Transformer-based method for identifying minerals on the Martian surface

Abstract

This invention provides a Transformer-based method for identifying Martian surface minerals, comprising: dividing a preprocessed Martian mineral standard hyperspectral dataset into a training dataset, a validation dataset, and a test dataset; then constructing a mineral identification Transformer model; training the mineral identification Transformer model using the training and validation datasets to obtain an optimized model; testing the optimized model using the test dataset and determining the test classification accuracy; finally, if the test accuracy meets a preset accuracy condition, inputting the Martian mineral hyperspectral data to be identified into the optimized mineral identification Transformer model, and outputting the Martian mineral classification and identification results. This effectively improves the accuracy of Martian mineral classification and identification.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, and more specifically to a method for identifying minerals on the surface of Mars based on Transformer. Background Technology

[0002] Hyperspectral remote sensing is a primary method for mineral identification on the Martian surface. Hyperspectral remote sensing data contains rich spectral and spatial information, offering a unified image and spectrum. Different types of minerals possess unique spectral characteristics, and mineral types and compositions can be identified based on these diagnostic spectral features. Therefore, Martian hyperspectral imagery has been widely used in research on mineral identification on the Martian surface.

[0003] Currently, methods based on machine learning and deep learning, such as Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), and Support Vector Machines (SVMs), are commonly used. While these methods have shown good results in identifying minerals on the Martian surface, they still have certain limitations. For example, they are insufficient in extracting global band information and cannot fully consider long-distance dependencies between spectral bands. These limitations result in existing Martian mineral classification models having low classification accuracy and low automation, making them difficult to apply on a large scale. Summary of the Invention

[0004] The technical problem to be solved by this invention is the inability to accurately classify and identify minerals on the surface of Mars.

[0005] To address the aforementioned technical problems, this invention provides a method for identifying minerals on the Martian surface based on Transformer, specifically employing the following technical solution:

[0006] This invention provides a Transformer-based method for identifying Martian surface minerals, comprising: First, acquiring a standard hyperspectral dataset of Martian minerals. Then, preprocessing the standard hyperspectral dataset to obtain a preprocessed standard hyperspectral dataset of Martian minerals. Next, dividing the preprocessed standard hyperspectral dataset of Martian minerals into a training dataset, a validation dataset, and a test dataset according to a preset partitioning ratio. Further, constructing a mineral recognition Transformer model for classifying and identifying Martian surface minerals. Then, training the mineral recognition Transformer model using the training and validation datasets to obtain an optimized model. Testing the optimized model using the test dataset and determining the test classification accuracy. Finally, if the test accuracy meets a preset accuracy condition, inputting the hyperspectral data of the Martian minerals to be identified into the optimized mineral recognition Transformer model, and outputting the Martian mineral classification and identification results.

[0007] This method, based on the Transformer model, uses deep learning to capture the internal correlations of Martian mineral standard hyperspectral data, focusing on valuable information while ignoring unimportant information. In this way, the mineral identification Transformer model can learn richer feature representations from the Martian mineral standard hyperspectral data, thereby effectively improving the accuracy of Martian mineral classification and identification.

[0008] In one alternative implementation, the aforementioned mineral recognition Transformer model includes: an embedding mapping function, an encoder, a fully connected layer, and an output layer. The embedding mapping function consists of a band embedding function and a position embedding function, and the encoder consists of multiple encoder layers connected in series.

[0009] In this implementation, the mineral identification Transformer model with the above structure can be used to classify and identify minerals from a Martian mineral hyperspectral dataset.

[0010] In one alternative implementation, training the mineral recognition Transformer model using the training and validation datasets to obtain an optimized model includes: First, inputting the training and validation datasets into an embedding mapping function to output a first embedding representation corresponding to the training dataset and a second embedding representation corresponding to the validation dataset. Then, inputting the first and second embedding representations into an encoder to output a first encoder output corresponding to the first embedding representation and a second encoder output corresponding to the second embedding representation. Further, inputting the first and second encoder outputs into a fully connected layer to output a first fully connected layer output corresponding to the first encoder output and a second fully connected layer output corresponding to the second encoder output. Next, inputting the first and second fully connected layer outputs into an output layer to output a first predicted classification result corresponding to the first fully connected layer output and a second predicted classification result corresponding to the second fully connected layer output. Finally, constructing a loss function based on the first predicted classification result and iteratively optimizing the hyperparameters of the mineral recognition Transformer model to converge the loss function, thus obtaining the hyperparameter training results. Finally, the hyperparameter training results are optimized and adjusted based on the second predicted classification results to obtain an optimized mineral recognition Transformer model.

