CN105389466A

CN105389466A - Middle and low resolution remote sensing product true value acquisition method for correcting scaling effect

Info

Publication number: CN105389466A
Application number: CN201510728112.3A
Authority: CN
Inventors: 吴骅; 李召良; 房世峰; 倪丽; 唐伯惠; 唐荣林
Original assignee: Institute of Geographic Sciences and Natural Resources of CAS
Current assignee: Institute of Geographic Sciences and Natural Resources of CAS
Priority date: 2015-11-02
Filing date: 2015-11-02
Publication date: 2016-03-09
Anticipated expiration: 2035-11-02
Also published as: CN105389466B

Abstract

The invention discloses a method for obtaining the true value of medium and low-resolution remote sensing products for correcting scale effects, which is divided into the following five steps: quantitative remote sensing product production; model second-order derivative estimation; spatial heterogeneity estimation; spatial scale effect estimation; spatial scale effect correction. The present invention gets rid of the shackles of matching high-resolution data requirements in the process of obtaining the true value of remote sensing products, breaks through the limitation of continuous guideability of remote sensing inversion models in the prior art, and considers intra-class heterogeneity and inter-class heterogeneity at the same time The resulting scale effect problem further improves the acquisition accuracy of the pixel-scale true value of medium and low-resolution remote sensing products, and finally realizes the simple and accurate acquisition of the pixel-scale true value of medium-low resolution remote sensing products.

Description

A Method for Acquiring the True Value of Medium and Low Resolution Remote Sensing Products Corrected for Scale Effect

技术领域technical field

本发明涉及一种中低分辨率遥感产品真值获取方法，尤其涉及一种校正尺度效应的中低分辨率遥感产品真值获取方法。The invention relates to a method for acquiring the true value of low- and medium-resolution remote sensing products, in particular to a method for obtaining the true value of low- and medium-resolution remote sensing products for correcting scale effects.

背景技术Background technique

20世纪六七十年代，随着遥感技术的不断发展，遥感科学为地表要素的区域测量和估算提供了强有力的手段，可以说是从传统的“点”测量向“面”测量发展的一次飞跃，遥感技术逐步成为目前获得区域尺度地表信息的唯一有效手段。然而，这光鲜的背后却隐藏着中低分辨率遥感产品真值获取仍旧并非易事，特别对于非均一的地表。究其原因，主要是因为遥感反演过程中使用的模型大多都是通过分析特定波长下实际地表参数同传感器测量的物理量之间的关系而建立起来的。简言之，这些遥感模型大多都是从点源观测发展而来的，通常都只适合于地表下垫面均一的情况。在实际应用中，由于地表的非均一性，中低分辨率像元多数都是混合像元。不加改正就使用这些遥感模型来反演地表参数势必会因为尺度的改变而带来较大的反演误差，造成所谓的尺度效应问题，影响了遥感产品地表真值的高精度获取，无法满足各行各业决策对于地表参数遥感反演精度的需求。In the 1960s and 1970s, with the continuous development of remote sensing technology, remote sensing science provided a powerful means for the regional measurement and estimation of surface elements, which can be said to be a development from traditional "point" measurement to "surface" measurement. With leaps and bounds, remote sensing technology has gradually become the only effective means of obtaining regional-scale surface information. However, behind this glamor lies that it is still not easy to obtain the true value of low- and medium-resolution remote sensing products, especially for non-uniform surfaces. The reason is mainly because most of the models used in the remote sensing inversion process are established by analyzing the relationship between the actual surface parameters at a specific wavelength and the physical quantities measured by the sensor. In short, most of these remote sensing models are developed from point source observations, and are usually only suitable for the case where the underlying surface is uniform. In practical applications, due to the heterogeneity of the ground surface, most of the low- and medium-resolution pixels are mixed pixels. Using these remote sensing models to invert surface parameters without correction will inevitably lead to large inversion errors due to scale changes, resulting in the so-called scale effect problem, which affects the high-precision acquisition of the ground truth value of remote sensing products, and cannot satisfy The decision-making needs of various industries for the accuracy of remote sensing retrieval of surface parameters.

