Remote sensing data normalization method and system based on geographically weighted nonparametric regression

Abstract

The invention discloses a remote sensing data normalization method and system based on geographically weighted nonparametric regression. The method comprises the following steps: establishing a geographically weighted nonparametric regression estimation model; obtaining a target image and a reference remote sensing image set, and forming a remote sensing image pair based on the target image and the corresponding reference remote sensing image; preprocessing a target image in the remote sensing image pair to obtain a resampled target image consistent with the reference remote sensing image in spatial resolution; judging a variable coefficient in each local window in the resampled target image and the reference remote sensing image, and screening out homogeneous pixels meeting conditions based on the variable coefficient and a preset threshold to form a homogeneous pixel set; and based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set, performing pixel-by-pixel calculation on a to-be-normalized image to obtain a normalization result. Remote sensing data with large breadth, high time resolution, consistent spectrum and consistent radiation can be used as a radiation reference, and normalization of any target image is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing data processing, and in particular to a remote sensing data normalization method and system based on geographically weighted nonparametric regression. Background Art

[0002] The rapid increase in the number of earth observation satellites is a necessary factor in promoting the construction of large-scale remote sensing datasets at global and regional scales, providing reliable support for high-frequency and high-precision surface monitoring. Although the surge in data volume and the diversity of sensors have improved the ability of remote sensing monitoring, traditional radiation correction technology still faces significant challenges in the radiation consistency or normalization processing of multi-phase and multi-source remote sensing images, which limits the coordinated use of massive multi-source remote sensing data. The key core of radiation normalization is to construct the radiation conversion relationship between multi-source remote sensing data. This radiation conversion relationship is complex, nonlinear, spatiotemporally heterogeneous, and highly variable based on differences in sensor characteristics, surface conditions, imaging processes, and atmospheric conditions. Currently, the more common methods include traditional regression estimation models and machine learning methods. Traditional regression estimation models include least squares linear model and Thiessen regression, and machine learning methods include support vector machine regression and random forest. However, these methods are currently insufficient to characterize the complex nonlinear spatial and temporal heterogeneous radiation conversion relationship caused by multiple factors. Therefore, it is difficult to deal with the radiation inconsistency problem under the influence of complex factors, such as surface differences, atmospheric interference and sensor differences, and also limit the radiation consistency effect of multi-source optical remote sensing data.

[0003] In order to effectively deal with the problem of radiation inconsistency and the complexity brought about by the temporal changes of land cover types and land objects, how to accurately characterize the correlation of radiation differences under the influence of multi-factor coupling is the key technical difficulty in achieving high-precision radiation normalization of multi-source remote sensing images. Summary of the invention

[0004] The present invention aims at the shortcomings of the prior art and provides a remote sensing data normalization method, system and device based on geographically weighted nonparametric regression.

[0005] In order to solve the above technical problems, the present invention is solved by the following technical solutions:

[0006] A remote sensing data normalization method based on geographically weighted nonparametric regression includes the following steps:

[0007] Build a geographically weighted nonparametric regression estimation model, combine the geographical location of the current pixel point, and determine the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the value range space bandwidth, where the adaptive bandwidth parameter selection weight is determined by the value range space weight and the geographical space weight;

[0008] Acquiring a target image and a set of reference remote sensing images, and forming a remote sensing image pair based on the target image and the corresponding reference remote sensing image;

[0009] Preprocessing the target image in the remote sensing image alignment to obtain a resampled target image having the same spatial resolution as the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric alignment;

[0010] Determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out homogeneous pixels that meet the conditions based on the coefficient of variation and a preset threshold to form a homogeneous pixel set;

[0011] Based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set, the target image is calculated pixel by pixel to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, and then the adaptive bandwidth parameter selection weight is determined.

[0012] As an implementable method, the geographically weighted nonparametric regression estimation model is constructed by the following steps:

[0013] Combined with the geographical location of the current pixel point, a linear regression estimation model is constructed. The linear regression estimation model is transformed into a non-parametric regression by Taylor expression, combining the regression coefficient, geographic space bandwidth and range space bandwidth to obtain a non-parametric linear regression estimation model.

[0014] The nonparametric linear regression estimation model is subjected to weighted least squares processing to obtain a geographically weighted nonparametric regression estimation model, and the predicted value at a certain geographical location is obtained through the geographically weighted nonparametric regression estimation model.

[0015] As an implementable method, the linear regression estimation model is expressed as follows:

[0016]

[0017] The nonparametric linear regression estimation model is expressed as follows:

[0019] The geographically weighted nonparametric regression estimation model is expressed as follows:

[0020]

[0021] Among them, W j p (x i ,θ i) represents the adaptive bandwidth parameter selection weight, b represents the bandwidth of the value domain space, h represents the bandwidth of the geographic space, and y j Indicates the actual value, Indicates that at the geographical location θ i and the estimated or predicted value of the regression coefficient β, w j (x i ) represents the jth homogeneous pixel in the range space relative to x i The weight, w j (θ i ) represents the jth homogeneous pixel in geographic space relative to the geographic location θ i The weight, θ i represents the geographical location of the i-th location, i.e., the spatial coordinates, Y j represents the reference remote sensing image, Indicates the adaptive bandwidth parameter selection weight.