[0011] In one alternative implementation, the expression for the above embedding mapping function is:

[0012] Em(x) = Eb(x) + Ep(x);

[0013] In the formula, Em(x) represents the embedding mapping function, Eb(x) represents the band embedding function, Ep(x) represents the location embedding function, and x represents the input hyperspectral data. The expression for Eb(x) is:

[0014] Eb(x)=W(x;θ e );

[0015] In the formula, W represents a linear mapping, and θ e This represents the parameter for band embedding. The expression for Ep(x) above is:

[0016]

[0017] In the formula, p represents the position of the band in the hyperspectral data x, p = 0, 1, ..., 232; 2i represents the even-numbered dimension of the embedding, 2i+1 represents the odd-numbered dimension of the embedding, and d represents the total dimension of the embedding.

[0018] In one alternative implementation, the function expression of the encoder described above is:

[0019] xn = Encoder(xe);

[0020] In the formula, xn represents the output of the encoder, and xe represents the output of the embedding mapping function.

[0021] The expansion of Encoder(xe) is:

[0022]

[0023] In the formula, t (i) ,i=1,2…N represents the output of the i-th encoder layer, θ n (i) For the parameters of the i-th encoder layer, en (i) This represents the i-th encoder layer.

[0024] In one alternative implementation, the expression for the fully connected layer described above is:

[0025] xl=D{LN[γ(L(xn,θ f ))]};

[0026] In the formula, xl represents the output of the fully connected layer, and θ f Here are the parameters for the fully connected layer, where γ represents the ReLU activation function, L represents the fully connected layer, LN represents the layer normalization, and D represents the Dropout function.

[0027] In one alternative implementation, the expression for the output layer is:

[0028]

[0029] In the formula, This represents the output of the output layer, and S represents the Softmax activation function.

[0030] In one alternative implementation, the loss function is the cross-entropy loss function, and the hyperparameters of the mineral recognition Transformer model include: the number of encoder layers, the model dimension, and the number of multi-head attention branches.

[0031] In one alternative implementation, the expression for the above test classification accuracy is:

[0032]

[0033] in,

[0034]

[0035] Where ACC represents the test classification accuracy, n ′ Indicates the size of the test dataset. This indicates the results of a mineral classification test, y i ′ This indicates the results of the mineral classification standards.

[0036] In one alternative implementation, the aforementioned Mars mineral standard hyperspectral dataset includes: iron / magnesium montmorillonite spectral data, magnesium carbonate spectral data, and fir peridot spectral data; the preset partition ratio is: training dataset: validation dataset: test dataset = 8:1:1. Attached Figure Description

[0037] Figure 1 A schematic diagram illustrating the principle of the Transformer-based mineral identification method for the Martian surface provided in this embodiment of the invention;

[0038] Figure 2 A schematic flowchart of the Transformer-based mineral identification method for the Martian surface provided in this embodiment of the invention;

[0039] Figure 3 This is a schematic diagram of the structure of the mineral identification Transformer model provided in an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the mineral mapping results provided in an embodiment of the present invention. Detailed Implementation

[0041] The embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described below do not represent all embodiments consistent with this application. They are merely examples of systems and methods consistent with some aspects of this application as detailed in the claims.

[0042] Hyperspectral remote sensing is a primary method for mineral identification on the Martian surface. Hyperspectral remote sensing data contains rich spectral and spatial information, offering a unified image and spectrum. Different types of minerals possess unique spectral characteristics, and mineral types and compositions can be identified based on these diagnostic spectral features. Therefore, Martian hyperspectral imagery has been widely used in research on mineral identification on the Martian surface.

[0043] Currently, methods based on machine learning and deep learning, such as Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), and Support Vector Machines (SVMs), are commonly used. While these methods have shown good results in identifying minerals on the Martian surface, they still have certain limitations. For example, they are insufficient in extracting global band information and cannot fully consider long-distance dependencies between spectral bands. These limitations result in existing Martian mineral classification models having low classification accuracy and low automation, making them difficult to apply on a large scale.

