Martian mineral spectral identification method and device based on deep learning

Abstract

The present invention relates to a method and device for identifying Martian mineral spectra based on deep learning. The Martian mineral spectra identification method based on deep learning uses four-dimensional constraints, namely, standard deviation, spectrum length, absorption depth, and overall reflectivity, to adaptively identify background spectra in images, and calculate ratio spectra based on dynamic background spectra, effectively eliminating interference from non-target factors such as atmospheric scattering and instrument noise. By directly extracting endmember spectra from hyperspectral images and combining the second ratio spectrum with a preset standard spectrum library, a mixed spectrum library containing Martian-specific minerals is constructed, thereby making up for the lack of data in the preset standard spectrum library and avoiding the problem of poor regional adaptability of directly using the preset standard spectrum library, that is, directly using the Earth spectrum library may not accurately reflect the surface characteristics of Mars, thereby improving the adaptability of the model to the Martian surface and ensuring that the model has higher recognition accuracy in the specific environment of Mars.

Description

Technical Field

[0001] The present invention relates to the technical field of Martian mineral identification, and in particular to a method and device for Martian mineral spectral identification based on deep learning. Background Art

[0002] The precise identification of Martian surface mineral composition holds important scientific value for revealing the planet's geological evolution, searching for traces of extraterrestrial life, and reconstructing paleoclimate environments. Orbital hyperspectral remote sensing, with its nanometer-scale spectral resolution, has become the most effective means of obtaining information on the spatial distribution of Martian minerals. This technology acquires continuous reflectance spectral signatures of Martian surface materials in the visible, near-infrared, and shortwave infrared bands, mapping spectral response mechanisms such as mineral crystal field effects and charge transfer with mineral chemical composition, enabling precise identification of mineral species.

[0003] Currently, the identification method based on spectral library matching is a classic paradigm in this field. Its technical path mainly includes two key links: first, constructing a standard spectral database covering typical Martian minerals, and then using similarity measurement algorithms such as SAM (SpectralAngleMapper) or SID (Spectral Information Divergence) to achieve mineral classification through pixel-by-pixel comparison. Although this method has shown feasibility in the identification of typical minerals such as hydrous sulfates and layered silicates, its engineering application faces significant bottlenecks: First, the nanophase iron oxide covering layer that is ubiquitous on the surface of Mars will cause spectral blue shifts, which, combined with atmospheric radiation transmission effects (including CO2 absorption, aerosol scattering, etc.), will cause characteristic distortion of the orbital spectrum; second, traditional spectral matching requires traversing the entire library to perform complex calculations, which is difficult to meet the needs of mineral mapping in large areas.

[0004] In recent years, deep learning architectures, such as convolutional neural networks and variational autoencoders, have demonstrated superior performance over traditional methods in complex tasks such as sulfate mineral subclassification by automatically extracting spectral features through multi-level nonlinear transformations. However, due to various noises present in spectral data (such as instrumental noise, random noise, and residual errors after atmospheric correction), deep learning techniques still suffer from low recognition accuracy. Summary of the Invention

[0005] Based on this, it is necessary to provide a Martian mineral spectral recognition method and device based on deep learning that can improve the recognition accuracy.

[0006] To achieve the above-mentioned objectives, on the one hand, an embodiment of the present application provides a method for spectral identification of Martian minerals based on deep learning, comprising: acquiring orbital hyperspectral image data;

[0007] Extract the spectral curve of any pixel in orbital hyperspectral image data;

[0008] Obtaining the spectral characteristics of each spectral curve, and calculating the standard deviation of each spectral curve based on the spectral characteristics;

[0009] Obtaining the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve;

[0010] Identify the spectral curve that meets the following conditions as the background spectrum in any image: the standard deviation is less than a first preset value, the length is less than a second preset value, the absorption depth is less than a third preset value, and the overall average reflectivity is less than a fourth preset value;

[0011] According to the background spectrum and the track hyperspectral image data, the first ratio spectrum of any pixel is obtained;

[0012] Using a pre-set standard spectral library, pure endmember spectral patches are matched in orbital hyperspectral imagery data;

[0013] Calculating the second ratio spectrum of the pure end-member spectrum block, and constructing a mineral end-member spectrum library according to the ratio spectrum and the second ratio spectrum of the preset standard spectrum library;

[0014] Construct a network model and use the mineral end-member spectrum library to train the network model to obtain a mineral identification model;

[0015] The first ratio spectrum is input into the mineral identification model to obtain a mineral spatial distribution map.

[0016] In one embodiment, it further includes:

[0017] Preprocessing the orbital hyperspectral image data to obtain a preprocessed spectral image;

[0018] The steps of extracting the spectral curve in any pixel of the orbital hyperspectral image data include:

[0019] Extract the spectral curve in any pixel of the preprocessed spectral image;

[0020] In one embodiment, the step of preprocessing the track hyperspectral image data to obtain a preprocessed spectral image includes:

[0021] The orbital hyperspectral image data is subjected to radiometric calibration and atmospheric correction to obtain a preprocessed spectral image.