[0022] As an implementable method, the adaptive bandwidth parameter selection weight is determined by a range space weight and a geographic space weight, comprising the following steps:

[0023] The corresponding range space bandwidth is determined by the pixels to be normalized in the target image and the hyperparameters, and the range space weight is obtained by the bi-square function of the range space bandwidth and the local feature adaptation;

[0024] The geospatial bandwidth is determined based on the spatial autocorrelation coefficient threshold, and then the geospatial weight is obtained through an adaptive bisquare function;

[0025] The value range weight is expressed as follows:

[0026]

[0027] in, b i Represents the bandwidth of the value range space, according to the pixel x to be normalized in the target image i With the hyperparameter d similar Settings, which change with the current pixel gray value;

[0028] The geospatial weight is expressed as follows:

[0029]

[0030] Among them, d j (θ i ) indicates that the pixel to be normalized in the target image and the geographical location of the homogeneous pixel set are θ i The distance between h represents the geographic spatial bandwidth, which is determined by the threshold R' of the spatial autocorrelation coefficient R;

[0031]

[0032] Where N is the number of pixels, ω ij (d) is the weight matrix of distance d, x i and x j are the reference values ​​of pixel i and pixel j respectively; is the average value of all pixels. When R takes the threshold R', the distance d is defined as the geographic space bandwidth h. The geographic space bandwidth h determines the neighborhood range in the geographic space. (u i ,v i ) represents the coordinates of the i-th homogeneous pixel in the homogeneous pixel set on the target image.

[0033] As an implementable method, the preprocessing of the target image in the remote sensing image to obtain a resampled target image having the same spatial resolution as the reference remote sensing image comprises the following steps:

[0034] The target image is resampled by pixel mean, and the resampled target image has the same spatial scale as the reference remote sensing image. Through geometric registration, the target image and all reference remote sensing images have the same geometric reference information.

[0035] As an implementable method, the determination of the coefficient of variation in each local window in the resampled target image and the reference remote sensing image, and screening out qualified homogeneous pixels based on the coefficient of variation and a preset threshold to form a homogeneous pixel set includes the following steps:

[0036] The coefficient of variation in each local window of the resampled target image and the reference remote sensing image is calculated by using the coefficient of variation calculation model;

[0037] The adaptive threshold of the preset adjustable hyperparameter is expressed as: homo And σ homo =λ·std+mean, and the pixels with coefficient of variation less than the adaptive threshold are screened out to form a homogeneous pixel set. The coefficient of variation is expressed as CV, and CV<σ is screened out. homo The pixels form a homogeneous pixel set Ω homo ;

[0038] The homogeneous pixel set is expressed as follows:

[0039] Ω homo ={(SI resampling1 ,RI1),(SI resampling2 ,RI2),...,(SI resamplingn ,RI n )}

[0040] Among them, Ω homo Represents a set of homogeneous pixels, SI resampling represents the pixel in the downsampled target image, SI represents the target image, RI represents the reference remote sensing image, n represents the number of homogeneous pixels, σ homo represents the adaptive threshold, mean and std represent the mean and standard deviation of the grayscale values ​​of all pixels in the current local window respectively.

[0041] As an implementable method, the coefficient of variation calculation model is expressed as follows:

[0042]

[0043] Among them, CV represents the coefficient of variation, mean and std represent the mean and standard deviation of the grayscale of all pixels in the current local window respectively.

[0044] As an implementable method, the homogeneous pixels are calculated based on the geographically weighted nonparametric regression estimation model to obtain a normalized result, specifically:

[0045] The homogeneous pixels are calculated pixel by pixel by pixel through the geographically weighted nonparametric regression estimation model to obtain a normalized result. The normalized result of the homogeneous pixel set is expressed as follows:

[0046]

[0047] Among them, SI i (x i ,θ i ) represents the geographical location θ of the i-th pixel in the homogeneous pixel set i The normalized result is represents the adaptive bandwidth parameter selection weight, θ i represents the geographical location of the ith pixel, i.e., the spatial coordinates, x i Represents the i-th pixel of a homogeneous pixel set, SI i Represents a set of homogeneous pixels, RI j represents the reference remote sensing image, Ω homo Represents a homogeneous set of pixels.

[0048] As an implementable method, the resampling is achieved by area-weighted data resampling, specifically: calculating the actual ground area in each local window in the converted remote sensing observation image;

[0049] Based on the actual ground area in each local window divided by the sum of all pixel areas, the weight in each local window is obtained, and then the weighted average is obtained through the weight in each local window to achieve image resampling.