[0044] To address the aforementioned issues, this application provides a method for identifying minerals on the Martian surface based on Transformer. Figure 1 This is a schematic diagram illustrating the principle of the Transformer-based Martian surface mineral identification method provided in this embodiment of the invention, as shown below. Figure 1 As shown, this method divides the preprocessed Martian mineral standard hyperspectral dataset into training, validation, and test datasets. Then, a mineral recognition Transformer model is trained and optimized using the training and validation datasets to obtain the optimized model. Further, the optimized model is tested using the test dataset. Finally, if the test classification accuracy meets a preset accuracy condition, the optimized Transformer model is used to classify and identify the Martian minerals from the hyperspectral data, outputting the Martian mineral classification results. In this way, the mineral recognition Transformer model can learn richer feature representations from the Martian mineral standard hyperspectral data, thereby effectively improving the accuracy of Martian mineral classification and identification.

[0045] The solutions provided in the embodiments of this application will be described below with reference to the accompanying drawings.

[0046] For details, see Figure 2 This is a flowchart illustrating the method for identifying Martian surface minerals based on Transformer provided in an embodiment of the present invention. Figure 2 As shown, the Mars surface mineral identification method based on Transformer provided by the present invention includes the following steps S101-S107:

[0047] S101. Obtain the standard hyperspectral dataset of Martian minerals.

[0048] In this embodiment of the application, the Mars mineral standard hyperspectral dataset can be a Compact Reconnaissance Imaging Spectrometer for Mars (CRI SM) dataset.

[0049] In one implementation, the Martian mineral standard hyperspectral dataset can be obtained from a standard spectral library obtained through CRI SM analysis (MICA).

[0050] In some embodiments, the Mars mineral standard hyperspectral dataset includes: iron / magnesium montmorillonite spectral data, magnesium carbonate spectral data, and fir olivine spectral data.

[0051] S102. Perform data preprocessing on the Martian mineral standard hyperspectral dataset to obtain the preprocessed Martian mineral standard hyperspectral dataset.

[0052] In some embodiments, data preprocessing of the Mars mineral standard hyperspectral dataset includes at least one of the following: cropping, smoothing, continuum removal, and normalization.

[0053] Specifically, the cropping process involves extracting hyperspectral data at the target spectral wavelength. For example, the spectral wavelength range in the MICA spectral library is 0-4 μm; cropping can extract the 1.0-2.6 μm portion as the preprocessed Mars mineral standard hyperspectral dataset. Smoothing involves denoising the Mars mineral standard hyperspectral dataset by removing sharp peaks. Continuum removal removes continuum from the hyperspectral data to highlight absorption features. Normalization normalizes the hyperspectral data values ​​to near 0, making the hyperspectral data in the Mars mineral standard hyperspectral dataset more regular.

[0054] S103. Divide the preprocessed Mars mineral standard hyperspectral dataset into three datasets according to a preset division ratio: training dataset, validation dataset, and test dataset.

[0055] Specifically, to facilitate the training and optimization of the mineral identification Transformer model, this application embodiment can divide the preprocessed Mars mineral standard hyperspectral dataset into: a training dataset, a validation dataset, and a test dataset.

[0056] In some embodiments, the preset partitioning ratio can be 8:1:1 for the training dataset, validation dataset, and test dataset. This allows for effective training of the mineral recognition Transformer model using a larger training dataset, thereby improving the accuracy of the model in classifying and identifying minerals on the Martian surface.

[0057] In one implementation, the training dataset can be represented by X1. Where n1 is the number of hyperspectral data points in the training dataset, and k1 is the dimension, i.e., the number of spectral channels. A single training data point in the training dataset is represented by x1. Therefore, the training dataset can be represented as The representations of the validation and test sets are consistent with those of the training set.

[0058] In one implementation, the annotations of the training dataset (i.e., the classification and recognition results corresponding to the training dataset) can be represented by Y1. Here, 4 represents four mineral labels after one-hot encoding: categories 1, 2, 3, and 4 represent iron / magnesium montmorillonite, magnesium carbonate, fir peridotite, and background pixels, respectively. In the training dataset annotation, a single data label is represented by y1, where y∈R. 1 X 4 Therefore, the annotations of the training dataset can be represented as:

[0059] It is understandable that the representation of the validation and test datasets can be the same as that of the training dataset, and will not be elaborated here.

[0060] S104. Construct a mineral identification Transformer model.

