Hyperspectral image band selection method, apparatus, and electronic device

Abstract

The application provides a hyperspectral image band selection method, device and electronic equipment. The method comprises: sampling each band image of a hyperspectral image in a three-dimensional space to obtain a sampling two-dimensional matrix; calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, selecting a sampling band corresponding to the maximum information entropy in each group to form an initial clustering center matrix; performing bias processing on the initial clustering center matrix to obtain a target clustering center matrix; based on the sampling two-dimensional matrix and the target clustering center matrix, calculating the membership degree of each sampling band in the sampling two-dimensional matrix to each clustering center in the target clustering center matrix to obtain an initial membership matrix; calculating a kernel matrix based on the sampling two-dimensional matrix; and determining a target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix. The method is efficient in calculation and improves the local optimal solution problem.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus and electronic device for selecting bands in hyperspectral images. Background Technology

[0002] Hyperspectral remote sensing technology can acquire hyperspectral images with more channels. Spectral values ​​from different bands of the same pixel can form an approximately continuous spectral curve, enhancing the ability to distinguish ground features. While more detailed spectral information about ground features is obtained based on finer spectral resolution, the data volume also increases compared to multispectral images, leading to longer data transmission and subsequent processing times and reduced efficiency. Furthermore, the high correlation and dependence between bands result in significant computational complexity and the Hughes effect, making dimensionality reduction necessary before data use.

[0003] Currently, there are two main methods for dimensionality reduction: feature extraction and band selection (also known as band selection). Band selection involves choosing a subset of bands from all bands based on certain conditions. This subset represents the original data. Compared to feature extraction methods, band selection preserves the physical meaning of the values ​​in each band of the original data, thus meeting specific application requirements.

[0004] In related technologies, hyperspectral image band selection techniques suffer from several drawbacks. Sparse-based methods yield non-unique results; search-based and deep learning-based methods are computationally time-consuming; and ranking-based methods cannot effectively overcome band correlation. While clustering methods can address these issues, most are based on hard clustering, which, due to the complexity of spectral bands, may result in inaccurate results. Fuzzy C-means clustering (FCM) is a type of soft clustering, but its computation requires iteration, consuming considerable time, and its results are often only locally optimal. Summary of the Invention

[0005] This invention provides a method, apparatus, and electronic device for selecting bands in hyperspectral images, which addresses the shortcomings of existing technologies such as long computation time, low efficiency, and the generation of local optima. It realizes a band selection method that simplifies the calculation process, improves computational efficiency, and solves local optima problems while achieving high-quality band selection.

[0006] This invention provides a method for selecting bands in a hyperspectral image, comprising:

[0007] In three-dimensional space, each band of the hyperspectral image is sampled to obtain a two-dimensional sampling matrix;

[0008] Calculate the information entropy of each sampling band in the sampling two-dimensional matrix and group the information entropy. Select the sampling band with the maximum information entropy in each group to form an initial cluster center matrix.

[0009] The initial cluster center matrix is ​​biased to obtain the target cluster center matrix;

[0010] Based on the sampling two-dimensional matrix and the target cluster center matrix, the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix is ​​calculated to obtain the initial membership degree matrix;

[0011] The kernel matrix is ​​calculated based on the sampling two-dimensional matrix.

[0012] Based on the kernel matrix and the initial membership matrix, the target band selected in the hyperspectral image is determined.

[0013] According to the present invention, a method for selecting bands in a hyperspectral image includes calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, and selecting the sampling band corresponding to the maximum information entropy in each group to form an initial clustering center matrix, comprising:

[0014] Calculate the information entropy of each sampling band in the two-dimensional sampling matrix;

[0015] All the information entropies are grouped according to a preset selection number, wherein each group includes at least two of the information entropies;

[0016] Select the index of the sampling band corresponding to the maximum information entropy in each group to obtain the index set;

[0017] The initial cluster center matrix is ​​formed by selecting the corresponding sampling bands based on the index set.

[0018] According to a hyperspectral image band selection method provided by the present invention, the method of grouping all the information entropy according to a preset selection quantity includes uniform grouping.