[0022] In one embodiment, it further includes:

[0023] Perform data enhancement processing on the mineral end-member spectral library to obtain a multi-condition training data set;

[0024] The steps for training the network model using the mineral endmember spectral library include:

[0025] The network model is trained using a multi-condition training dataset.

[0026] In one embodiment, the step of performing data enhancement processing on the mineral end-member spectral library includes:

[0027] Generate a random number within a preset range;

[0028] Superimposing random numbers on the reflectance of the spectrum of each band in the mineral end-member spectrum library; and / or,

[0029] The reflectance of the spectrum of each band in the mineral end-member spectral library is superimposed with linear trend noise; and / or,

[0030] is the reflectance of the spectrum of each band in the mineral end-member spectral library, superimposed with multiplicative noise.

[0031] In one embodiment, the step of calculating the second ratio spectrum of the clean endmember spectrum block includes:

[0032] Confirm the background spectrum of the pure endmember spectrum tile;

[0033] A second ratio spectrum is obtained according to the background spectrum of the pure endmember spectrum block and the pure endmember spectrum block.

[0034] In one embodiment, the preset wavelength bands are 4 μm, 1.9 μm, 2.2 μm, and 2.3 μm.

[0035] On the one hand, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program and the processor executes the steps of the above method.

[0036] On the one hand, an embodiment of the present invention provides a Martian mineral spectrum identification device based on deep learning, comprising:

[0037] A first acquisition module is used to acquire orbital hyperspectral image data;

[0038] An extraction module, used to extract the spectral curve in any pixel of the orbital hyperspectral image data;

[0039] The second acquisition module is used to obtain the spectral characteristics of each spectral curve and calculate the standard deviation of each spectral curve based on the spectral characteristics;

[0040] A third acquisition module is used to obtain the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve;

[0041] a confirmation module, configured to confirm a spectral curve that meets the following conditions as a background spectrum in any image: a standard deviation less than a first preset value, a length less than a second preset value, an absorption depth less than a third preset value, and an overall average reflectivity less than a fourth preset value;

[0042] A ratio spectrum acquisition module is used to obtain a first ratio spectrum of any pixel based on the background spectrum and track hyperspectral image data;

[0043] A screening module is used to match pure endmember spectral blocks in orbital hyperspectral image data using a preset standard spectral library;

[0044] A mineral end-member spectrum library construction module is used to calculate the second ratio spectrum of the pure end-member spectrum block and construct the mineral end-member spectrum library according to the ratio spectrum and the second ratio spectrum of the preset standard spectrum library;

[0045] The network model building module is used to build a network model and train the network model using the mineral end-member spectrum library to obtain a mineral identification model;

[0046] The identification module is used to input the first ratio spectrum into the mineral identification model to obtain a mineral spatial distribution map.

[0047] On the other hand, the present application provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of the above method when run.

[0048] One of the above technical solutions has the following advantages and beneficial effects:

[0049] The above-mentioned deep learning-based Martian mineral identification method adaptively identifies the background spectrum in the image through four-dimensional constraints of standard deviation, spectral length, absorption depth, and overall reflectivity. It calculates the ratio spectrum based on the dynamic background spectrum, which can effectively eliminate interference from non-target factors such as atmospheric scattering and instrument noise. By directly extracting endmember spectra from hyperspectral images and combining the second ratio spectrum with a preset standard spectral library, a hybrid spectral library containing Martian-specific minerals is constructed. This makes up for the lack of data in the preset standard spectral library and avoids the problem of poor regional adaptability of directly using the preset standard spectral library. In other words, directly using the Earth spectral library may not accurately reflect the surface characteristics of Mars. This improves the model's adaptability to the Martian surface, ensures that the model has higher recognition accuracy in the specific environment of Mars, and solves the problems of spectral noise and insufficient training samples that hinder the spectral identification of Martian minerals, which can improve the accuracy of mineral identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0052] Figure 1 1 is a schematic flow chart of a method for identifying Martian mineral spectra based on deep learning in one embodiment;

[0053] Figure 2 is a schematic flow chart of the steps of performing data enhancement processing on a mineral end-member spectral library in one embodiment;

[0054] Figure 3 FIG. 1 is a schematic flow chart of the steps of calculating a second ratio spectrum of a clean endmember spectrum block in one embodiment. DETAILED DESCRIPTION

[0055] To facilitate understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings. The accompanying drawings provide embodiments of the present application. However, the present application may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the present application more thorough and comprehensive.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are for the purpose of describing specific embodiments only and are not intended to limit this application.

[0057] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of this application and have no specific meaning. Therefore, "module" and "component" can be used interchangeably.

[0058] It can be understood that the “connection” in the following embodiments should be understood as “electrical connection”, “communication connection”, etc. if there is transmission of electrical signals or data between the connected circuits, modules, units, etc.

[0059] When used herein, the singular forms "a", "an", and "the" may also include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, integers, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof.

[0060] In one embodiment, Figure 1 As shown, a method for identifying Martian mineral spectra based on deep learning is provided, comprising: S101, acquiring orbital hyperspectral image data;

[0061] Orbital hyperspectral imagery refers to spectral data covering a large area acquired by orbital remote sensing instruments. This data contains reflectance spectral information from visible light to shortwave infrared wavelengths over the target area and can be used to analyze the composition and distribution of surface materials.

