Pixel spectral database optimization method and related device

Abstract

The invention relates to the technical field of data processing, and provides a pixel spectral database optimization method and a related device, and the method comprises the steps: obtaining an initial pixel spectral database; screening the initial pixel spectral data in the initial pixel spectral database by adopting a linear separability quantitative analysis method to obtain a pixel spectral database subset after the first screening; and screening the pixel spectral database subset after the first screening by adopting a nonlinear separability method to obtain a pixel spectral database subset after the second screening, and obtaining an optimized pixel spectral database. According to the image element spectrum database screening method, the image element spectrum database can be screened twice from the low-latitude linear angle and the high-latitude nonlinear angle, redundant data and noise in the image element spectrum database are eliminated, the image element spectrum database is optimized, and the data quality of the image element spectrum database is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a pixel spectral database optimization method and related devices. Background Art

[0002] In the field of remote sensing image processing and hyperspectral data analysis, pixel spectral database plays a vital role. However, with the continuous advancement of spectral data acquisition technology, the initial pixel spectral database often contains a large amount of redundant data and noise, which brings great challenges to subsequent data processing and analysis.

[0003] Existing pixel spectral database optimization methods mainly rely on simple statistical analysis and threshold screening. Although they can remove some redundant data to a certain extent, they often ignore the complex relationship and intrinsic connection between spectral data, resulting in a large amount of noise and interference information in the optimized database. Summary of the invention

[0004] The embodiment of the present application provides a pixel spectral database optimization method and related devices, which can screen the pixel spectral database twice from a linear angle at a low latitude and a nonlinear angle at a high latitude, eliminate redundant data and noise in the pixel spectral database, and then optimize the pixel spectral database.

[0005] A first aspect of an embodiment of the present application provides a method for optimizing a pixel spectral database, the method comprising:

[0006] Obtaining an initial pixel spectrum database;

[0007] The linear separability quantitative analysis method is used to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain the pixel spectrum database subset after the first screening;

[0008] The nonlinear separability method is used to screen the subset of the pixel spectrum database after the first screening, and the subset of the pixel spectrum database after the second screening is obtained, and the optimized pixel spectrum database is obtained.

[0009] Furthermore, the linear separability quantitative analysis method is used to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain a subset of the pixel spectrum database after the first screening, including:

[0010] Confirming reference spectral data according to the initial pixel spectral data;

[0011] Calculating the SAM value of the initial pixel spectrum data and the corresponding reference spectrum data by using the spectral angle mapping method to obtain the initial pixel spectrum SAM value;

[0012] Normalizing the initial pixel spectrum SAM value to obtain a pixel spectrum measurement value;

[0013] The spectral similarity principle method was used to confirm the SAM threshold;

[0014] The initial pixel spectral data are screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening.

[0015] Further, the initial pixel spectral data is screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening, including:

[0016] Compare the pixel spectral measurement value with the reference spectral data, and extract the spectral data and reference spectral data corresponding to the pixel spectral measurement value with the highest similarity;

[0017] The spectral data corresponding to the pixel spectral measurement values ​​that do not meet the first condition are eliminated to obtain the pixel spectral data after the first screening, wherein the first condition is specifically that the spectral data corresponding to the extracted pixel spectral measurement value and the reference spectral data belong to the same category, the pixel spectral measurement value is the minimum value, and the pixel spectral measurement value is less than the SAM threshold.

[0018] Further, confirming reference spectral data according to the initial pixel spectral data includes:

[0019] Extracting initial pixel spectrum label information from the initial pixel spectrum data;

[0020] Classifying the initial pixel spectral data according to the initial pixel spectral label information to obtain n categories of initial pixel spectral data;

[0021] Reference values ​​are calculated for the n categories of initial pixel spectral data respectively to obtain reference spectral data corresponding to the initial pixel spectral data.