[0050] A remote sensing data normalization system based on geographically weighted nonparametric regression, including a model building module, a data processing module and a result calculation module;

[0051] The model building module builds a geographically weighted nonparametric regression estimation model, combines the geographical location of the current pixel point, and determines the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the range space bandwidth, wherein the adaptive bandwidth parameter selection weight is determined by the range space weight and the geographical space weight;

[0052] The data processing module is used to obtain a target image and a reference remote sensing image set, and form a remote sensing image pair based on the target image and the corresponding reference remote sensing image; preprocess the target image in the remote sensing image pair to obtain a resampled target image with a spatial resolution consistent with the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric alignment; determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out qualified homogeneous pixels based on the coefficient of variation and a preset threshold to form a homogeneous pixel set;

[0053] The result calculation module performs pixel-by-pixel calculation on the target image based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, thereby determining the adaptive bandwidth parameter selection weight.

[0054] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following method is implemented:

[0055] Build a geographically weighted nonparametric regression estimation model, combine the geographical location of the current pixel point, and determine the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the value range space bandwidth, where the adaptive bandwidth parameter selection weight is determined by the value range space weight and the geographical space weight;

[0056] Acquiring a target image and a set of reference remote sensing images, and forming a remote sensing image pair based on the target image and the corresponding reference remote sensing image;

[0057] Preprocessing the target image in the remote sensing image alignment to obtain a resampled target image having the same spatial resolution as the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric alignment;

[0058] Determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out homogeneous pixels that meet the conditions based on the coefficient of variation and a preset threshold to form a homogeneous pixel set;

[0059] Based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set, the target image is calculated pixel by pixel to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, and then the adaptive bandwidth parameter selection weight is determined.

[0060] A remote sensing data normalization device based on geographically weighted nonparametric regression comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the following method when executing the computer program:

[0061] Build a geographically weighted nonparametric regression estimation model, combine the geographical location of the current pixel point, and determine the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the value range space bandwidth, where the adaptive bandwidth parameter selection weight is determined by the value range space weight and the geographical space weight;

[0062] Acquiring a target image and a set of reference remote sensing images, and forming a remote sensing image pair based on the target image and the corresponding reference remote sensing image;

[0063] Preprocessing the target image in the remote sensing image alignment to obtain a resampled target image having the same spatial resolution as the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric alignment;

[0064] Determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out homogeneous pixels that meet the conditions based on the coefficient of variation and a preset threshold to form a homogeneous pixel set;

[0065] Based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set, the target image is calculated pixel by pixel to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, and then the adaptive bandwidth parameter selection weight is determined.

[0066] The present invention has significant technical effects due to the adoption of the above technical solution:

[0067] The present invention constructs a geographically weighted nonparametric regression estimation model, and establishes a nonparametric model in a data-driven manner, that is, the model has no fixed form and is good at processing complex nonlinear relationships, and adds a geographic space weighting term to the geographically weighted nonparametric regression estimation model to realize the modeling of the geographically weighted nonparametric regression estimation model, and in the modeling process, through the construction of spatial and spectral domain bandwidth weight functions, the mining and utilization of spatial-spectral dimension radiation correlation is realized, and the geographically weighted nonparametric regression estimation model can solve the nonlinear and spatial non-stationary modeling problems that are common in the field of earth science.

[0068] The present invention uses remote sensing data with wide bandwidth, high temporal resolution, consistent spectrum and consistent radiation as a radiation reference benchmark, and can realize the normalization of any target image.

[0069] The present invention determines the bandwidth parameter in geographically weighted nonparametric regression through adaptive bandwidth parameters, solves the problem that fixed parameter models are insufficient in coping with local models, and can determine different parameters according to data characteristics under different spatiotemporal conditions and under different conditions, thereby providing spatially different radiation conversion models to better adapt to changes in local characteristics and improve generalization performance and radiation normalization accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0071] Figure 1 It is a schematic diagram of the overall process of the method of the present invention;

[0072] Figure 2 It is a schematic diagram of the overall structure of the system of the present invention;

[0073] Figure 3 It is a schematic diagram of the comparison of the target image, the reference image and the normalization result of the present invention;

[0074] Figure 4 It is a schematic diagram of visual representation obtained by various methods of the present invention. DETAILED DESCRIPTION

[0075] The present invention is further described in detail below in conjunction with embodiments. The following embodiments are for explanation of the present invention but the present invention is not limited to the following embodiments.

[0076] Embodiment 1:

[0077] A remote sensing data normalization method based on geographically weighted nonparametric regression, such as Figure 1 As shown, the following steps are included:

[0078] S100, building a geographically weighted nonparametric regression estimation model, combining the geographical location of the current pixel point, and determining the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the range space bandwidth, wherein the adaptive bandwidth parameter selection weight is determined by the range space weight and the geographical space weight;

[0079] S200, acquiring a target image and a reference remote sensing image set, and forming a remote sensing image pair based on the target image and the corresponding reference remote sensing image;

[0080] S300, preprocessing the target image in the remote sensing image to obtain a resampled target image having the same spatial resolution as the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric registration;

[0081] S400, determining the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and selecting homogeneous pixels that meet the conditions based on the coefficient of variation and a preset threshold to form a homogeneous pixel set;

[0082] S500. Based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set, the target image is calculated pixel by pixel to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, and then the adaptive bandwidth parameter selection weight is determined.