[0019] According to a hyperspectral image band selection method provided by the present invention, determining the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix includes:

[0020] The initial membership matrix is ​​updated iteratively based on the kernel matrix until a preset condition is met, at which point the update stops, and the target membership matrix is ​​obtained.

[0021] In the target membership matrix, the index of the sampling band with the highest membership degree is selected for each cluster center to obtain the target index set;

[0022] The target band selected in the hyperspectral image is determined based on the target index set.

[0023] According to a hyperspectral image band selection method provided by the present invention, the preset conditions include: a preset number of iterations or a preset threshold.

[0024] The process of iteratively updating the initial membership matrix based on the kernel matrix until a preset condition is met, and then stopping the update to obtain the target membership matrix, includes:

[0025] The initial membership matrix is ​​updated iteratively based on the kernel matrix;

[0026] Record the number of iterations.

[0027] Calculate the absolute difference matrix between the updated membership matrices obtained after two consecutive iterations;

[0028] Compare the maximum value in the absolute difference matrix with the preset threshold;

[0029] The update stops when the number of iterations reaches the preset number of iterations, or when the maximum value in the absolute difference matrix is ​​less than the preset threshold.

[0030] The updated membership matrix mentioned above is the target membership matrix.

[0031] According to the present invention, a method for selecting bands in a hyperspectral image, wherein sampling each band of the hyperspectral image to obtain a two-dimensional sampling matrix includes:

[0032] A two-dimensional sampling matrix is ​​obtained by sampling each band of the hyperspectral image using a preset sampling method, wherein the preset sampling method includes one of row sampling method, column sampling method and cross sampling method.

[0033] The present invention also provides a hyperspectral image band selection device, comprising:

[0034] The sampling module is used to sample each band of the hyperspectral image in three-dimensional space to obtain a two-dimensional sampling matrix;

[0035] The calculation selection module is used to calculate the information entropy of each sampling band in the sampling two-dimensional matrix, group the information entropy, and select the sampling band corresponding to the maximum information entropy in each group to form an initial cluster center matrix.

[0036] The bias processing module is used to bias the initial cluster center matrix to obtain the target cluster center matrix;

[0037] The first calculation module is used to calculate the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix based on the sampling two-dimensional matrix and the target cluster center matrix to obtain an initial membership degree matrix;

[0038] The second calculation module is used to calculate the kernel matrix based on the sampling two-dimensional matrix;

[0039] A determination module is used to determine the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix.

[0040] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the hyperspectral image band selection method as described above.

[0041] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the hyperspectral image band selection method as described above.

[0042] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the hyperspectral image band selection method as described above.

[0043] The hyperspectral image band selection method, apparatus, and electronic device provided by this invention reduce the amount of computational data by sampling hyperspectral images in three-dimensional space, simplify the calculation process by using a kernel matrix calculated through a kernel function, and improve the efficiency of band selection; by calculating the information entropy of the bands, the problem of local optima is solved to a certain extent. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0045] Figure 1 This is a flowchart illustrating the hyperspectral image band selection method provided by the present invention;

[0046] Figure 2 This is a flowchart illustrating the method for obtaining the initial cluster center matrix provided by the present invention;

[0047] Figure 3 A flowchart illustrating the method for obtaining the target band provided by this invention;

[0048] Figure 4 This is a flowchart illustrating the method for obtaining the target membership matrix provided by the present invention;

[0049] Figure 5 This is a schematic diagram of the row sampling method provided by the present invention;

[0050] Figure 6 This is a schematic diagram of the column sampling method provided by the present invention;

[0051] Figure 7 This is a schematic diagram of the cross-sampling method provided by the present invention;

[0052] Figure 8 This is a schematic diagram of the hyperspectral image band selection device provided by the present invention;

[0053] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0055] The following is combined Figures 1-6 This invention describes a hyperspectral image band selection method, the method comprising:

[0056] S101: In three-dimensional space, each band of the hyperspectral image is sampled to obtain a two-dimensional sampling matrix.