[0022] Furthermore, the pixel spectrum database subset after the first screening is screened by using a nonlinear separability method to obtain the pixel spectrum database subset after the second screening, and the optimized pixel spectrum database is obtained, including:

[0023] Extracting k nearest neighbors of the target pixel spectral data and constructing a neighborhood relationship graph, wherein the target pixel spectral data is pixel spectral data arbitrarily selected from the pixel spectral database subset after the first screening;

[0024] The Gaussian kernel function is used to calculate the edge weights of the neighborhood relationship graph;

[0025] Extracting label information of the first optimized pixel spectral data, using the Kronecker function to calculate the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data, and obtaining the pixel spectral data corresponding to the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data;

[0026] Eliminate the target pixel spectral data that does not meet the second condition, wherein the second condition is that if the label of the pixel spectral data corresponding to the maximum value of the weighted label is compared with the label of the target pixel spectral data, the labels are consistent;

[0027] Repeat the step of extracting k nearest neighbors of the target pixel spectral data, constructing a neighborhood relationship graph to eliminate the target pixel spectral data that does not meet the second condition, until all pixel points in the first optimized pixel spectral data are screened, obtaining a pixel spectral database subset after the second screening, and obtaining an optimized pixel spectral database.

[0028] In this example, the initial pixel spectral data are first screened using a linear separability quantitative analysis method to remove low-latitude redundant data and noise in the pixel spectral database, and a pixel spectral data subset after the first screening is obtained. Then, the nonlinear separability method is used to perform a second screening on the pixel spectral data after the first screening to obtain a second screened pixel spectral data subset as an optimized pixel spectral database, thereby removing the remaining high-latitude redundant data and noise in the pixel spectral database, thereby optimizing the pixel spectral database and removing most of the redundant data and noise in the pixel spectral database.

[0029] A second aspect of the embodiments of the present application provides a pixel spectrum database optimization device, the device comprising: a first acquisition unit, used to acquire an initial pixel spectrum database;

[0030] The first processing unit is used to screen the initial pixel spectrum data in the initial pixel spectrum database by using a linear separability quantitative analysis method to obtain a pixel spectrum database subset after the first screening;

[0031] The second processing unit is used to use a nonlinear separability method to screen the pixel spectrum database subset after the first screening to obtain the pixel spectrum database subset after the second screening, and obtain the optimized pixel spectrum database.

[0032] Further, in the aspect of using the linear separability quantitative analysis method to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain a pixel spectrum database subset after the first screening, the first processing unit is used to:

[0033] Confirming reference spectral data according to the initial pixel spectral data;

[0034] Calculating the SAM value of the initial pixel spectrum data and the corresponding reference spectrum data by using the spectral angle mapping method to obtain the initial pixel spectrum SAM value;

[0035] Normalizing the initial pixel spectrum SAM value to obtain a pixel spectrum measurement value;

[0036] The spectral similarity principle method was used to confirm the SAM threshold;

[0037] The initial pixel spectral data are screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening.

[0038] Further, in the aspect of using the nonlinear separability method to screen the pixel spectrum database subset after the first screening to obtain the pixel spectrum database subset after the second screening to obtain the optimized pixel spectrum database, the second processing unit is used to:

[0039] Extracting k nearest neighbors of the target pixel spectral data and constructing a neighborhood relationship graph, wherein the target pixel spectral data is pixel spectral data arbitrarily selected from the pixel spectral database subset after the first screening;

[0040] The Gaussian kernel function is used to calculate the edge weights of the neighborhood relationship graph;

[0041] Extracting label information of the first optimized pixel spectral data, using the Kronecker function to calculate the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data, and obtaining the pixel spectral data corresponding to the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data;

[0042] Eliminate the target pixel spectral data that does not meet the second condition, wherein the second condition is that if the label of the pixel spectral data corresponding to the maximum value of the weighted label is compared with the label of the target pixel spectral data, the labels are consistent;

[0043] Repeat the step of extracting k nearest neighbors of the target pixel spectral data, constructing a neighborhood relationship graph to eliminate the target pixel spectral data that does not meet the second condition, until all pixel points in the first optimized pixel spectral data are screened, obtaining a pixel spectral database subset after the second screening, and obtaining an optimized pixel spectral database.

[0044] A third aspect of an embodiment of the present application provides a terminal, comprising a processor, an input device, an output device and a memory, wherein the processor, input device, output device and memory are interconnected, wherein the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the step instructions of the pixel spectral database optimization method in the first aspect of the embodiment of the present application.