[0061] The Transformer model for mineral identification is used to classify and identify minerals on the Martian surface. In some embodiments, Figure 3 This is a schematic diagram of the structure of the mineral recognition Transformer model provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the mineral recognition Transformer model includes an embedding mapping function, an encoder, a fully connected layer, and an output layer. The embedding mapping function consists of a band embedding function and a position embedding function, and the encoder is composed of multiple encoder layers connected in series.

[0062] Specifically, Martian mineral hyperspectral data can be input into an embedding mapping function. The output of the embedding mapping function (i.e., the embedding representation) is input into an encoder. The encoder's output is input into a fully connected layer, and the fully connected layer's output is input into an output layer. Finally, the output layer outputs the predicted classification and recognition results.

[0063] In some embodiments, the expression for the above embedding mapping function is:

[0064] Em(x) = Eb(x) + Ep(x);

[0065] In the formula, Em(x) represents the embedding mapping function, Eb(x) represents the band embedding function, Ep(x) represents the position embedding function, and x represents the input hyperspectral data.

[0066] The expression for Eb(x) is:

[0067] Eb(x)=W(x;θ e );

[0068] In the formula, W represents a linear mapping, and θ e Parameters indicating band embedding;

[0069] The expression for Ep(x) is:

[0070]

[0071] In the formula, p represents the position of the band in the hyperspectral data x, i.e. the band index, p = 0, 1, ..., 232; 2i represents the even-numbered dimension of the embedding, 2i+1 represents the odd-numbered dimension of the embedding, and d represents the total dimension of the embedding, i.e. the model dimension.

[0072] In some embodiments, the function expression of the encoder described above is:

[0073] xn = Encoder(xe);

[0074] In the formula, xn represents the output of the encoder, and xe represents the output of the embedding mapping function;

[0075] The expansion of Encoder(xe) is:

[0076]

[0077] In the formula, t (i) ,i=1,2…N represents the output of the i-th encoder layer, θ n (i) For the parameters of the i-th encoder layer, en (i) This represents the i-th encoder layer.

[0078] In some embodiments, the expression for the fully connected layer is:

[0079] xl=D{LN[γ(L(xn,θ f ))]};

[0080] In the formula, xl represents the output of the fully connected layer, and θ f Here are the parameters for the fully connected layer, where γ represents the ReLU activation function, L represents the fully connected layer, LN represents the layer normalization, and D represents the Dropout function.

[0081] In some embodiments, the expression for the output layer is:

[0082]

[0083] In the formula, This represents the output of the output layer, and S represents the Softmax activation function.

[0084] S105. Train the mineral recognition Transformer model based on the training dataset and the validation dataset to obtain the optimized mineral recognition Transformer model.

[0085] In some embodiments, with Figure 3 Taking the mineral identification Transformer model shown as an example, S105 may specifically include the following steps S1051-S1056:

[0086] S1051. Input the training dataset and validation dataset into the embedding mapping function, and output the first embedding representation corresponding to the training dataset and the second embedding representation corresponding to the validation dataset.

[0087] The first embedding representation and the second embedding representation can be determined based on the expression of the above embedding mapping function.

[0088] S1052. Input the first embedded representation and the second embedded representation into the encoder, and output the first encoder output result corresponding to the first embedded representation and the second encoder output result corresponding to the second embedded representation.

[0089] The output results of the first encoder and the second encoder can be determined based on the function expression of the encoder described above.

[0090] S1053. Input the output results of the first encoder and the second encoder into the fully connected layer, and output the first fully connected layer output result corresponding to the first encoder output result and the second fully connected layer output result corresponding to the second encoder output result.

[0091] The output results of the first fully connected layer and the second fully connected layer can be determined based on the expression of the fully connected layer mentioned above.

[0092] S1054. Input the output results of the first fully connected layer and the second fully connected layer into the output layer, and output the first predicted classification and recognition result corresponding to the output result of the first fully connected layer and the second predicted classification and recognition result corresponding to the output result of the second fully connected layer.

[0093] The first and second predicted classification results can be determined based on the expression of the output layer.

[0094] S1055. Construct a loss function based on the first predicted classification result, iteratively optimize the hyperparameters of the mineral recognition Transformer model to make the loss function converge, and obtain the hyperparameter training results.

[0095] In some embodiments, the loss function can be the cross-entropy loss function. For example, the cross-entropy loss function can be expressed as:

[0096]

[0097] Where d represents the model dimension. i This represents the category label in the predicted classification and recognition results of the training data.