现有的获取方法存在的缺陷是：The disadvantages of existing acquisition methods are:

中低分辨率遥感产品真值的估算脱离不了高分辨率数据。配套的高分辨率数据本身获取就较为棘手，成本代价较高。此外要求高分辨率数据和中低分辨率数据的成像时间、观测几何等条件均近似保持一致，否则空间异质性的估算会带来一定的偏差，进而影响到中低分辨率遥感产品真值的获取精度。这一局限性导致了在实际应用过程中由于高分辨率数据的缺乏而使得中低分辨率的遥感产品真值无法有效精确获取。The estimation of the true value of medium and low resolution remote sensing products cannot be separated from high resolution data. Acquiring the supporting high-resolution data itself is more difficult and costly. In addition, the imaging time and observation geometry of high-resolution data and low-resolution data are required to be approximately consistent, otherwise the estimation of spatial heterogeneity will bring a certain deviation, which will affect the true value of low- and medium-resolution remote sensing products. Acquisition accuracy. This limitation leads to the lack of high-resolution data in the actual application process, so that the true value of low-resolution remote sensing products cannot be effectively and accurately obtained.

在空间尺度效应的数学推导中，采用泰勒级数展开公式对空间尺度效应进行了定量估算。这一过程要求遥感反演模型连续可导，因此当违背这一假设，即选择的遥感反演模型不连续或者不可导时，就无法估算空间尺度效应，从而无法获取中低分辨率的遥感产品真值。In the mathematical derivation of the spatial scale effect, the quantitative estimation of the spatial scale effect was carried out by using the Taylor series expansion formula. This process requires the remote sensing inversion model to be continuously and differentiable. Therefore, when this assumption is violated, that is, the selected remote sensing inversion model is discontinuous or non-conductive, the spatial scale effect cannot be estimated, and low- and medium-resolution remote sensing products cannot be obtained. true value.

空间异质性决定了空间尺度效应的大小，但现有技术仅考虑了类内异质性导致的尺度效应，而忽略了类间异质性导致的尺度效应。然而不同地表类型间这种明显的类间异质性才是决定空间尺度效应的关键因素，如果不能有效刻画类间异质性的影响将无法保证尺度效应校正的精度，进而影响中低分辨率遥感产品真值的获取精度。Spatial heterogeneity determines the size of the spatial scale effect, but the existing techniques only consider the scale effect caused by intra-class heterogeneity, while ignoring the scale effect caused by between-class heterogeneity. However, the obvious inter-class heterogeneity among different surface types is the key factor to determine the spatial scale effect. If the impact of inter-class heterogeneity cannot be effectively described, the accuracy of scale effect correction will not be guaranteed, which will affect the low-to-medium resolution. The acquisition accuracy of the true value of remote sensing products.

发明内容Contents of the invention

为了解决上述技术所存在的不足之处，本发明提供了一种校正尺度效应的中低分辨率遥感产品真值获取方法。In order to solve the deficiencies of the above-mentioned technologies, the present invention provides a method for obtaining the true value of medium and low resolution remote sensing products that corrects the scale effect.

为了解决以上技术问题，本发明采用的技术方案是：一种校正尺度效应的中低分辨率遥感产品真值获取方法，获取方法分为以下步骤：In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for obtaining the true value of the medium and low resolution remote sensing products that corrects the scale effect, and the obtaining method is divided into the following steps:

(1)、定量遥感产品生产：(1) Production of quantitative remote sensing products:

定量遥感产品生产具体分为5个步骤：第一、获取公里级中低分辨率遥感数据r，其中r是中低分辨率遥感传感器多个通道的观测值；第二、通过配套的地表下垫面的分类图估算不同地表类型的组分比例ω_d；第三、根据地表下垫面的分类图，针对不同的地表下垫面d，选择合适的遥感反演模型f_d，通常地表下垫面类型不同，遥感反演模型的形式或者模型系数也不同；第四、将遥感数据r输入到对应的遥感反演模型f_d中，直接获取不同地表下垫面组分对应的定量遥感产品p_d，即：p_d＝f_d(r)；第五、通过面积加权的方式获取中低分辨率遥感产品的粗略值即：The production of quantitative remote sensing products is specifically divided into five steps: first, obtain the kilometer-level medium and low resolution remote sensing data r, where r is the observation value of multiple channels of the medium and low resolution remote sensing sensor; The classification map of different surface types is used to estimate the component ratio ω _d of different surface types; thirdly, according to the classification map of the underlying surface, select the appropriate remote sensing inversion model f _d for different underlying surfaces d, usually the underlying surface Different surface types have different forms or model coefficients of the remote sensing inversion model; fourth, input the remote sensing data r into the corresponding remote sensing inversion model f _d , and directly obtain the quantitative remote sensing products p corresponding to different underlying surface components _d , that is: p _d =f _d (r); Fifth, obtain the rough value of low- and medium-resolution remote sensing products by means of area weighting which is:

$f f ((\overset{&OverBar; &OverBar;}{r r})) = = {Σ Σ}_{d d = = 11}^{c c} {ω ω}_{d d} {f f}_{d d} ((r r)) - - - - - - ((11))$

其中，c为地表不同下垫面类型的总个数；ω_d为不同地表类型的组分比例；f_d(r)表示对应的定量遥感产品；Among them, c is the total number of different underlying surface types on the surface; ω _d is the component ratio of different surface types; f _d (r) represents the corresponding quantitative remote sensing products;

(2)、模型二阶导估算：(2) Estimation of the second derivative of the model:

遥感反演函数将被作为黑箱，采用离散求解的方式估算模型二阶导和二阶偏导k_i，j，即：The remote sensing inversion function will be used as a black box, and the second-order derivative and second-order partial derivative k _{i, j} of the model will be estimated by discrete solution, namely:

${k k}_{i i,, j j} = = \frac{{\partial \partial}^{22} f f}{\partial \partial {r r}_{i i} \partial \partial {r r}_{j j}} = = \frac{\frac{\partial \partial f f}{\partial \partial {r r}_{i i}} ((r r + + {Δr Δr}_{j j})) - - \frac{\partial \partial f f}{\partial \partial {r r}_{i i}} ((r r - - {Δr Δr}_{j j}))}{22 {Δr Δr}_{j j}} - - - - - - ((22))$

其中，是模型的一阶导，可以表示为r是中低分辨率遥感传感器多个通道的观测值；Δr_i和Δr_j都是一个足够小的增量，如10^-3；表示数学上求取函数的偏导数；f表示抽象的遥感反演模型；in, is the first derivative of the model, which can be expressed as r is the observation value of multiple channels of medium and low resolution remote sensing sensors; both Δr _i and Δr _j are a sufficiently small increment, such as 10 ^-3 ; Represents the partial derivative of a mathematical function; f represents an abstract remote sensing inversion model;

(3)、空间异质性估算：(3) Estimation of spatial heterogeneity:

首先根据配套的地表下垫面的分类图，针对每一地表类型，采用分层随机采样的方式，随机布设采样点，并依赖地面观测实验来获取样本；采样点的个数需根据标准差、抽样误差以及置信度的先验知识来确定；在每一地表类型内，通过统计比较特定滞后距离分隔的观测变量的差异，计算相隔滞后距离h，然后观测变量的变异性，获取变异函数γ(h)，即：Firstly, according to the matching classification map of the underlying surface, for each surface type, adopt the method of stratified random sampling, randomly arrange sampling points, and rely on ground observation experiments to obtain samples; the number of sampling points should be based on the standard deviation, The sampling error and the prior knowledge of the confidence level are determined; within each surface type, the difference between the observed variables separated by a specific lag distance is statistically compared, the lag distance h is calculated, and then the variability of the observed variables is obtained to obtain the variation function γ( h), namely:

$γ γ ((h h)) = = \frac{11}{22 {N N}_{h h}} {Σ Σ}_{00}^{{N N}_{h h}} {((r r ((x x + + h h)) - - r r ((x x))))}^{22} - - - - - - ((33))$

其中N_h是相隔滞后距离h的观测变量数据对的个数；r(x)是观测变量值；Where N _h is the number of observed variable data pairs separated by a lag distance h; r(x) is the observed variable value;

为了刻画数据的多尺度特性，变异函数被认为是球状模型和指数模型的线性组合；通过对理论变异函数模型的拟合，获取到描述变异函数的重要参数：球状模型和指数模型的权重、基台值以及变程；相应的类内异质性可以表示为：In order to describe the multi-scale characteristics of the data, the variogram is considered to be a linear combination of the spherical model and the exponential model; through the fitting of the theoretical variogram model, important parameters describing the variogram are obtained: the weight of the spherical model and the exponential model, the basis values and ranges; the corresponding intraclass heterogeneity It can be expressed as:

${ν ν}_{i i,, j j}^{s the s} = = \frac{11}{| | ν ν {| |}^{22}} \underset{x x &Element; &Element; ν ν}{&Integral; &Integral;} \underset{y the y &Element; &Element; ν ν}{&Integral; &Integral;} γ γ | | | | x x - - y the y | | | | d d x x d d y the y - - - - - - ((44))$

其中，|v|为根据地表下垫面分类图获取的像元尺度s内的某一地表类型的面积，x和y为像元内在水平和垂直方向上的空间位置的坐标，同时，类间异质性直接利用组分比例ω_d来衡量；整个像元的空间异质性由类内异质性和类间异质性来共同表示；Among them, |v| is the area of a certain surface type within the pixel scale s obtained according to the classification map of the subsurface surface, x and y are the coordinates of the spatial position in the horizontal and vertical directions within the pixel, and at the same time, the inter-class The heterogeneity is directly measured by the component ratio _ωd ; the spatial heterogeneity of the entire pixel is jointly represented by intra-class heterogeneity and inter-class heterogeneity;

(4)、空间尺度效应估算：(4) Estimation of spatial scale effect:

通过继续分析同一区域高分辨率数据和中低分辨率数据表达同一客观事物的差异，结合现有技术方法中对空间尺度效应的数学简化近似和推导，尺度效应可以表示为：By continuing to analyze the difference between high-resolution data and low-resolution data in the same area to express the same objective thing, combined with the mathematical simplified approximation and derivation of the spatial scale effect in the existing technical methods, the scale effect It can be expressed as:

${Δ Δ}_{s the s c c a a l l e e}^{N N e e w w} = = \frac{11}{22} {Σ Σ}_{d d = = 11}^{c c} {ω ω}_{d d} (({Σ Σ}_{i i = = 11}^{n no} {Σ Σ}_{j j = = 11}^{n no} {k k}_{d d,, i i,, j j} {ν ν}_{d d,, i i,, j j}^{s the s})) - - - - - - ((55))$

其中，k_d，i，j是下垫面d对通道i和j的模型二阶偏导，是下垫面d在像元尺度s下i和j通道的类间异质性；上式中组分比例ω_d解释了类间异质性对尺度效应的影响，k_d，i，j和解释了模型非线性程度和类内异质性对尺度效应的影响；Among them, k _{d, i, j} are the second-order partial derivatives of the model of the underlying surface d with respect to channels i and j, is the inter-class heterogeneity of the i and j channels of the underlying surface d at the pixel scale s; the component ratio ω _d in the above formula explains the impact of the inter-class heterogeneity on the scale effect, k _{d, i, j} and The effect of model nonlinearity and intraclass heterogeneity on scale effects is explained;

(5)、空间尺度效应改正：(5) Correction of spatial scale effect:

结合公式(1)和(5)，利用像元内的组分比例就可以对中低分辨率反演获取的遥感产品进行空间尺度效应校正，获取中低分辨率遥感产品像元尺度真值 Combining formulas (1) and (5), using the component ratio in the pixel, the spatial scale effect correction can be performed on the remote sensing products obtained by medium and low resolution inversion, and the true value of the pixel scale of medium and low resolution remote sensing products can be obtained

${f f}_{c c o o r r r r}^{N N e e w w} = = f f ((\overset{&OverBar; &OverBar;}{r r})) + + {Δ Δ}_{s the s c c a a l l e e}^{N N e e w w} {Σ Σ}_{d d = = 11}^{c c} {ω ω}_{d d} (({f f}_{d d} ((r r)) + + \frac{11}{22} {Σ Σ}_{i i = = 11}^{n no} {Σ Σ}_{j j = = 11}^{R R} {k k}_{d d,, i i,, j j} {v v}_{d d,, i i,, j j}^{s the s})) - - - - - - ((66))$