[0045] The fourth aspect of the embodiments of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps described in the pixel spectral database optimization method in the first aspect of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 A schematic diagram of the overall process steps of a pixel spectral database optimization method is provided for an embodiment of the present application;

[0048] Figure 2 A schematic diagram of the first screening steps of a pixel spectrum database optimization method is provided for an embodiment of the present application;

[0049] Figure 3 A schematic diagram of specific steps of the first screening of a pixel spectrum database optimization method is provided for an embodiment of the present application;

[0050] Figure 4 A schematic diagram of the steps of the second screening of a pixel spectrum database optimization method is provided for an embodiment of the present application;

[0051] Figure 5 A display diagram of an initial pixel spectral library in an example of a pixel spectral database optimization method is provided for an embodiment of the present application;

[0052] Figure 6 A display diagram of a reference spectrum in an example of a pixel spectrum database optimization method is provided for an embodiment of the present application;

[0053] Figure 7 A schematic diagram of an optimized pixel spectrum database in an example of a pixel spectrum database optimization method is provided for an embodiment of the present application;

[0054] Figure 8 A schematic diagram of the structure of a pixel spectrum database optimization device is provided for an embodiment of the present application;

[0055] Fig. 9 A schematic diagram of the structure of a terminal provided in an embodiment of the present application;

[0056] Reference numerals:

[0057] First acquisition unit-1, first processing unit-2, second processing unit-3. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0059] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0060] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0061] In order to better understand the pixel spectral database optimization method provided in the embodiment of the present application, the following first briefly introduces the scenario in which the pixel spectral database optimization method is applied. The quality of pixel spectra is easily affected by the quality of remote sensing images, image processing process, homogeneity of ground objects, etc., and there is still no effective quality optimization method. As a result, the acquired spectra are confused, poorly separable, and of low quality, which makes it easy to cause error accumulation in the analysis and modeling process, seriously restricting the quantitative investigation and monitoring capabilities of natural resource elements.

[0062] The pixel spectrum database optimization method is applied to the pixel spectrum database optimization device. Figure 1 A schematic diagram of the overall process steps of a pixel spectral database optimization method is shown in FIG. Figure 1 As shown, the method includes:

[0063] S1. Obtain an initial pixel spectral database. The initial pixel spectral database should contain label information and spectral information. The initial pixel optical database can be an existing pixel spectral database, or it can be extracted and produced in ENVI software by selecting a ROI region of interest.

[0064] S2, using the linear separability quantitative analysis method to screen the initial pixel spectrum data in the initial pixel spectrum database, and obtain the pixel spectrum database subset after the first screening. This embodiment is preferred, Figure 2 A schematic diagram of the first screening steps of a pixel spectrum database optimization method is shown, Figure 2 As shown, step S2 includes:

[0065] S201, confirming reference spectral data according to the initial pixel spectral data. Specifically, it includes:

[0066] S2011, classifying the initial pixel spectral data according to the characteristics of the initial pixel spectral data to obtain n categories of initial pixel spectral data. In this embodiment, by comparing the similarity between the pixel spectral data as the characteristics of the pixel spectral data, the pixel spectral data with the same similarity are classified into the same type. Specifically, it includes:

[0067] Extract initial pixel spectrum label information from the initial pixel spectrum data.

[0068] According to the initial pixel spectrum label information, the initial pixel spectrum data is classified to obtain n categories of initial pixel spectrum data.

[0069] Reference values ​​are calculated for the n categories of initial pixel spectral data respectively to obtain reference spectral data corresponding to the initial pixel spectral data.

[0070] S2012, respectively calculate reference values ​​for the n categories of initial pixel spectral data to obtain reference spectral data corresponding to the initial pixel spectral data. Specifically, calculate the average spectral curve of the pixel spectral data of each category in turn as the reference spectral data of the category, and the calculation formula of the reference spectral data is as follows:

[0071]

[0072] Among them, X is the spectral information of the initial pixel spectrum library, which is a D*N dimensional matrix, D is the number of spectral bands of the initial pixel spectrum library, and N is the total number of pixel spectra. is the spectral information of the reference spectrum, which is a D*C dimensional matrix, where C is the number of object categories. c represents any object category, c = {1, 2, 3, ..., C}, and Nc represents the number of pixel spectra of the c-th object in the initial pixel spectrum library. Xn represents any pixel spectrum in the c-th object, and Xn is a D*1 dimensional matrix, where n = {1, 2, 3, ..., Nc}. Y is the label information of the initial pixel spectrum, which is a 1*N dimensional matrix, where Yn represents the label of the n-th pixel spectrum in the initial pixel spectrum, and n = {1, 2, 3, ..., N}. is the label information of the reference spectrum, which is a 1*C dimensional matrix. Yc represents the label of the c-th type of feature. To simplify the calculation, Yc={1,2,3,...,C}. If the initial pixel spectrum library also contains time and space coordinate information. The reference spectrum can also be calculated by using the spatial fitting method of high-resolution remote sensing images, natural resource feature type vector data and initial pixel spectrum of the same period to find the reference spectrum. The obtained reference spectrum contains 1 sample for each feature category.