[0083] Nonparametric regression model is a statistical method that does not assume a fixed model form when modeling and analyzing data. Unlike traditional parametric regression models, nonparametric regression does not require the structure of the model to be defined in advance, making nonparametric regression models very flexible in dealing with complex data relationships, especially when the exact relationship between the data is unclear or when the relationship is difficult to express in a simple mathematical form. Nonparametric regression methods are able to capture nonlinear patterns and structures in the data without specifying the model form in advance, so they can adapt to the structure of the data regardless of its complexity. It has a wide range of applicability and can be applied to various types of data and relationships, including continuous and discrete variables; in addition, the results of nonparametric regression models are usually easier to interpret because they can be derived directly from the data without complex mathematical assumptions.

[0084] Based on this, the present invention establishes a non-parametric model driven by data, that is, the model has no fixed form and is good at handling complex nonlinear relationships. In the process of model construction, the first law of geography is considered, that is, objects that are close in space have similarities, and a geographic spatial weighting term is constructed to realize spatial non-stationary modeling. This model will also help solve the nonlinear and spatial non-stationary modeling problems that are common in the field of earth sciences.

[0085] Without assuming any particular form of the regression function, the goal is to find an estimator of the regression function The initial nonparametric regression estimation model is as follows:

[0086]

[0087] Local parameterization is performed using a Taylor expression of order p, such as when x is close to x i When , the local parameterized linear regression estimation model is obtained, which is expressed as follows:

[0088]

[0089] Construct regression coefficient β c =f (j) (x) / c! , transform the nonparametric regression problem into a linear regression problem and generate a linear regression estimation model, which is expressed as follows:

[0090]

[0091] Based on each sample The contribution to the forecast is weighted according to and weighted by the proximity ,in, is the observation bandwidth, For a custom kernel function, the above formula is modified as follows:

[0093] It can be understood that the geographically weighted nonparametric regression model (GWNPR) of the present invention is implemented on this basis in combination with the geographically weighted regression (GWR) solution, that is, it can allow the regression coefficient to vary spatially in different locations. The specific implementation includes the following steps: This embodiment takes into account the spatial geographic location and defines the regression function as a spatial variable that depends on the geographic location. It can weight all observations according to their spatial proximity to the regression observations based on the distance decay function. Assume that a sample set is , for any point The coordinates are Its geographical location is represented by , geospatial weight , is the spatial bandwidth. Then the non-parametric regression estimation model initially constructed considering the spatial geographical location is expressed as: ; Through Taylor expression, the linear regression estimation model is transformed into a non-parametric regression by combining the regression coefficient, geographic space bandwidth and range space bandwidth, and the non-parametric linear regression estimation model is obtained, which is expressed as: ; The nonparametric linear regression estimation model is subjected to weighted least squares processing to obtain a geographically weighted nonparametric regression estimation model. The predicted value at a certain geographical location is obtained through the geographically weighted nonparametric regression estimation model. The geographically weighted nonparametric regression estimation model is expressed as: ; in, represents the adaptive bandwidth parameter selection weight, , represents the bandwidth of the value range space, represents the geospatial bandwidth, Indicates the actual value, Indicates geographical location The regression coefficient The estimated or predicted value of Indicates the value range space Homogeneous pixels relative to The weight of Represents the geographic space Homogeneous pixels relative to their geographic location The weight of Indicates The geographical location is the spatial coordinates. represents the reference remote sensing image, Indicates the adaptive bandwidth parameter selection weight.

[0099] The geographically weighted nonparametric regression estimation model constructed by this method can simplify complex relationships through modeling to find reasonable correlations, and can reasonably use spatial geographic locations in the modeling process, and then determine the weight of adaptive bandwidth parameter selection by constructing geographic spatial bandwidth and value range spatial bandwidth weights, thereby realizing the mining of spatial spectrum dimension radiation correlation and subsequent applications. Then, the data with wide bandwidth, high temporal resolution, consistent spectrum and consistent radiation are used as the radiation reference benchmark to normalize any target image.

[0100] In addition, in other embodiments, an adaptive bandwidth parameter determination method is proposed to determine the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model, which can solve the problem that the fixed parameter model is insufficient to cope with the local model. The specific scheme is: the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model is determined by the geographical space bandwidth and the value range space bandwidth. The adaptive bandwidth parameter selection weight can be expressed as:

[0101] The corresponding range space bandwidth is determined by the pixels to be normalized in the target image and the hyperparameters; the range space weight is obtained by the bi-square function of the range space bandwidth and the local feature adaptation, and the range space weight is expressed as:

[0102] The geographic spatial bandwidth is determined based on the spatial autocorrelation coefficient threshold, and then the geographic spatial weight is obtained through the adaptive bisquare function. The geographic spatial weight can be expressed as:

[0103]