[0098] S1056. Based on the second predicted classification and recognition results, the hyperparameter training results are optimized and adjusted to obtain an optimized mineral recognition Transformer model.

[0099] Specifically, the hyperparameters of the mineral recognition Transformer model include: the number of encoder layers, the model dimension, and the number of multi-head attention branches.

[0100] The hyperparameter training results can be further optimized and adjusted based on the second prediction classification results to determine the optimized hyperparameters, thereby obtaining an optimized mineral recognition Transformer model.

[0101] S106. Test the optimized mineral recognition Transformer model based on the test dataset and determine the test classification accuracy.

[0102] Furthermore, the test dataset can be input into the optimized mineral identification Transformer model to obtain the third predicted classification result, and then compared with the standard classification result in the test dataset to determine the test classification accuracy.

[0103] In some embodiments, the expression for the above-mentioned test classification accuracy is:

[0104]

[0105] in,

[0106]

[0107] In the formula, ACC represents the test classification accuracy, and n ′ Indicates the size of the test dataset. This represents the mineral classification test results (e.g., the third predicted classification identification results mentioned above), y i ′ This indicates the results of mineral classification standards (such as the classification and identification results mentioned above).

[0108] S107. If the test accuracy meets the preset accuracy conditions, input the hyperspectral data of the Martian minerals to be identified into the optimized mineral identification Transformer model, and output the Martian mineral classification and identification results.

[0109] Specifically, taking the classification accuracy ACC of the above test as an example, the higher the ACC, the more accurate the classification result of the optimized mineral recognition Transformer model.

[0110] In some embodiments, corresponding preset accuracy conditions can be set according to different application scenarios and application requirements. For example, corresponding preset accuracy conditions can be set according to different Martian regions. The preset accuracy conditions can be, for example, greater than 85%, or greater than 90%, etc. If the test accuracy meets the preset accuracy conditions, it can be determined that the optimized mineral identification Transformer model can meet the application requirements.

[0111] Furthermore, if the test accuracy meets the preset accuracy conditions, the hyperspectral data of the Martian minerals to be identified can be input into the optimized mineral identification Transformer model to output the Martian mineral classification and identification results.

[0112] In some embodiments, the Martian mineral classification and identification results can be mineral mapping results marked with different colors, wherein each color corresponds to a Martian mineral type.

[0113] For example, consider an MTRDR image from CR I SM as the source of the hyperspectral data for the Martian mineral to be identified. Here, the MTRDR image data is represented by X_im, and the mineral classification and identification result is represented by Y_im. The spectral data corresponding to a single pixel in the MTRDR image is represented by x_im, i.e. The classification result corresponding to a single pixel is represented by y_im, i.e. Where m is the total number of pixels in the MTRDR image. Inputting X_im into the optimized mineral recognition Transformer model yields Y_im. The mineral category recognition result for each pixel is then annotated with different colors in the MTRDR image, resulting in the mineral mapping result as shown below. Figure 4 As shown in the mineral mapping results, the blue area 401 represents the mineral category of iron / magnesium montmorillonite, the green area 402 represents the mineral category of magnesium carbonate, and the red area 403 represents the mineral category of fir peridotite.

[0114] It should be noted that the embodiments provided in this application... Figure 4The colors shown are illustrative and intended only to distinguish different Martian mineral types; they are not intended to be limiting. The colors corresponding to different Martian mineral types can be set according to actual application needs, and this application does not impose specific limitations on this.

[0115] The Martian surface mineral identification method based on Transformer provided in this invention captures the internal correlations of Martian mineral standard hyperspectral data through deep learning using the Transformer model. It focuses on valuable information within the Martian mineral standard hyperspectral data while ignoring unimportant information. In this way, the mineral identification Transformer model can learn richer feature representations from the Martian mineral standard hyperspectral data, thereby improving the accuracy of Martian mineral classification and identification. This method also has significant guiding significance for the development of Martian resource exploration projects and the exploitation and utilization of deep space resources.

[0116] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0117] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0118] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0119] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0120] Similar parts between the embodiments provided in this application can be referred to mutually. The specific implementation methods provided above are only a few examples under the overall concept of this application and do not constitute a limitation on the scope of protection of this application. For those skilled in the art, any other implementation methods extended from the solution of this application without creative effort shall fall within the scope of protection of this application.