[0073] S202, using the spectral angle mapping method to calculate the SAM value of the initial pixel spectrum data and the reference spectrum data, to obtain the initial pixel spectrum SAM value, and to obtain the pixel spectrum SAM value F. Specifically, the calculation formula of the SAM value is as follows:

[0074]

[0075] Among them, F is the pixel spectrum SAM value, which is an N*C dimensional matrix, N is the number of initial pixel spectra, and C is the number of ground object categories. Fnc represents the SAM value between the initial pixel spectrum n and the reference spectrum c, n = {1,2,3,...,N}, c = {1,2,3,...,C}.

[0076] S203, normalize the initial pixel spectrum SAM value to obtain the pixel spectrum measurement value. Obtain the pixel spectrum measurement value. According to the minimum-maximum normalization calculation rule, obtain the normalized measurement value F so that the measurement value is within the range of [0,1]. Specifically, the normalization formula is as follows:

[0077]

[0078] in, is the pixel spectral measurement value, Fmin is the minimum value in the matrix F, and Fmax is the maximum value in the matrix F. It is an N*C dimensional matrix formed after F is normalized by MIN-MAX.

[0079] S204, using the spectrum similarity principle method to determine the SAM threshold F. According to the principle that the smaller the SAM value of the same type of spectrum, the stronger the spectrum similarity, the intra-class SAM threshold T1 is set.

[0080] S205, screening the initial pixel spectrum data according to the pixel spectrum measurement value, reference spectrum data and SAM threshold, and obtaining a pixel spectrum database subset after the first screening. Figure 3 A schematic diagram of specific steps of the first screening of a pixel spectrum database optimization method is shown, including:

[0081] S2051: Compare the pixel spectral measurement value with reference spectral data, and extract spectral data and reference spectral data corresponding to the pixel spectral measurement value with the highest similarity.

[0082] S2052, eliminating the spectral data corresponding to the pixel spectral measurement value that does not meet the first condition, and obtaining the pixel spectral data after the first screening, wherein the first condition is specifically that the spectral data corresponding to the extracted pixel spectral measurement value and the reference spectral data belong to the same category, the pixel spectral measurement value is the minimum value, and the pixel spectral measurement value is less than the SAM threshold. Specifically, the following formula is used for discrimination:

[0083] right Before making a decision, it is necessary to clarify Is it a homogeneous SAM value or a heterogeneous SAM value.

[0084]

[0085] Wherein, S is an N*C dimensional matrix. When Snc=1, it indicates that the label information of the nth initial spectrum and the cth reference spectrum are the same. When Snc=0, it indicates that the label information of the nth initial spectrum and the cth reference spectrum are different.

[0086] make Compare Is it If so, determine whether Snc is equal to 1, and If the value is less than or equal to the threshold value T1 and meets the above conditions, the pixel point is retained. After the determination is completed one by one, the refined initial pixel spectrum forms a subset A.

[0087] A is the spectral information of the spectral subset selected by the above steps, which is a D*M dimensional matrix, where D is the number of spectral bands and M is the number of pixel spectra after selection. YA is the label information of the spectral subset, which is a 1*M dimensional matrix, where YAi represents the label of the i-th pixel spectrum in A, i={1,2,3,...,M}, YAi ∈{1,2,3,...,C}.

[0088] S3. Use the nonlinear separability method to screen the subset of the pixel spectrum database after the first screening, obtain the subset of the pixel spectrum database after the second screening, and obtain the optimized pixel spectrum database.

[0089] This embodiment is preferred, Figure 4 The figure shows a schematic diagram of the second screening steps of a pixel spectral database optimization method. It should be noted that the first screening is to remove the noise of low-latitude low-spectral data, and the second screening is to remove the noise of high-spectral data of high latitude. Hyperspectral data has high-dimensional nonlinear structural characteristics. Manifold learning can effectively mine the intrinsic geometric structure of nonlinear data, mine low-dimensional manifolds, and better grasp the key features of spectral low-dimensional mapping. Figure 4 As shown, step S3 includes:

[0090] S301, extracting k nearest neighbors of the target pixel spectral data, and constructing a neighborhood relationship diagram. The target pixel spectral data is the pixel spectral data arbitrarily selected from the first optimized pixel spectral data. k is a given arbitrary number. i (i=1,...M), determine its nearest neighbors Represents the neighboring points of Ai in ascending order. The domain relationship graph G(V, E) takes point Ai as the vertex and the spectral information divergence is S SID (A i ,A j ) edges, building a symmetrical and sparse graph structure. In this sparse graph structure, each pixel point is usually connected to only a small number of pixel points by edges.

