进展

发布时间：2018-04-09 22:14

本文选题：遥感　切入点：模型　出处：《农业工程学报》2017年01期

【摘要】：利用遥感手段,基于冠层反射率(canopy reflectance,CR)模型反演农作物参数具有经济、高效、普适性好的特点,是智慧农业快速、精确监测区域尺度农情信息的理想方法。然而,CR模型反演过程受"病态反演"问题影响。针对此,前人提出了多阶段目标决策(multi-stage,sample-direction dependent,target-decisions,MSDT)法和面向对象(object-based)反演法。分别依据CR模型参数的敏感性和不确定性,以及作物参数的空间分布特征,将反演过程划分为若干阶段,每阶段只反演部分参数,前阶段反演结果作为后阶段反演的先验知识,以此减少CR模型参数优化的不确定性,改善"病态反演"问题。该文系统总结了MSDT法与面向对象反演法,将其归纳为统一的"多阶段反演"方法,并提出概念模型。基于此,总结、讨论了多阶段反演中如下三方面共性问题,试分析可能的解决途径:1)多阶段反演决策还需要广泛比较、科学论证与改进,以确保其合理性和有效性;未来研究中,应将MSDT法与面向对象反演方法有机结合,在统一的多阶段反演技术框架下,制定更加合理的反演决策方法。2)CR模型的参数化精度可能影响多阶段反演;未来应尝试利用"天空地一体化"遥感技术和尺度转换方法获取先验知识,提高CR模型参数化精度。3)多阶段反演过程中,反演误差逐级传递;未来研究中,一方面应尝试识别并纠正前阶段反演中的误差,另一方面应合理利用前阶段反演结果,避免前阶段反演误差影响后阶段的反演。
[Abstract]:The retrieval of crop parameters based on canopy reflectivity and canopy reflectance model is economical, efficient and universal. It is an ideal method for rapid and accurate monitoring of agricultural information on regional scale by means of remote sensing.However, the inversion process of CR model is affected by the problem of ill-conditioned inversion.In order to solve this problem, multi-stage objective decision making (multi-stageering) and object-oriented object-based (OO) inversion have been proposed in this paper.According to the sensitivity and uncertainty of CR model parameters and the spatial distribution characteristics of crop parameters, the inversion process is divided into several stages, in which only some parameters are retrieved in each stage, and the inversion results in the previous stage are regarded as the prior knowledge of the inversion in the later stage.In order to reduce the uncertainty of CR model parameter optimization, improve the "ill-conditioned inversion" problem.In this paper, the MSDT method and the object oriented inversion method are systematically summarized, and they are summarized as the unified "multi-stage inversion" method, and the conceptual model is put forward.Based on this, the following three common problems in multi-stage inversion are discussed. The possible solution to multi-stage inversion decision is to be compared widely and scientifically proved and improved to ensure its rationality and effectiveness.In the future research, the MSDT method and the object-oriented inversion method should be combined organically. Under the framework of the unified multi-stage inversion technology, a more reasonable inversion decision method, the parameterized precision of the two-stage CR model, may affect the multi-stage inversion.In the future, we should try to acquire a priori knowledge by using the remote sensing technology and scale conversion method of "integration of sky and earth" to improve the parameterized accuracy of CR model. 3) in the multi-stage inversion process, the inversion error should be transmitted step by step; in the future research,On the one hand, we should try to identify and correct the errors in the pre-stage inversion, on the other hand, we should make rational use of the pre-stage inversion results to avoid the influence of the pre-stage inversion error on the post-stage inversion.
【作者单位】：四川省农业科学院遥感应用研究所/农业部遥感应用中心成都分中心;中国农业科学院农业资源与农业区划研究所/农业部农业信息技术重点实验室;
【基金】：四川省财政创新能力提升工程青年基金(2015QNJJ-023) 四川省财政创新能力提升工程新兴学科专项(2013XXXK-024) 四川省财政创新能力工程高新领域扩展专项基金(2016GXTZ-011)
【分类号】：S127

【相似文献】