当遥感反演模型在整个输入参数分布区间内不连续或者不可导的时候，可根据遥感反演模型不连续和不可导的位置，将输入参数分布区间重新划分为若干子区间，使得在每个子区间内遥感反演模型连续且可导，并同时估算每个子分布区间所占的比例和等效参数值，结合公式(6)估算中低分辨率遥感产品像元尺度真值；n为通道数。When the remote sensing inversion model is discontinuous or non-differentiable in the entire input parameter distribution interval, the input parameter distribution interval can be re-divided into several sub-intervals according to the discontinuous and non-differentiable positions of the remote sensing inversion model, so that in each sub-interval The remote sensing inversion model in the interval is continuous and derivable, and the proportion and equivalent parameter value of each sub-distribution interval are estimated at the same time, combined with formula (6) to estimate the true value of the pixel scale of the medium and low resolution remote sensing products; n is the number of channels .

本发明摆脱了遥感产品真值获取过程中对于配套高分辨率数据需求的束缚，突破现有技术中要求遥感反演模型连续可导的限制，同时考虑类内异质性和类间异质性导致的尺度效应问题，进一步提高中低分辨率遥感产品像元尺度真值的获取精度，最终实现中低分辨率遥感产品像元尺度真值的简便精确获取。The present invention gets rid of the shackles of matching high-resolution data requirements in the process of obtaining the true value of remote sensing products, breaks through the limitation of continuous guideability of remote sensing inversion models in the prior art, and considers intra-class heterogeneity and inter-class heterogeneity at the same time The resulting scale effect problem further improves the acquisition accuracy of the pixel-scale true value of medium and low-resolution remote sensing products, and finally realizes the simple and accurate acquisition of the pixel-scale true value of medium-low resolution remote sensing products.

附图说明Description of drawings

下面结合附图和具体实施方式对本发明作进一步详细的说明。The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

图1为本发明的流程图。Fig. 1 is a flowchart of the present invention.

具体实施方式detailed description

本发明提出的中低分辨率遥感产品真值的获取方法仍旧是结合遥感模型反演和尺度效应校正来实现的。但对于尺度效应的校正，是通过结合下垫面地表覆盖类型的分类图，在不同地表类型布设采用点的方式，利用地统计中的变异函数来估算地表空间异质性。最后结合不同地表类型的组分比例，同时校正类内和类间异质性导致的尺度效应，进而获取到中低分辨率遥感产品像元尺度真值。本发明的总体技术流程图如图1所示，分为以下五个步骤：1、定量遥感产品生产；2、模型二阶导估算；3、空间异质性估算；4、空间尺度效应估算；5、空间尺度效应改正。The method for obtaining the true value of the medium and low resolution remote sensing products proposed by the present invention is still realized by combining remote sensing model inversion and scale effect correction. However, for the correction of the scale effect, it is combined with the classification map of the underlying land surface cover type, and the method of laying out points for different surface types, and using the variation function in geostatistics to estimate the spatial heterogeneity of the land surface. Finally, the scale effect caused by intra-class and inter-class heterogeneity is corrected by combining the component proportions of different surface types, and then the true value of the pixel scale of medium and low resolution remote sensing products is obtained. The overall technical flow chart of the present invention is shown in Figure 1, is divided into following five steps: 1, quantitative remote sensing product production; 2, model second-order derivative estimation; 3, spatial heterogeneity estimation; 4, spatial scale effect estimation; 5. Correction of spatial scale effect.

1、定量遥感产品生产：1. Production of quantitative remote sensing products:

分为5个步骤：第一、获取公里级中低分辨率遥感数据r，其中r是中低分辨率遥感传感器多个通道的观测值；第二、通过配套的地表下垫面的分类图估算不同地表类型的组分比例ω_d；第三、根据地表下垫面的分类图，针对不同的地表下垫面d，选择合适的遥感反演模型f_d(通常地表下垫面类型不同，遥感反演模型的形式或者模型系数也不同)；第四、将遥感数据r输入到对应的遥感反演模型f_d中，直接获取不同地表下垫面组分对应的定量遥感产品p_d，即：p_d＝f_d(r)；第五、通过面积加权的方式获取中低分辨率遥感产品的粗略值即：It is divided into 5 steps: first, obtain the kilometer-level medium and low resolution remote sensing data r, where r is the observation value of multiple channels of the medium and low resolution remote sensing sensor; second, estimate through the matching subsurface classification map The component ratio of different surface types ω _d ; thirdly, according to the classification map of the underlying surface, select the appropriate remote sensing inversion model f _d for different underlying surfaces d (usually the types of underlying surfaces are different, and remote sensing The form or model coefficient of the inversion model is also different); fourth, input the remote sensing data r into the corresponding remote sensing inversion model f _d , and directly obtain the quantitative remote sensing product p _d corresponding to different subsurface components, namely: p _d = f _d (r); Fifth, obtain the rough value of low- and medium-resolution remote sensing products by means of area weighting which is:

其中，c为地表不同下垫面类型的总个数；ω_d为不同地表类型的组分比例；f_d(r)表示对应的定量遥感产品。Among them, c is the total number of different underlying surface types on the surface; ω _d is the component ratio of different surface types; f _d (r) represents the corresponding quantitative remote sensing products.

2、模型二阶导估算：2. Estimation of the second order derivative of the model:

3、空间异质性估算：3. Spatial heterogeneity estimation:

首先根据配套的地表下垫面的分类图，针对每一地表类型，采用分层随机采样的方式，随机布设采样点，并依赖地面观测实验来获取样本。采样点的个数需根据标准差、抽样误差以及置信度的先验知识来确定。在每一地表类型内，通过统计比较特定滞后距离分隔的观测变量的差异，计算相隔滞后距离h，然后观测变量的变异性，获取变异函数γ(h)，即：Firstly, according to the matching classification map of the underlying surface, for each surface type, the stratified random sampling method is adopted to randomly arrange sampling points, and rely on ground observation experiments to obtain samples. The number of sampling points needs to be determined according to the prior knowledge of standard deviation, sampling error and confidence. Within each land surface type, by statistically comparing the difference of observed variables separated by a specific lag distance, the lag distance h separated by the interval is calculated, and then the variability of the observed variables is obtained to obtain the variation function γ(h), namely:

其中N_h是相隔滞后距离h的观测变量数据对的个数；r(x)是观测变量值；为了刻画数据的多尺度特性，变异函数被认为是球状模型和指数模型的线性组合。通过对理论变异函数模型的拟合，获取到描述变异函数的重要参数：球状模型和指数模型的权重、基台值以及变程。相应的类内异质性可以表示为：where N _h is the number of observed variable data pairs separated by a lag distance h; r(x) is the observed variable value; in order to describe the multi-scale characteristics of the data, the variation function is considered to be a linear combination of the spherical model and the exponential model. Through the fitting of the theoretical variogram model, the important parameters describing the variogram are obtained: the weight, sill value and range of the spherical model and the exponential model. corresponding intraclass heterogeneity It can be expressed as:

其中，|v|为根据地表下垫面分类图获取的像元尺度s内的某一地表类型的面积，x和y为像元内在水平和垂直方向上的空间位置的坐标，同时，类间异质性直接利用组分比例ω_d来衡量。整个像元的空间异质性由类内异质性和类间异质性来共同表示。Among them, |v| is the area of a certain surface type within the pixel scale s obtained according to the classification map of the subsurface surface, x and y are the coordinates of the spatial position in the horizontal and vertical directions within the pixel, and at the same time, the inter-class Heterogeneity is measured directly using the component proportion _ωd . The spatial heterogeneity of the entire pixel is jointly represented by intra-class heterogeneity and between-class heterogeneity.

4、空间尺度效应估算：4. Spatial scale effect estimation:

其中，k_d，i，j是下垫面d对通道i和j的模型二阶偏导，是下垫面d在像元尺度s下i和j通道的类间异质性。上式中组分比例ω_d解释了类间异质性对尺度效应的影响，k_d，i，j和解释了模型非线性程度和类内异质性对尺度效应的影响。Among them, k _{d, i, j} are the second-order partial derivatives of the model of the underlying surface d with respect to channels i and j, is the inter-class heterogeneity of the i and j channels of the underlying surface d at the pixel scale s. The proportion of components in the above formula ω _d explains the effect of between-class heterogeneity on the scale effect, k _{d, i, j} and The degree of model nonlinearity and the effect of within-class heterogeneity on scale effects are explained.