[0091]

[0092] Among them, p i =A i / ΣA,q j =A j / ΣA, Ai, Aj are any points of the pixel spectrum subset A after the first optimization, S SID (A i ,A j ) is the spectral information divergence of Ai and Aj. p and q are the probability density vectors of the spectral curves of Ai and Aj, respectively, p=(p1,p2,...,p L )q=(q1,q2,...,q L ), L is the number of bands of the pixel spectrum. is the relative entropy of Aj with respect to Ai, is the relative entropy of Ai with respect to Aj. l =Ail / ∑A i ,q l =A jl / ∑A j , l∈[1,...,L].

[0093] S302, using the Gaussian kernel function method to calculate the edge weights of the neighborhood relationship graph. Specifically, due to the constructed domain relationship graph, k nearest neighbors The contribution to Ai is often different, and this contribution is usually related to the similarity between data points, which is expressed by edge weight. The formula for calculating the Gaussian kernel function method is as follows:

[0094]

[0095] Among them, Ai is the pixel spectral data randomly selected from the first optimized pixel spectral data, Aj is the neighbor of Ai, ∥A i -A j ∥ 2 represents the distance between Ai and its nearest neighbor Aj; δ is the scale parameter. The edge weights of non-nearest neighbors are set to 0.

[0096] S303, extracting label information of the first optimized pixel spectral data, using Kronecker function to calculate the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data, and obtaining the pixel spectral data corresponding to the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data.

[0097] Specifically, the calculation formula for the maximum value of the weighted label is as follows:

[0098]

[0099] Where δ is the Kronecker function, when YA i When =c, output 1, otherwise output 0.

[0100] S304, eliminating target pixel spectral data that does not meet the second condition, wherein the second condition is that if the label of the pixel spectral data corresponding to the maximum value of the weighted label is compared with the label of the target pixel spectral data, the labels are consistent. i * =YA i , it indicates that Ai is nonlinearly reconstructible, then Ai is retained, otherwise it is removed.

[0101] S305, repeatedly extracting k nearest neighbors of the target pixel spectral data, constructing a neighborhood relationship graph to eliminate the target pixel spectral data that does not meet the second condition, until all pixel points in the first optimized pixel spectral data are screened, and obtaining a pixel spectral database subset after the second screening.

[0102] S4. Summarize the pixel spectrum database subsets after the second screening to obtain the pixel spectrum subset B after the second optimization as the optimized pixel spectrum database.

[0103] In this example, the following example is taken as a specific usage scenario. The Zhuhai No. 1 hyperspectral remote sensing image is used as the basic image, and the image is preprocessed by atmospheric correction, geometric correction and other preprocessing work.

[0104] In the first step, ArcGIS 10.6 software was used to preliminarily screen samples of 8 types of land features such as rice and corn in Nanchuan District, Chongqing, and the pixel sampling area was delineated. The sampling area range graphics were sorted by comparing with the high-precision remote sensing image map, so that the range contains as many pure pixels as possible. The Vector to ROI tool in the Region of intersect of ENVI 5.3 software was used to convert the sampling area vector range into the raster data of the interest area corresponding to the hyperspectral remote sensing image space, and the attribute fields were labeled for each raster. At the same time, the ROI separability tool was used to analyze the separability of the ROI. According to the analysis results, the spectra with a separability less than 1.8 were re-sorted to make the selected ROI as small, numerous and precise as possible. After multiple rounds of data sorting, the initial pixel spectral library was formed. The number of spectra of each type of land feature varies according to the size of the divided sampling area. The initial number of spectra of rice is the largest, 1010, and the initial number of spectra of pear trees is the smallest, 75. The initial pixel spectral library has a total of 3927, and the number of spectra of each category is shown in the following table.

[0105]

[0106] The initial pixel spectral library is shown as Figure 5 shown.

[0107] In the second step, the reference spectrum of each ground feature spectrum was calculated using the following averaging method. This work was implemented in MATLAB software.