[0057] Specifically, the spectral resolution is 10 -2 Spectral images within the order of λ are called hyperspectral images. In three-dimensional space, hyperspectral images... Each band image H i Sampling is performed to obtain the sampling bands of each band image.

[0058]

[0059] Among them, H i This is the image of the i-th band of the hyperspectral image, and the subscript row indicates the preset sampling method.

[0060] The multiple sampling bands obtained through the above formula are used to form a two-dimensional sampling matrix X. For example, the band vectors of the sampling bands of the first band image can be used as the first column of the two-dimensional matrix, the band vectors of the sampling bands of the second band image can be used as the second column, the band vectors of the sampling bands of the third band image can be used as the third column, and so on. Alternatively, the band vectors of the sampling bands of the first band image can be used as the first row of the two-dimensional matrix, the band vectors of the sampling bands of the second band image can be used as the second row, the band vectors of the sampling bands of the third band image can be used as the third row, and so on.

[0061] S102: Calculate the information entropy of each sampling band in the sampling two-dimensional matrix, group the information entropy, and select the sampling band corresponding to the maximum information entropy in each group to form an initial cluster center matrix.

[0062] Specifically, the information entropy of each sampling band is calculated, and the multiple information entropies are grouped according to the order of the sampling bands. The sampling band corresponding to the maximum information entropy in each group is the cluster center, and multiple cluster centers form an initial cluster center matrix.

[0063] S103: The initial cluster center matrix is ​​biased to obtain the target cluster center matrix.

[0064] Specifically, the biasing process for the initial cluster center matrix C is as follows: subtract a number near 0 from each element in the initial cluster center matrix. This number can be an integer or a non-integer.

[0065] S104: Based on the sampling two-dimensional matrix and the target cluster center matrix, calculate the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix to obtain the initial membership degree matrix.

[0066] Specifically, the initial membership matrix U is obtained by calculating the membership degree of each sampled band in the sampled two-dimensional matrix to each cluster center in the target cluster center matrix using the sampled two-dimensional matrix and the target cluster center matrix. The first row and first column of the initial membership matrix represent the membership degree of the first sampled band in the sampled two-dimensional matrix to the first cluster center in the target cluster center matrix; the first row and second column represent the membership degree of the first sampled band in the sampled two-dimensional matrix to the second cluster center in the target cluster center matrix; and so on. Similarly, the second row and first column represent the membership degree of the second sampled band in the sampled two-dimensional matrix to the first cluster center in the target cluster center matrix; the second row and second column represent the membership degree of the second sampled band in the sampled two-dimensional matrix to the second cluster center in the target cluster center matrix; and so on.

[0067] For example, the membership degree u of the i-th band to the j-th cluster center ij The following formula is used to calculate:

[0068]

[0069] and u ij The following constraints must be met:

[0070]

[0071] Where, x i Let c be the vector of the i-th sampling band. j Let c be the vector of the j-th cluster center in the target cluster center matrix. k Let u be the k-th cluster center vector in the target cluster center matrix. ij denoted as the membership degree of the i-th sampling band to the j-th cluster center, m is the number of cluster centers in the target cluster center matrix, which is also the number of target bands to be selected, or the number of groups for information entropy grouping, n is the number of band images in the hyperspectral image, and q is the preset weighting index.

[0072] When the calculated initial membership matrix does not meet the above constraints, data processing is required. Specifically, for a given sampling band, sum the membership degrees of that band with respect to the m cluster centers to obtain a total membership degree. Then, divide each of the m membership degrees of that band with respect to the m cluster centers by the total membership degree. During processing, either only the membership degrees of the sampling bands that do not meet the constraints are processed, or the entire initial membership matrix U can be processed uniformly.

[0073] S105: Calculate the kernel matrix based on the sampling two-dimensional matrix.

[0074] Specifically, kernel functions include linear kernel functions. Polynomial kernel function Gaussian kernel function wait.

[0075] For example, any one of the kernel functions k(x) mentioned above can be selected. i ,x j The kernel matrix K is calculated based on the sampling two-dimensional matrix.