Claims

1. A method for identifying minerals on the surface of Mars based on Transformer, characterized in that, include: Obtain a standard hyperspectral dataset of Martian minerals; The Martian mineral standard hyperspectral dataset is preprocessed to obtain the preprocessed Martian mineral standard hyperspectral dataset. The preprocessed Mars mineral standard hyperspectral dataset is divided into three datasets according to a preset division ratio: training dataset, validation dataset, and test dataset. A mineral recognition Transformer model is constructed, which is used to classify and identify minerals on the surface of Mars. The mineral recognition Transformer model includes: an embedding mapping function, an encoder, a fully connected layer, and an output layer; The embedding mapping function consists of a band embedding function and a position embedding function, and the encoder consists of multiple encoder layers connected in series. The expression for the embedding mapping function is: ； In the formula, This indicates an embedded mapping function. Represents the band embedding function. This indicates a positional embedding function. This represents the input hyperspectral data; Among them, the The expression is: ； In the formula, Represents a linear mapping. Parameters indicating band embedding; The The expression is: ； In the formula, Representing hyperspectral data The location of the medium band, ; Indicates the even-numbered dimension of the embedding. Indicates the odd-numbered dimensions of the embedding. This represents the total dimension of the embedding; The mineral recognition Transformer model is trained using the training dataset and the validation dataset to obtain an optimized mineral recognition Transformer model. The optimized mineral identification Transformer model was tested based on the test dataset, and the test classification accuracy was determined. If the test accuracy meets the preset accuracy conditions, the hyperspectral data of the Martian minerals to be identified is input into the optimized mineral identification Transformer model, and the Martian mineral classification and identification results are output.

2. The method according to claim 1, characterized in that, The step of training the mineral recognition Transformer model based on the training dataset and the validation dataset to obtain an optimized mineral recognition Transformer model includes: The training dataset and the validation dataset are input into the embedding mapping function, and the function outputs a first embedding representation corresponding to the training dataset and a second embedding representation corresponding to the validation dataset. The first embedding representation and the second embedding representation are input into the encoder, and the first encoder output result corresponding to the first embedding representation and the second encoder output result corresponding to the second embedding representation are output. The first encoder output and the second encoder output are input into the fully connected layer, and the first fully connected layer output corresponding to the first encoder output and the second fully connected layer output corresponding to the second encoder output are output. The output results of the first fully connected layer and the second fully connected layer are input into the output layer to obtain a first predicted classification and recognition result corresponding to the first fully connected layer output result and a second predicted classification and recognition result corresponding to the second fully connected layer output result. Based on the first predicted classification and recognition result, a loss function is constructed, and the hyperparameters of the mineral recognition Transformer model are iteratively optimized to make the loss function converge, thereby obtaining the hyperparameter training result; The hyperparameter training results are optimized and adjusted based on the second predicted classification and recognition results to obtain the optimized mineral recognition Transformer model.

3. The method according to claim 1 or 2, characterized in that, The function expression of the encoder is: ； In the formula, This represents the output of the encoder. This represents the output of the embedding mapping function; in, The expansion is: ； In the formula, This represents the output of the i-th encoder layer. For the parameters of the i-th encoder layer, This represents the i-th encoder layer.

4. The method according to claim 1 or 2, characterized in that, The expression for the fully connected layer is: ； In the formula, This represents the output of the fully connected layer. These are the parameters of the fully connected layer. This represents the ReLU activation function. Indicates a fully connected layer. Representation layer normalization, This represents the Dropout function.

5. The method according to claim 1 or 2, characterized in that, The expression for the output layer is: ； In the formula, This indicates the output result of the output layer. This represents the Softmax activation function.

6. The method according to claim 2, characterized in that, The loss function is the cross-entropy loss function, and the hyperparameters of the mineral recognition Transformer model include: the number of encoder layers, the model dimension, and the number of multi-head attention branches.

7. The method according to claim 1, characterized in that, The expression for the test classification accuracy is: ； in, ； in, This indicates the test classification accuracy. Indicates the size of the test dataset. This indicates the results of the mineral classification test. This indicates the results of the mineral classification standards.

8. The method according to claim 1, characterized in that, The Mars mineral standard hyperspectral dataset includes: iron / magnesium montmorillonite spectral data, magnesium carbonate spectral data, and fir peridot spectral data; the preset division ratio is: training dataset: validation dataset: test dataset = 8:1:1.

Images