[0108]

[0109] The obtained reference spectrum contains one sample for each type of ground feature. The sample data is displayed using ENVI5.3 software. See below for details. Figure 6 shown.

[0110] The third step is to use the spectral angle mapping (SAM) method to measure the linear separability between spectral data and to conduct a quantitative analysis of the linear separability between the initial pixel spectrum and the reference spectrum. This work is implemented in MATLAB software.

[0111] First, the spectral angle mapping method is used to calculate the SAM value F of the initial pixel spectrum and the reference spectrum. Then, the normalized metric value is obtained according to the minimum-maximum normalization method. Make the metric value in the range of [0,1]. Finally, set the intra-class threshold T1 = 0.1, compare the points with the greatest similarity between the initial spectrum and the reference spectrum, and see if they meet the same category and If the value is less than T1, the pixels that meet the conditions are retained to form the screened spectral subset A.

[0112] The number of spectra in the screened spectral subset A is 3043, and spectra with small intra-class similarity are eliminated.

[0113] The number of rice has been greatly refined. The spectral number of each category of pixels is shown in the following table.

[0114]

[0115] The fourth step is to construct the nonlinear separability between the spectral data using the graph Laplacian construction based on the Laplace eigenmap (LE) manifold learning algorithm. This work is implemented in MATLAB software.

[0116] First, the nearest neighbor k is set to 8, and the neighborhood relationship graph G(V, E) is constructed. Then, the edge weight W of the graph is calculated using the Gaussian kernel function method, and the Gaussian kernel function scale parameter δ is set to 0.01. According to the label information of the pixel spectrum subset, the Kronecker function is used to calculate the maximum value of the weighted label in the nearest neighbor, and the pixel spectrum with the same label is retained to obtain the pixel spectrum subset B after secondary optimization.

[0117] After the second optimization, the number of spectra in spectral subset B is 3019, and the pixels that cannot be reconstructed nonlinearly are eliminated. Among them, the spectra of pear trees and bamboo forests are further optimized.

[0118]

[0119] The optimized pixel spectral library is shown below Figure 7 shown.

[0120] See also Figure 8 , Figure 8 The present invention provides a schematic diagram of the structure of a pixel spectrum database optimization device. Figure 8 As shown, including:

[0121] The first acquisition unit 1 is used to acquire an initial pixel spectrum database.

[0122] The first processing unit 2 is used to screen the initial pixel spectrum data in the initial pixel spectrum database by using a linear separability quantitative analysis method to obtain a pixel spectrum database subset after the first screening.

[0123] The second processing unit 3 is used to use a nonlinear separability method to screen the pixel spectrum database subset after the first screening to obtain the pixel spectrum database subset after the second screening, and obtain an optimized pixel spectrum database.

[0124] In this example, the initial pixel spectral data are first screened by using a linear separability quantitative analysis method to remove redundant data and noise in the low latitudes of the pixel spectral database, thereby obtaining a subset of pixel spectral data after the first screening. Then, a nonlinear separability method is used to perform a second screening on the pixel spectral data after the first screening, thereby obtaining a subset of pixel spectral data after the second screening as an optimized pixel spectral database, thereby removing the remaining redundant data and noise in the high latitudes of the pixel spectral database, thereby optimizing the pixel spectral database and removing most of the redundant data and noise in the pixel spectral database.

[0125] In a possible implementation, in the aspect of using the linear separability quantitative analysis method to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain a pixel spectrum database subset after the first screening, the first processing unit is used to:

[0126] Confirming reference spectral data according to the initial pixel spectral data;

[0127] Calculate the SAM value of the reference spectral data corresponding to the initial pixel spectral data by using the spectral angle mapping method to obtain the initial pixel spectral SAM value;

[0128] Normalizing the initial pixel spectrum SAM value to obtain a pixel spectrum measurement value;

[0129] The spectral similarity principle method was used to confirm the SAM threshold;

[0130] The initial pixel spectral data are screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening.