[0104] That is to say, in the whole calculation process, the value space weight in the variable value space is calculated by the local feature adaptive bisquare function, and the value space bandwidth b is determined by the data characteristics of the pixels to be normalized in the current target image and the hyperparameter d similar Decision, adaptive changes based on local data characteristics, b i Represents the bandwidth of the value range space, according to the pixel x to be normalized in the target image i With the hyperparameter d similar This adaptive bandwidth setting can better fit the data and adjust the weights according to the local characteristics of the data, thereby more accurately describing the non-stationarity of the radiation transformation relationship. If the bandwidth is not set appropriately, such as too high, it will be similar to the global OLS model, which is difficult to describe the non-stationarity of the radiation transformation relationship; if it is too low, it will lead to higher variance because the model is only applicable to observations closer to the regression point; the geospatial bandwidth in geographic space is calculated by the spatial distance adaptive bisquare function, d j (θ i ) indicates that the pixel to be normalized in the target image and the geographical location of the homogeneous pixel set are θ i The distance between them, combined with the samples, can be known The geographic space bandwidth h is determined by the threshold R' of the spatial autocorrelation coefficient R;

[0105]

[0106] Where N is the number of observation points, ω ij (d) is the weight matrix of distance d, x i and x j are the observation values ​​of observation points i and j respectively; It is the average value of all observation points. When R takes the threshold R', the distance d is defined as the geographic space bandwidth h. The geographic space bandwidth h determines the "neighborhood" range in the geographic space and affects which observation points are considered when calculating the weight.

[0107] The fixed bandwidth settings involved in these two weights weight data at different locations in geographic space, which will make observations closer to the regression point have greater weights than those farther away. Observations outside the specified neighborhood are assigned zero and will not affect the local regression of the target point, better considering the impact of geographic location on the regression relationship. This method can be used to determine parameters under different conditions based on data features under different spatiotemporal conditions, better adapt to changes in local characteristics, and improve the generalization performance of the model and the accuracy of radiation normalization.

[0108] After constructing the geographically weighted nonparametric regression estimation model, the process of achieving radiation normalization through the geographically weighted nonparametric regression estimation model is as follows:

[0109] A target image and a reference remote sensing image set are obtained, and a remote sensing image pair is formed based on the target image and the corresponding reference remote sensing image; the target image in the remote sensing image pair is preprocessed to obtain a resampled target image with the same spatial resolution as the reference remote sensing image; the coefficient of variation in each local window between the resampled target image and the reference remote sensing image is determined, and qualified homogeneous pixels are screened out based on the coefficient of variation and a preset threshold to form a homogeneous pixel set; and then the target image is calculated pixel by pixel in combination with the homogeneous pixel set through a geographically weighted nonparametric regression estimation model to obtain a normalized result.

[0110] The specific process is as follows:

[0111] After obtaining the target image and the reference remote sensing image set, a remote sensing image pair is formed from the target image and the corresponding reference remote sensing image. As an example, the target image is resampled, and the resampled target image is geometrically registered with the reference remote sensing image. In other words, the resampled target image and the reference remote sensing image are in a corresponding relationship. Specifically, the target image is denoted as SI, the reference remote sensing image is denoted as RI, the reference remote sensing image is a coarse resolution image, and the points of each image in the image pair are represented by corresponding values ​​and positions (i.e., spatial coordinates). For example, SI i =(x i ,θ i ) represents the i-th point from the target image, and its geographical location is denoted by θ iThe normalized image (i.e., the target image) is resampled by pixel mean. The resampled target image has the same spatial scale and geometric features as the reference remote sensing image. The pixels in the downsampled target image can be recorded as SI resampling .

[0112] In order to match the pixels in the reference remote sensing image with the pixels in the target image, it is also necessary to screen out the homogeneous pixels that meet the conditions, that is, by calculating the coefficient of variation in each local window in the resampled target image and the reference remote sensing image and extracting the homogeneous pixels, the homogeneous pixels that meet the conditions are screened out according to the coefficient of variation and the preset threshold to form a homogeneous pixel set, which will be used as the benchmark or medium for the pixels to be normalized in the target image in the target image. The specific steps are as follows:

[0113] According to the pixels of the resampled target image, the coefficient of variation calculation model CV is used to measure the coefficient of variation in each local window in the resampled target image; the adaptive threshold σ is adjusted by the hyperparameter λ homo =λ·std+mean, filter out CV<σ homo The pixels of the same quality are composed of homogeneous pixels as the homogeneous pixel set Ω homo , where mean and std are the mean and standard deviation of the grayscale values ​​of all pixels in a local window, and the homogeneous pixel set is expressed as:

[0114] Ω homo ={(SI resampling1 ,RI1),(SI resampling2 ,RI2),...,(SI resamplingn ,RI n )}.