[0076] K = [K ij =k(x i ,x j )] n×n

[0077] Where, x iLet x be the vector of the i-th sampling band. j Let K be the vector of the j-th sampling band, n be the number of band images in the hyperspectral image, which is also the number of sampling bands, and K be the vector of the j-th sampling band. ij This represents the element in the i-th row and j-th column of the kernel matrix.

[0078] S106: Based on the kernel matrix and the initial membership matrix, determine the target band selected in the hyperspectral image.

[0079] The hyperspectral image band selection method provided by this invention reduces the amount of computational data by sampling the hyperspectral image in three-dimensional space, simplifies the calculation process by using the kernel matrix calculated by the kernel function, and improves the efficiency of band selection; by calculating the information entropy of the band, it solves the problem of local optima to a certain extent.

[0080] In one embodiment, the step of calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, and selecting the sampling band corresponding to the maximum information entropy in each group to form an initial cluster center matrix includes:

[0081] S201: Calculate the information entropy of each sampling band in the sampling two-dimensional matrix.

[0082] Specifically, the first step is to calculate the information entropy of each sampling band in the two-dimensional sampling matrix:

[0083]

[0084] Where Ω is the color grayscale space, p(ω) represents the probability of a pixel with grayscale value ω appearing in the image, which can be calculated based on the image's grayscale histogram, x i Let be the vector of the i-th sampling band.

[0085] S202: Group all the information entropies according to a preset selection number, wherein each group includes at least two of the information entropies.

[0086] Specifically, the preset selection quantity is the number of target bands to be selected in the hyperspectral image, which is also the number of cluster centers m mentioned above. When grouping, the number of groups is the same as the number of bands to be selected. The information entropy is grouped in order according to the preset selection quantity, and each group must include at least two information entropies.

[0087] For example, define ie i =H(x) i If the set of information entropy for all bands is IE = {ie1, ie2, ..., ie...}, then the set of information entropy for all bands is IE = {ie1, ie2, ..., ie...}. n}, GIE j Then it represents the set of the j-th information entropy group.

[0088] S203: Select the index of the sampling band corresponding to the maximum information entropy in each group to obtain the index set.

[0089] Specifically, in the information entropy grouping, the index information set I0 is obtained by selecting the index information of the sampling band corresponding to the maximum information entropy in each group. Among them, GIE j Let I be the set of the j-th information entropy groups, m be the number of groups, and I0 be the set of indices of the maximum information entropy obtained from each of the m information entropy groups.

[0090] S204: Select the corresponding sampling bands according to the index set to form the initial cluster center matrix.

[0091] Specifically, after obtaining the index set I0, the corresponding sampling bands can be selected from all sampling bands of the sampling two-dimensional matrix to form the initial cluster center matrix based on the index information in the index set I0.

[0092] In one embodiment, grouping all the information entropy according to a preset selection number includes uniform grouping.

[0093] Specifically, information entropy can be evenly grouped sequentially according to a preset selection quantity. During even grouping, if the information entropy cannot be evenly divided, it is rounded down. For example, if there are 100 information entropies, and each group is evenly divided into 13 information entropies, then there are 7 groups. In this case, only the first 91 information entropies out of the 100 information entropies are grouped, and the last 9 information entropies are discarded. Evenly grouping the information entropy sequentially ensures that the cluster centers selected based on the maximum information entropy after grouping can better cover the global band, and also reduces the complexity of writing the grouping program.

[0094] In one embodiment, determining the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix includes:

[0095] S301: The initial membership matrix is ​​updated iteratively based on the kernel matrix until a preset condition is met, and then the update is stopped to obtain the target membership matrix.

[0096] Specifically, the preset condition is a pre-defined condition used to constrain the stopping of iterative calculations. Once the preset condition is met, the iterative calculations will stop.

[0097] S302: In the target membership matrix, select the index of the sampling band corresponding to the highest membership degree for each cluster center to obtain the target index set.