[0131] In a possible implementation, in the aspect of using the nonlinear separability method to screen the pixel spectrum database subset after the first screening to obtain the pixel spectrum database subset after the second screening to obtain the optimized pixel spectrum database, the second processing unit is used to:

[0132] Extracting k nearest neighbors of the target pixel spectral data and constructing a neighborhood relationship graph, wherein the target pixel spectral data is pixel spectral data arbitrarily selected from the pixel spectral database subset after the first screening;

[0133] The Gaussian kernel function is used to calculate the edge weights of the neighborhood relationship graph;

[0134] Extracting label information of the first optimized pixel spectral data, using the Kronecker function to calculate the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data, and obtaining the pixel spectral data corresponding to the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data;

[0135] Eliminate the target pixel spectral data that does not meet the second condition, wherein the second condition is that if the label of the pixel spectral data corresponding to the maximum value of the weighted label is compared with the label of the target pixel spectral data, the labels are consistent;

[0136] Repeat the step of extracting k nearest neighbors of the target pixel spectral data, constructing a neighborhood relationship graph to eliminate the target pixel spectral data that does not meet the second condition, until all pixel points in the first optimized pixel spectral data are screened, obtaining a pixel spectral database subset after the second screening, and obtaining an optimized pixel spectral database.

[0137] For the above embodiments, please refer to Fig. 9 , Fig. 9 A schematic diagram of the structure of a terminal provided in an embodiment of the present application, as shown in the figure, includes a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, the processor is configured to call the program instructions, and the program includes instructions for executing the following steps;

[0138] Obtaining an initial pixel spectrum database;

[0139] The linear separability quantitative analysis method is used to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain the pixel spectrum database subset after the first screening;

[0140] The nonlinear separability method is used to screen the subset of the pixel spectrum database after the first screening, and the subset of the pixel spectrum database after the second screening is obtained, and the optimized pixel spectrum database is obtained.

[0141] In this example, the initial pixel spectral data are first screened using a linear separability quantitative analysis method to remove low-latitude redundant data and noise in the pixel spectral database, and a pixel spectral data subset after the first screening is obtained. Then, the nonlinear separability method is used to perform a second screening on the pixel spectral data after the first screening to obtain a second screened pixel spectral data subset as an optimized pixel spectral database, thereby removing the remaining high-latitude redundant data and noise in the pixel spectral database, thereby optimizing the pixel spectral database and removing most of the redundant data and noise in the pixel spectral database.

[0142] The above mainly introduces the scheme of the embodiment of the present application from the perspective of the execution process on the method side. It is understandable that in order to realize the above functions, the terminal includes a hardware structure and / or software module corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0143] The embodiment of the present application can divide the terminal into functional units according to the above method example. For example, each functional unit can be divided according to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. It should be noted that the division of units in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0144] An embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute part or all of the steps of any one of the pixel spectral database optimization methods recorded in the above method embodiments.

[0145] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program enables a computer to execute part or all of the steps of any one of the pixel spectral database optimization methods recorded in the above method embodiments.

[0146] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0147] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0148] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the device or unit can be electrical or other forms.

[0149] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0150] In addition, the functional units in the various embodiments of the application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software program modules.

[0151] If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, disk or optical disk and other media that can store program codes.

[0152] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which can include: a flash drive, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc.

[0153] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A pixel spectrum database optimization method, characterized in that: include: Obtaining an initial pixel spectrum database; The linear separability quantitative analysis method is used to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain the pixel spectrum database subset after the first screening; The nonlinear separability method is used to screen the subset of the pixel spectrum database after the first screening, and the subset of the pixel spectrum database after the second screening is obtained, and the optimized pixel spectrum database is obtained.

2. The pixel spectrum database optimization method according to claim 1, characterized in that: The linear separability quantitative analysis method is used to screen the initial pixel spectral data in the initial pixel spectral database to obtain the pixel spectral database subset after the first screening, including: Confirming reference spectral data according to the initial pixel spectral data; Calculating the SAM value of the initial pixel spectrum data and the reference spectrum data by using the spectral angle mapping method to obtain the initial pixel spectrum SAM value; Normalizing the initial pixel spectrum SAM value to obtain a pixel spectrum measurement value; The initial pixel spectral data are screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening.

3. The pixel spectrum database optimization method according to claim 2, characterized in that: The initial pixel spectral data is screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening, including: Compare the pixel spectral measurement value with the reference spectral data, and extract the spectral data and reference spectral data corresponding to the pixel spectral measurement value with the highest similarity; The spectral data corresponding to the pixel spectral measurement values ​​that do not meet the first condition are eliminated to obtain the pixel spectral data after the first screening, wherein the first condition is specifically that the spectral data corresponding to the extracted pixel spectral measurement value and the reference spectral data belong to the same category, the pixel spectral measurement value is the minimum value, and the pixel spectral measurement value is less than the SAM threshold.