5、空间尺度效应改正：5. Spatial scale effect correction:

当遥感反演模型在整个输入参数分布区间内不连续或者不可导的时候，可根据遥感反演模型不连续和不可导的位置，将输入参数分布区间重新划分为若干子区间，使得在每个子区间内遥感反演模型连续且可导，并同时估算每个子分布区间所占的比例和等效参数值，结合公式(6)估算中低分辨率遥感产品像元尺度真值。n为通道数。本发明上述公式中同样的符号表示同样的含义。When the remote sensing inversion model is discontinuous or non-differentiable in the entire input parameter distribution interval, the input parameter distribution interval can be re-divided into several sub-intervals according to the discontinuous and non-differentiable positions of the remote sensing inversion model, so that in each sub-interval The remote sensing inversion model in the interval is continuous and derivable, and the proportion and equivalent parameter value of each sub-distribution interval are estimated at the same time, combined with formula (6) to estimate the true value of the pixel scale of the medium and low resolution remote sensing products. n is the number of channels. The same symbols in the above formulas of the present invention represent the same meanings.

上述实施方式并非是对本发明的限制，本发明也并不仅限于上述举例，本技术领域的技术人员在本发明的技术方案范围内所做出的变化、改型、添加或替换，也均属于本发明的保护范围。The above-mentioned embodiments are not limitations to the present invention, and the present invention is not limited to the above-mentioned examples. Changes, modifications, additions or substitutions made by those skilled in the art within the scope of the technical solution of the present invention also belong to this invention. protection scope of the invention.

Claims

1. A method for obtaining the true value of medium and low resolution remote sensing products for correcting scale effects, characterized in that: the obtaining method is divided into the following steps:

(1) Production of quantitative remote sensing products:

The production of quantitative remote sensing products is specifically divided into five steps: first, obtain the kilometer-level medium and low resolution remote sensing data r, where r is the observation value of multiple channels of the medium and low resolution remote sensing sensor; The classification map of different surface types is used to estimate the component ratio ω _d of different surface types; thirdly, according to the classification map of the underlying surface, select the appropriate remote sensing inversion model f _d for different underlying surfaces d, usually the underlying surface Different surface types have different forms or model coefficients of the remote sensing inversion model; fourth, input the remote sensing data r into the corresponding remote sensing inversion model f _d , and directly obtain the quantitative remote sensing products p corresponding to different underlying surface components _d , that is: p _d =f _d (r); Fifth, obtain the rough value of low- and medium-resolution remote sensing products by means of area weighting which is:

f f ((\overset{&OverBar; &OverBar;}{r r})) = = {Σ Σ}_{d d = = 11}^{c c} {ω ω}_{d d} {f f}_{d d} ((r r)) - - - - - - ((11))

Among them, c is the total number of different underlying surface types on the surface; ω _d is the component ratio of different surface types; f _d (r) represents the corresponding quantitative remote sensing products;

(2) Estimation of the second derivative of the model:

The remote sensing inversion function will be used as a black box, and the second-order derivative and second-order partial derivative k _{i, j} of the model will be estimated by discrete solution, namely:

{k k}_{i i,, j j} = = \frac{{\partial \partial}^{22} f f}{\partial \partial {r r}_{i i} \partial \partial {r r}_{j j}} = = \frac{\frac{\partial \partial f f}{\partial \partial {r r}_{i i}} ((r r + + {Δr Δr}_{j j})) - - \frac{\partial \partial f f}{\partial \partial {r r}_{i i}} ((r r - - {Δr Δr}_{j j}))}{22 {Δr Δr}_{j j}} - - - - - - ((22))

in, is the first derivative of the model, which can be expressed as

\frac{\partial f}{\partial r_{i}} = (f (r + {Δr}_{i}) - f (r - {Δr}_{i})) / (2 {Δr}_{i});

r is the observation value of multiple channels of medium and low resolution remote sensing sensors; both Δr _i and Δr _j are a sufficiently small increment, such as 10 ^-3 ; Represents the partial derivative of a mathematical function; f represents an abstract remote sensing inversion model;

(3) Estimation of spatial heterogeneity:

First of all, according to the classification map of the underlying surface of the ground, for each surface type, adopt the method of stratified random sampling, randomly arrange sampling points, and rely on ground observation experiments to obtain samples; the number of sampling points needs to be based on the standard deviation, The sampling error and the prior knowledge of the confidence level are determined; within each surface type, the differences of the observed variables separated by a specific lag distance are statistically compared, the lag distance h is calculated, and then the variability of the observed variables is obtained to obtain the variation function γ( h), namely:

γ γ ((h h)) = = \frac{11}{22 {N N}_{h h}} {Σ Σ}_{00}^{{N N}_{h h}} {((r r ((x x + + h h)) - - r r ((x x))))}^{22} - - - - - - ((33))

Where N _h is the number of observed variable data pairs separated by a lag distance h; r(x) is the observed variable value;

In order to describe the multi-scale characteristics of the data, the variogram is considered to be a linear combination of the spherical model and the exponential model; through the fitting of the theoretical variogram model, important parameters describing the variogram are obtained: the weight of the spherical model and the exponential model, the basis values and ranges; the corresponding intraclass heterogeneity It can be expressed as:

{v v}_{i i,, j j}^{s the s} = = \frac{11}{| | v v {| |}^{22}} \underset{x x &Element; &Element; v v}{&Integral; &Integral;} \underset{y the y &Element; &Element; v v}{&Integral; &Integral;} γ γ | | | | x x - - y the y | | | | d d x x d d y the y - - - - - - ((44))

Among them, |v| is the area of a certain surface type within the pixel scale s obtained according to the classification map of the subsurface surface, x and y are the coordinates of the spatial position in the horizontal and vertical directions within the pixel, and at the same time, the inter-class The heterogeneity is directly measured by the component ratio _ωd ; the spatial heterogeneity of the entire pixel is jointly represented by intra-class heterogeneity and inter-class heterogeneity;

(4) Estimation of spatial scale effect:

By continuing to analyze the difference between high-resolution data and low-resolution data in the same area to express the same objective thing, combined with the mathematical simplified approximation and derivation of the spatial scale effect in the existing technical methods, the scale effect It can be expressed as:

{Δ Δ}_{s the s c c a a l l e e}^{N N e e w w} = = \frac{11}{22} {Σ Σ}_{d d = = 11}^{c c} {ω ω}_{d d} (({Σ Σ}_{i i = = 11}^{n no} {Σ Σ}_{j j = = 11}^{n no} {k k}_{d d,, i i,, j j} {v v}_{d d,, i i,, j j}^{s the s})) - - - - - - ((55))

Among them, k _{d, i, j} are the second-order partial derivatives of the model of the underlying surface d with respect to channels i and j, is the inter-class heterogeneity of the i and j channels of the underlying surface d at the pixel scale s; the component ratio ω _d in the above formula explains the impact of the inter-class heterogeneity on the scale effect, k _{d, i, j} and The effect of model nonlinearity and intraclass heterogeneity on scale effects is explained;

(5) Correction of spatial scale effect:

Combining formulas (1) and (5), using the component ratio in the pixel, the spatial scale effect correction can be performed on the remote sensing products obtained by medium and low resolution inversion, and the true value of the pixel scale of medium and low resolution remote sensing products can be obtained

{f f}_{c c o o r r r r}^{N N e e w w} = = f f ((\overset{&OverBar; &OverBar;}{r r})) + + {Δ Δ}_{s the s c c a a l l e e}^{N N e e w w} = = {Σ Σ}_{d d = = 11}^{c c} {ω ω}_{d d} (({f f}_{d d} ((r r)) + + \frac{11}{22} {Σ Σ}_{i i = = 11}^{n no} {Σ Σ}_{j j = = 11}^{n no} {k k}_{d d,, i i,, j j} {v v}_{d d,, i i,, j j}^{s the s})) - - - - - - ((66))

When the remote sensing inversion model is discontinuous or non-differentiable in the entire input parameter distribution interval, the input parameter distribution interval can be re-divided into several sub-intervals according to the discontinuous and non-differentiable positions of the remote sensing inversion model, so that in each sub-interval The remote sensing inversion model in the interval is continuous and derivable, and the proportion and equivalent parameter value of each sub-distribution interval are estimated at the same time, combined with formula (6) to estimate the true value of the pixel scale of the medium and low resolution remote sensing products; n is the number of channels .