[0098] Specifically, in the target membership matrix, the index information of the sampling band corresponding to the highest membership degree is selected for each cluster center, resulting in the target index set I. Among them, u j Let be the membership vector of the j-th cluster center, m be the number of cluster centers, and I represent the set of subscripts of the maximum membership obtained for each of the m cluster centers, i.e., the target index set, where each subscript has a corresponding target band.

[0099] S303: Determine the target band selected in the hyperspectral image based on the target index set.

[0100] After obtaining the target index set I, the corresponding target band can be selected from the hyperspectral image based on the index information in the target index set I.

[0101] In one embodiment, the preset conditions include: a preset number of iterations or a preset threshold;

[0102] The process of iteratively updating the initial membership matrix based on the kernel matrix until a preset condition is met, and then stopping the update to obtain the target membership matrix, includes:

[0103] S401: The initial membership matrix is ​​updated based on the kernel matrix through iterative calculation.

[0104] Specifically, preset conditions can include a preset number of iterations V or a preset threshold ε. V is a fixed constant, typically between 30 and 100. Generally, more iterations result in higher accuracy, but also longer computation time. Typically, 30 iterations are sufficient to obtain a reasonably effective result. Considering both accuracy and computation time, 50 iterations are optimal, as the accuracy difference between 50 and 100 iterations is minimal, while the computation time for 50 iterations is significantly shorter than that for 100 iterations. The preset threshold ε is a fixed constant, typically 0.001 or 0.0001.

[0105] For example, the calculation process of iterative computation is as follows:

[0106]

[0107] in, u ij Let x be the membership degree of the i-th sampling band with respect to the j-th cluster center. i Let c be the vector of the i-th sampling band. j Let c be the vector of the j-th cluster center in the target cluster centers. kLet be the k-th cluster center vector in the target cluster center matrix, m be the number of cluster centers in the target cluster center matrix (i.e., the number of target bands to be selected, which is also the number of groups for information entropy grouping), q be the preset weighting exponent, K be the kernel matrix, n be the number of band images in the hyperspectral image, and u be the number of band images in the hyperspectral image. j It is the membership vector of the j-th cluster center, T denotes the matrix transpose, and K ii This represents the element in the i-th row and i-th column of the kernel matrix.

[0108] S402: Record the number of iterations.

[0109] For example, the number of iterations t is recorded during the iterative calculation process.

[0110] S403: Calculate the absolute difference matrix between the updated membership matrices obtained after two consecutive iterations.

[0111] For example, during the iterative calculation process, the absolute difference matrix between the updated membership matrices obtained after two adjacent iterations is calculated:

[0112] |U (t+1) -U (t) |

[0113] Among them, U (t+1) U represents the updated membership matrix obtained after the (t+1)th iteration of the initial membership matrix. (t) This represents the updated membership matrix obtained after the t-th iteration of the initial membership matrix.

[0114] S404: Compare the maximum value in the absolute difference matrix with the preset threshold.

[0115] S405: When the number of iterations reaches the preset number of iterations, or when the maximum value in the absolute difference matrix is ​​less than the preset threshold, the update is stopped.

[0116] Specifically, when the number of iterations reaches the preset number of iterations, i.e., t = V, or when the maximum value in the absolute difference matrix is ​​less than the preset threshold, i.e., max|U (t+1) -U (t) When | < ε, the iterative calculation ends and the update stops.

[0117] S406: The updated membership matrix mentioned above is the target membership matrix.

[0118] In one embodiment, sampling each band of the hyperspectral image to obtain a two-dimensional sampling matrix includes:

[0119] A two-dimensional sampling matrix is ​​obtained by sampling each band of the hyperspectral image using a preset sampling method, wherein the preset sampling method includes one of row sampling method, column sampling method and cross sampling method.

[0120] For details, please refer to Figure 5-7 As shown, the preset sampling methods can include three types. Figure 5 The row sampling method shown Figure 6 The column sampling method shown and Figure 7 The cross-sampling method shown, wherein, Figure 5-7 The gray pixels in the image are the sampled pixels.

[0121] For example, row sampling samples the band image in row order, where only one pixel in every two adjacent rows is sampled. Column sampling samples in column order, where only one pixel in every two adjacent columns is sampled. In cross-sampling, only one pixel in every two adjacent rows and columns is sampled.