4. The pixel spectrum database optimization method according to claim 2, characterized in that: Confirming reference spectral data according to the initial pixel spectral data includes: Extracting initial pixel spectrum label information from the initial pixel spectrum data; Classifying the initial pixel spectral data according to the initial pixel spectral label information to obtain n categories of initial pixel spectral data; Reference values ​​are calculated for the n categories of initial pixel spectral data respectively to obtain reference spectral data corresponding to the initial pixel spectral data.

5. The pixel spectrum database optimization method according to claim 1, characterized in that: The pixel spectrum database subset after the first screening is screened by using the nonlinear separability method to obtain the pixel spectrum database subset after the second screening, and the optimized pixel spectrum database is obtained, including: Extracting k nearest neighbors of the target pixel spectral data and constructing a neighborhood relationship graph, wherein the target pixel spectral data is pixel spectral data arbitrarily selected from the pixel spectral database subset after the first screening; The Gaussian kernel function is used to calculate the edge weights of the neighborhood relationship graph; Extracting label information of the first optimized pixel spectral data, using the Kronecker function to calculate the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data, and obtaining the pixel spectral data corresponding to the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data; Eliminate the target pixel spectral data that does not meet the second condition, wherein the second condition is that if the label of the pixel spectral data corresponding to the maximum value of the weighted label is compared with the label of the target pixel spectral data, the labels are consistent; Repeat the step of extracting k nearest neighbors of the target pixel spectral data, constructing a neighborhood relationship graph to eliminate the target pixel spectral data that does not meet the second condition, until all pixel points in the first optimized pixel spectral data are screened, obtaining a pixel spectral database subset after the second screening, and obtaining an optimized pixel spectral database.

6. A pixel spectrum database optimization device, characterized in that: include: A first acquisition unit, used for acquiring an initial pixel spectrum database; The first processing unit is used to screen the initial pixel spectrum data in the initial pixel spectrum database by using a linear separability quantitative analysis method to obtain a pixel spectrum database subset after the first screening; The second processing unit is used to use a nonlinear separability method to screen the pixel spectrum database subset after the first screening to obtain the pixel spectrum database subset after the second screening, and obtain the optimized pixel spectrum database.

7. The pixel spectrum database optimization device according to claim 6, characterized in that: In the aspect of using the linear separability quantitative analysis method to screen the initial pixel spectrum data in the initial pixel spectrum database to obtain a pixel spectrum database subset after the first screening, the first processing unit is used to: Confirming reference spectral data according to the initial pixel spectral data; Calculate the SAM value of the reference spectral data corresponding to the initial pixel spectral data by using the spectral angle mapping method to obtain the initial pixel spectral SAM value; Normalizing the initial pixel spectrum SAM value to obtain a pixel spectrum measurement value; The spectral similarity principle method was used to confirm the SAM threshold; The initial pixel spectral data are screened according to the pixel spectral measurement value, the reference spectral data and the SAM threshold to obtain a pixel spectral database subset after the first screening.

8. The pixel spectrum database optimization device according to claim 6, characterized in that: In the aspect of using the nonlinear separability method to screen the pixel spectrum database subset after the first screening to obtain the pixel spectrum database subset after the second screening to obtain the optimized pixel spectrum database, the second processing unit is used to: Extracting k nearest neighbors of the target pixel spectral data and constructing a neighborhood relationship graph, wherein the target pixel spectral data is pixel spectral data arbitrarily selected from the pixel spectral database subset after the first screening; The Gaussian kernel function is used to calculate the edge weights of the neighborhood relationship graph; Extracting label information of the first optimized pixel spectral data, using the Kronecker function to calculate the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data, and obtaining the pixel spectral data corresponding to the maximum value of weighted labels among the k nearest neighbors of the target pixel spectral data; Eliminate the target pixel spectral data that does not meet the second condition, wherein the second condition is that if the label of the pixel spectral data corresponding to the maximum value of the weighted label is compared with the label of the target pixel spectral data, the labels are consistent; Repeat the step of extracting k nearest neighbors of the target pixel spectral data, constructing a neighborhood relationship graph to eliminate the target pixel spectral data that does not meet the second condition, until all pixel points in the first optimized pixel spectral data are screened, obtaining a pixel spectral database subset after the second screening, and obtaining an optimized pixel spectral database.

9. A terminal, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute the pixel spectral database optimization method as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the pixel spectrum database optimization method according to any one of claims 1 to 5.

Images