[0115] When dealing with the scale effect in radiation normalization, it is necessary to measure the homogeneity of pixels in order to select appropriate pixels for radiation transformation model fitting. The coefficient of variation calculation model (CV) is one of the commonly used homogeneity measurement methods. It is a standardized probability distribution or frequency distribution dispersion measurement method that can reflect the diversity of data. It can eliminate the influence of different means on the comparison of discrete degrees, and then reflect the statistical changes of pixels. Here, the coefficient of variation calculation model applied is expressed as: {(std,mean)∈R 2 :mean≠0}, mean, std are the mean and standard deviation of the grayscale values ​​of all pixels in the current local window respectively. Since the coefficient of variation calculation model comprehensively considers the deviation of all pixels from the mean, it provides more comprehensive pixel discreteness information for the homogeneous pixel screening process.

[0116] After obtaining the homogeneous pixel set, the target image is calculated pixel by pixel through the geographically weighted nonparametric regression estimation model and the homogeneous pixel set to obtain the normalized result. In the calculation process, the process of determining the weight of the adaptive bandwidth parameter selection is also included. It can be understood that this process is the process of determining the weight of the adaptive bandwidth parameter selection in the geographically weighted nonparametric regression estimation model. The corresponding domain space bandwidth is determined by the gray value difference between the target image pixel to be normalized and the homogeneous pixel and the hyperparameter; the domain space weight is obtained through the domain space bandwidth and the local feature adaptive bisquare function, and the domain space weight is expressed as:

[0117] The bandwidth b of the value range is determined by the data characteristics of the pixels to be normalized in the current target image and the hyperparameter d similar Decision, adaptive changes based on local data characteristics, b i Represents the bandwidth of the value range space, according to the pixel x to be normalized in the target image i With the hyperparameter d similar Set and change with the current pixel grayscale value.

[0118] The geospatial bandwidth is determined based on the geospatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, and then the geospatial weight is obtained through an adaptive bisquare function. The geospatial weight is expressed as:

[0119]

[0120] That is, the spatial weight of the value range needs to be determined by the geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image. In the whole process, the observation points in the model are the set of homogeneous pixels. The sample points in The geographic space bandwidth h is determined by the threshold R' of the spatial autocorrelation coefficient R;

[0121]

[0122] Finally, the adaptive bandwidth parameter selection weight is determined:

[0123] Finally, by combining the geographically weighted nonparametric regression estimation model with a homogeneous pixel set, the normalized result of pixel-by-pixel calculation of the target image is:

[0124] In addition, the technical effects of the present invention can be explained by using specific images. Figure 3 As shown, Figure 3 (a) is the target image to be normalized, Figure 3 (b) is the normalized result. Figure 3 (c) is the reference remote sensing image. As can be seen from the attached figure, the normalized result obtained after normalization processing is Figure 3 (b) and the target image to be normalized Figure 3 (a) Compared with the normalized results, there is an obvious grayscale change. Figure 3 (b) Compared with the reference remote sensing image Figure 3 Compared with (c), we can also intuitively see the similar distribution pattern and gray value dynamic range.

[0125] Experiments were conducted using NDVI data from synchronized cloud-free Landsat ETM+ and ASTER images of a certain border region in a certain area using a variety of methods. The normalized visual representation results are shown in Figure 2. Figure 4 As shown, specifically: (a) scatter plot of original data; (b) relative radiation normalization result based on least squares regression LSQ; (c) relative radiation normalization result based on object model CLM; (d) relative radiation normalization result based on support vector machine regression SVM; (e) relative radiation normalization result based on local calibration object model LCLM; (f) relative radiation normalization result based on geographically weighted nonparametric regression GWNPR of the present invention; after comparison, it can be seen that: the scatter points between the normalized results obtained by the present invention and the evaluation reference data, whose points are closely aligned with the 1:1 line, indicate that the two data are highly overlapped. The scatter points before normalization show obvious nonlinear patterns, which are consistent with the assumption of nonlinear relationship between multi-source data. Compared with other radiation conversion relationship modeling methods, such as least squares regression LSQ, global object classification model CLM, support vector machine regression SVM, and local calibration object classification model LCLM, the scatter points of the relative radiation normalization result of the modeling method GWNPR of the present invention are closer to the true value, and the fitting curve is more in line with the distribution of data. This can directly prove the superiority of the present invention in modeling between nonlinear and non-stationary data.

[0126] In one embodiment, it is necessary to explain the method of data collection: the atmospheric radiation transmission 6S model is used to simulate the atmospheric radiation transmission process, and the remote sensing observation data under different atmospheric conditions are simulated. The 6S model is mainly used to simulate the transmission process of solar radiation in the ground-atmosphere system and calculate the radiation brightness at the satellite entrance pupil. It can consider the absorption of water vapor, CO2, O3 and O2, the scattering of molecules and aerosols, and the influence of non-uniform ground and bidirectional reflectivity distribution under the assumption of a cloudless atmosphere. The input parameters of the 6S model include: the geometric conditions of the satellite, including the solar zenith angle, solar azimuth, observation zenith angle, observation azimuth, etc.; the atmospheric model defines the basic composition of the atmosphere and the temperature and humidity profile, and you can choose a preset mode or customize the input measured data; the aerosol model defines the main global aerosol parameters, such as aerosol phase function, asymmetry factor and single scattering albedo; the spectral characteristics of the sensor define the spectral response function of the sensor channel; the surface reflectivity defines the reflectivity model of the surface, including two cases of uniform surface and non-uniform surface.