[0122] After sampling, the pixels obtained from sampling each band image are read and arranged into sampled band row vectors or column vectors. The sampled band column vectors or row vectors are then arranged according to the order of the band images to obtain a two-dimensional sampling matrix. When reading pixels, they can be read sequentially from left to right in row order or sequentially from top to bottom in column order; no specific limitation is made here.

[0123] Below, for reference Figure 8 The hyperspectral image band selection apparatus provided by the present invention will be described below. The hyperspectral image band selection apparatus described below corresponds to the hyperspectral image band selection method described above. The apparatus includes:

[0124] The sampling module 801 is used to sample each band of the hyperspectral image in three-dimensional space to obtain a two-dimensional sampling matrix;

[0125] The calculation selection module 802 is used to calculate the information entropy of each sampling band in the sampling two-dimensional matrix, group the information entropy, and select the sampling band corresponding to the maximum information entropy in each group to form an initial clustering center matrix.

[0126] The bias processing module 803 is used to perform bias processing on the initial cluster center matrix to obtain the target cluster center matrix;

[0127] The first calculation module 804 is used to calculate the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix based on the sampling two-dimensional matrix and the target cluster center matrix to obtain an initial membership degree matrix;

[0128] The second calculation module 805 is used to calculate the kernel matrix based on the sampling two-dimensional matrix;

[0129] The determination module 806 is used to determine the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix.

[0130] The hyperspectral image band selection device provided by this invention reduces the amount of computational data by sampling the hyperspectral image in three-dimensional space, simplifies the calculation process by using the kernel matrix calculated by the kernel function, and improves the efficiency of band selection; by calculating the information entropy of the band, it solves the problem of local optima to a certain extent.

[0131] Figure 9 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8 As shown, the electronic device may include: a processor 910, a communication interface 920, a memory 930, and a communication bus 940, wherein the processor 910, the communication interface 920, and the memory 930 communicate with each other through the communication bus 940. The processor 910 can call logic instructions in the memory 930 to execute a hyperspectral image band selection method, which includes: sampling each band of the hyperspectral image in three-dimensional space to obtain a sampling two-dimensional matrix; calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, and selecting the sampling band corresponding to the maximum information entropy in each group to form an initial cluster center matrix; biasing the initial cluster center matrix to obtain a target cluster center matrix; calculating the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix based on the sampling two-dimensional matrix and the target cluster center matrix to obtain an initial membership degree matrix; calculating a kernel matrix based on the sampling two-dimensional matrix; and determining the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix.

[0132] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0133] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the hyperspectral image band selection method provided by the above methods. The method includes: sampling each band of the hyperspectral image in three-dimensional space to obtain a sampling two-dimensional matrix; calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, and selecting the sampling band corresponding to the maximum information entropy in each group to form an initial cluster center matrix; biasing the initial cluster center matrix to obtain a target cluster center matrix; calculating the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix based on the sampling two-dimensional matrix and the target cluster center matrix to obtain an initial membership degree matrix; calculating a kernel matrix based on the sampling two-dimensional matrix; and determining the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix.

[0134] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the hyperspectral image band selection method provided by the methods described above. The method includes: sampling each band of a hyperspectral image in three-dimensional space to obtain a sampling two-dimensional matrix; calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, and selecting the sampling band corresponding to the largest information entropy in each group to form an initial cluster center matrix; biasing the initial cluster center matrix to obtain a target cluster center matrix; calculating the membership degree of each sampling band in the sampling two-dimensional matrix to each cluster center in the target cluster center matrix based on the sampling two-dimensional matrix and the target cluster center matrix to obtain an initial membership degree matrix; calculating a kernel matrix based on the sampling two-dimensional matrix; and determining the target band selected in the hyperspectral image based on the kernel matrix and the initial membership matrix.

[0135] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hyperspectral image waveband selection method characterized by, The method comprises the following steps: sampling each band image of the hyperspectral image in three-dimensional space to obtain a sampling two-dimensional matrix; calculating the information entropy of each sampling band in the sampling two-dimensional matrix, grouping the information entropy, and selecting the sampling band corresponding to the maximum information entropy in each group to form an initial clustering center matrix; performing bias processing on the initial clustering center matrix to obtain a target clustering center matrix; based on the sampling two-dimensional matrix and the target clustering center matrix, calculating the membership degree of each sampling band in the sampling two-dimensional matrix to each clustering center in the target clustering center matrix to obtain an initial membership matrix; calculating a kernel matrix based on the sampling two-dimensional matrix; updating the initial membership matrix based on the kernel matrix through iterative calculation until the preset condition is met to stop updating, and obtaining a target membership matrix; selecting the index of the sampling band corresponding to the highest membership degree for each clustering center in the target membership matrix to obtain a target index set; determining the target band selected in the hyperspectral image according to the target index set.

2. The hyperspectral image waveband selection method of claim 1, wherein, The calculation of the information entropy of each sampling band in the sampling two-dimensional matrix and the grouping of the information entropy, and the selection of the sampling band corresponding to the maximum information entropy in each group to form an initial clustering center matrix, comprises: calculating the information entropy of each sampling band in the sampling two-dimensional matrix; grouping all the information entropy according to a preset selection number, wherein each group includes at least two information entropy; selecting the index of the sampling band corresponding to the maximum information entropy in each group to obtain an index set; selecting the corresponding sampling band according to the index set to form an initial clustering center matrix.

3. The method of claim 2, wherein, The grouping of all the information entropy according to a preset selection number comprises uniform grouping.

4. The method of claim 1, wherein, The preset condition comprises a preset iteration number or a preset threshold value. The updating of the initial membership matrix based on the kernel matrix through iterative calculation until the preset condition is met to stop updating, and the obtaining of a target membership matrix, comprises: updating the initial membership matrix based on the kernel matrix through iterative calculation; recording the number of iterative calculations; calculating the absolute difference value matrix between the updated membership matrices obtained after two adjacent iterations; comparing the maximum value in the absolute difference value matrix with the preset threshold value; stopping updating when the number of iterative calculations reaches the preset iteration number, or when the maximum value in the absolute difference value matrix is less than the preset threshold value; the current updated membership matrix is the target membership matrix.

5. The hyperspectral image waveband selection method according to any one of claims 1 to 4, characterized in that, The sampling of each band image of the hyperspectral image to obtain a sampling two-dimensional matrix comprises: sampling each band image of the hyperspectral image by a preset sampling method to obtain a sampling two-dimensional matrix, wherein the preset sampling method comprises one of a row sampling method, a column sampling method, and a cross sampling method.

6. A hyperspectral image waveband selection apparatus characterized by, The method comprises the following steps: a sampling module is configured to sample each band image of the hyperspectral image in three-dimensional space to obtain a sampling two-dimensional matrix; The computing selection module is configured to compute information entropy of each sampling wave band in the sampling two-dimensional matrix, group the information entropy, and select a sampling wave band corresponding to maximum information entropy in each group to form an initial clustering center matrix; The bias processing module is configured to perform bias processing on the initial clustering center matrix to obtain a target clustering center matrix; The first computing module is configured to compute membership degrees of each sampling wave band in the sampling two-dimensional matrix to each clustering center in the target clustering center matrix based on the sampling two-dimensional matrix and the target clustering center matrix to obtain an initial membership matrix; The second computing module is configured to compute a kernel matrix based on the sampling two-dimensional matrix; The determining module is configured to update the initial membership matrix based on the kernel matrix through iterative computation until a preset condition is met to stop updating, obtain a target membership matrix, select, in the target membership matrix, an index of a sampling wave band corresponding to a highest membership degree for each clustering center to obtain a target index set, and determine a target wave band selected in the hyperspectral image according to the target index set.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the hyperspectral image wave band selection method according to any one of claims 1 to 5 when executing the program.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the hyperspectral image wave band selection method according to any one of claims 1 to 5.

9. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the hyperspectral image wave band selection method according to any one of claims 1 to 5.

Images