[0127] Based on the spectral response function, data conversion between different sensors is achieved by calculating the spectral matching conversion factor between two sensors; data resampling is the process of converting a data set from one resolution or format to another resolution or format. Basically, the actual ground area of ​​each pixel in the converted remote sensing observation image is calculated, and the weight of each pixel is obtained based on the actual ground area of ​​each pixel divided by the sum of the areas of all pixels, and then the weighted average is obtained through the weight of each pixel to achieve image resampling.

[0128] Finally, the normalized result, i.e., the normalized image, will be used later. The normalized image will make the results of the affine reconstruction model or other models more accurate in subsequent applications.

[0129] Embodiment 2:

[0130] A remote sensing data normalization system based on geographically weighted nonparametric regression, such as Figure 2 As shown, it includes a model building module 100, a data processing module 200 and a result calculation module 300;

[0131] The model building module 100 builds a geographically weighted nonparametric regression estimation model, and determines the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model by the geographical space bandwidth and the range space bandwidth in combination with the geographical location of the current pixel point, wherein the adaptive bandwidth parameter selection weight is determined by the range space weight and the geographical space weight;

[0132] The data processing module 200 is used to obtain a target image and a reference remote sensing image set, and form a remote sensing image pair based on the target image and the corresponding reference remote sensing image; preprocess the target image in the remote sensing image pair to obtain a resampled target image with the same spatial resolution as the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric registration; determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out qualified homogeneous pixels based on the coefficient of variation and a preset threshold to form a homogeneous pixel set;

[0133] The result calculation module 300 performs pixel-by-pixel calculation on a homogeneous pixel set based on the geographically weighted nonparametric regression estimation model to obtain a normalized result.

[0134] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0135] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0136] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0137] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0138] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0139] It should be noted that:

[0140] The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0141] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for normalizing remote sensing data based on geographically weighted nonparametric regression, characterized in that: The following steps are involved: Build a geographically weighted nonparametric regression estimation model, combine the geographical location of the current pixel point, and determine the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the value range space bandwidth; Acquiring a target image and a set of reference remote sensing images, and forming a remote sensing image pair based on the target image and the corresponding reference remote sensing image; Preprocessing the target image in the remote sensing image alignment to obtain a resampled target image having the same spatial resolution as the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric alignment; Determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out homogeneous pixels that meet the conditions based on the coefficient of variation and a preset threshold to form a homogeneous pixel set; Based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set, the target image is calculated pixel by pixel to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, and then the adaptive bandwidth parameter selection weight is determined.

2. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 1, characterized in that: The geographically weighted nonparametric regression estimation model is constructed by the following steps: Combined with the geographical location of the current pixel point, a linear regression estimation model is constructed. The linear regression estimation model is transformed into a non-parametric regression by Taylor expression, combining the regression coefficient, geographic space bandwidth and range space bandwidth to obtain a non-parametric linear regression estimation model. The nonparametric linear regression estimation model is subjected to weighted least squares processing to obtain a geographically weighted nonparametric regression estimation model, and the predicted value at a certain geographical location is obtained through the geographically weighted nonparametric regression estimation model.

3. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 2, characterized in that: The linear regression estimation model is expressed as follows: The nonparametric linear regression estimation model is expressed as follows: The geographically weighted nonparametric regression estimation model is expressed as follows: in, represents the adaptive bandwidth parameter selection weight, , represents the bandwidth of the value range space, represents the geospatial bandwidth, Indicates the actual value, Indicates geographical location The regression coefficient The estimated or predicted value of Indicates the value range space Homogeneous pixels relative to The weight of Represents the geographic space Homogeneous pixels relative to their geographic location The weight of Indicates The geographical location is the spatial coordinates. represents the reference remote sensing image, Indicates the adaptive bandwidth parameter selection weight.

4. The remote sensing data normalization method based on geographically weighted nonparametric regression according to any one of claims 1 to 3, characterized in that: The adaptive bandwidth parameter selection weight is determined by the value range space weight and the geographic space weight, and includes the following steps: The corresponding range space bandwidth is determined by the pixels to be normalized in the target image and the hyperparameters, and the range space weight is obtained by the bi-square function of the range space bandwidth and the local feature adaptation; The geospatial bandwidth is determined based on the spatial autocorrelation coefficient threshold, and then the geospatial weight is obtained through an adaptive bisquare function; The value range weight is expressed as follows: in, b i Represents the bandwidth of the value range space, according to the pixel x to be normalized in the target image i With the hyperparameter d similar Settings, which change with the current pixel gray value; The geospatial weight is expressed as follows: Among them, d j (θ i ) indicates that the pixel to be normalized in the target image and the geographical location of the homogeneous pixel set are θ i The distance between h represents the geographic spatial bandwidth, which is determined by the threshold R' of the spatial autocorrelation coefficient R; Where N is the number of pixels, ω ij (d) is the weight matrix of distance d, x i and x j are the reference values ​​of pixel i and pixel j respectively; is the average value of all pixels. When R takes the threshold R', the distance d is defined as the geographic space bandwidth h. The geographic space bandwidth h determines the neighborhood range in the geographic space. (u i ,v i ) represents the coordinates of the i-th homogeneous pixel in the homogeneous pixel set on the target image.

5. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 1, characterized in that: The method of preprocessing the target image in the remote sensing image to obtain a resampled target image having the same spatial resolution as the reference remote sensing image comprises the following steps: The target image is resampled by pixel mean, and the resampled target image has the same spatial scale as the reference remote sensing image. Through geometric registration, the target image and all reference remote sensing images have the same geometric reference information.

6. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 1, characterized in that: The method comprises the following steps: determining the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and selecting homogeneous pixels that meet the conditions based on the coefficient of variation and a preset threshold value to form a homogeneous pixel set. The coefficient of variation in each local window of the resampled target image and the reference remote sensing image is calculated by using the coefficient of variation calculation model; The adaptive threshold of the preset adjustable hyperparameter is expressed as: homo And σ homo =λ·std+mean, and the pixels with coefficient of variation less than the adaptive threshold are screened out to form a homogeneous pixel set. The coefficient of variation is expressed as CV, and CV<σ is screened out. homo The pixels form a homogeneous pixel set Ω homo ; The homogeneous pixel set is expressed as follows: oh homo ={(YES resampling1 ,RI1),(YES resampling2 ,RI2),...,(YES resamplingn ,RE n )} Among them, Ω homo Represents a set of homogeneous pixels, SI resampling represents the pixel in the downsampled target image, SI represents the target image, RI represents the reference remote sensing image, n represents the number of homogeneous pixels, σ homo represents the adaptive threshold, mean and std represent the mean and standard deviation of the grayscale values ​​of all pixels in the current local window respectively.

7. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 1 or 6, characterized in that: The coefficient of variation calculation model is expressed as follows: Among them, CV represents the coefficient of variation, mean and std represent the mean and standard deviation of the grayscale of all pixels in the current local window respectively.

8. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 1, characterized in that: The homogeneous pixels are calculated based on the geographically weighted nonparametric regression estimation model to obtain a normalized result, which is specifically: The homogeneous pixels are calculated pixel by pixel by pixel through the geographically weighted nonparametric regression estimation model to obtain a normalized result. The normalized result of the homogeneous pixel set is expressed as follows: Among them, SI i (x i ,θ i ) represents the geographical location θ of the i-th pixel in the homogeneous pixel set i The normalized result is represents the adaptive bandwidth parameter selection weight, θ i represents the geographical location of the ith pixel, i.e., the spatial coordinates, x i Represents the i-th pixel in a homogeneous pixel set, RI j represents the reference remote sensing image, Ω homo Represents a homogeneous set of pixels.

9. The remote sensing data normalization method based on geographically weighted nonparametric regression according to claim 1, characterized in that: The resampling is achieved by performing data resampling by area weighting, specifically: calculating the actual ground area in each local window in the converted remote sensing observation image; Based on the actual ground area in each local window divided by the sum of all pixel areas, the weight in each local window is obtained, and then the weighted average is obtained through the weight in each local window to achieve image resampling.

10. A remote sensing data normalization system based on geographically weighted nonparametric regression, characterized in that: It includes model building module, data processing module and result calculation module; The model building module builds a geographically weighted nonparametric regression estimation model, combines the geographical location of the current pixel point, and determines the adaptive bandwidth parameter selection weight in the geographically weighted nonparametric regression estimation model through the geographical space bandwidth and the range space bandwidth, wherein the adaptive bandwidth parameter selection weight is determined by the range space weight and the geographical space weight; The data processing module is used to obtain a target image and a reference remote sensing image set, and form a remote sensing image pair based on the target image and the corresponding reference remote sensing image; preprocess the target image in the remote sensing image pair to obtain a resampled target image with a spatial resolution consistent with the reference remote sensing image, wherein the preprocessing at least includes resampling and geometric alignment; determine the coefficient of variation in each local window of the resampled target image and the reference remote sensing image, and screen out qualified homogeneous pixels based on the coefficient of variation and a preset threshold to form a homogeneous pixel set; The result calculation module performs pixel-by-pixel calculation on the target image based on the geographically weighted nonparametric regression estimation model and the homogeneous pixel set to obtain a normalized result, wherein the corresponding value range spatial bandwidth and geographic spatial bandwidth are determined through the grayscale value difference and geographic spatial position relationship between the pixels to be normalized and the homogeneous pixels in the target image, thereby determining the adaptive bandwidth parameter selection weight.

11. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

12. A remote sensing data normalization device based on geographically weighted nonparametric regression, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 9 is implemented.

Images