[0062] Hyperspectral imaging data is typically acquired using hyperspectral imagers on satellites or probes, such as CRISM (Compact Reconnaissance Imaging Spectrometer for Mars). These instruments are capable of high-resolution imaging across multiple wavelengths. Through precise spectral acquisition, the reflectance of each pixel in a specific wavelength band can be determined, providing a data foundation for subsequent mineral identification.

[0063] S103, extracting a spectral curve in any pixel of the track hyperspectral image data;

[0064] Among them, the spectral curve refers to the reflectance change diagram of a specific pixel in multiple bands, which reflects the spectral characteristics of the material in the pixel.

[0065] Specifically, extracting the spectral curve of a single pixel from hyperspectral imagery is a crucial step in analyzing specific surface materials. The information contained in each pixel's spectral curve can reveal the region's mineral composition, chemical properties, and physical state. By analyzing the spectral curve, it's possible to identify the presence of a specific mineral or minerals in that pixel.

[0066] S105, obtaining spectral characteristics of each spectral curve, and calculating the standard deviation of each spectral curve based on the spectral characteristics;

[0067] Spectral features refer to important parameters in a spectral curve, such as absorption peaks and reflectance values. These features can help distinguish different mineral compositions. Standard deviation is a statistic that describes the degree of variation in spectral curve data.

[0068] Specifically, key features can be extracted from spectral curves, such as reflectance peaks and absorption characteristics in specific wavelength bands. These features not only aid in mineral identification but also reflect the impact of environmental changes on mineral spectra. The standard deviation can be calculated by first taking the mean of the spectral curve at each wavelength, then calculating the square of the deviation at each wavelength, and finally summing and taking the square root of these values ​​to obtain the standard deviation.

[0069] S107, obtaining the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve;

[0070] Specifically, the above parameters can be obtained by any means known in the art, such as stretching the spectrum curve into a line segment and calculating the segment length. The absorption depth can be determined by selecting a target wavelength (e.g., 1.9 μm) and calculating the difference between the peak and the baseline. The overall average reflectance can be obtained by averaging the reflectance across all wavelengths.

[0071] S109, identifying a spectral curve that meets the following conditions as a background spectrum in any image: a standard deviation less than a first preset value, a length less than a second preset value, an absorption depth less than a third preset value, and an overall average reflectivity less than a fourth preset value;

[0072] The background spectrum refers to the spectrum where the mineral absorption characteristics are not obvious (such as flat sand or dust-covered areas without spectral characteristics).

[0073] Specifically, a series of conditions, including standard deviation, length, absorption depth, and overall average reflectance, were set to screen out background spectra. These conditions helped ensure that the selected spectral curves were statistically reliable, effectively eliminating inaccurate data caused by noise or environmental variations. The identification of background spectra provides a benchmark for subsequent precise mineral identification.

[0074] In one of the optional implementations, the first preset value, the second preset value, the third preset value and the fourth preset value can be further processed by the following means. The initial preset value can be obtained by a sample learning method. Taking into account that the temperature changes during the day and night will affect the temperature, pressure and chemical composition of the Martian atmosphere, thereby affecting the spectral characteristics. Therefore, a day and night on Mars is divided into four time periods, namely the first time period T1, the second time period T2, the third time period T3 and the fourth time period T4. According to the rotation period of Mars (24 hours and 37 minutes, approximately 24 hours) and the law of spectral changes, a day is divided into 4 time periods (each time is about 6 hours):

[0075] T1 period (00:00-06:00): pre-dawn low temperature and low pressure period, CO2 frost condensation is significant;

[0076] T2 period (06:00-12:00): Morning warming period, with enhanced atmospheric turbulence;

[0077] T3 period (12:00-18:00): afternoon high temperature period, peak water vapor / ozone content;

[0078] T4 period (18:00-24:00): Dusk cooling period, the probability of dust storm activity increases;

[0079] The first hour and second hour of each time period are assigned values ​​of 1 and 2, respectively, and so on. Obtain spectral data for different time periods and calculate the characteristics of the background spectrum in each scene (such as standard deviation, length, absorption depth, and overall average reflectance). Then calculate the correlation coefficient between each feature and each time period. The specific calculation formula can be referred to as the following formula:

[0080]

[0081] Where n is the sample size; x i is the i-th observation value of the feature; Y i is the i-th observation value of the scene (the first hour is 1; the second hour is 2; the third hour is 3; the fourth hour is 4, ...); is the average of all observations; is the mean of the observations for all scenes. First, a baseline coefficient W1 is determined based on the feature importance, and then the baseline coefficient is adjusted based on the correlation coefficient r between the scene and the feature, such as W2 = W1(1+r). This results in a final coefficient for each feature in each time period. The initial values ​​of the first, second, third, and fourth preset values ​​are multiplied by the corresponding final coefficients to obtain the final preset values ​​for each time period. The orbital hyperspectral image data also includes time points, and the spectral curve that meets the following conditions is identified as the background spectrum in any image: the standard deviation is less than the first preset value, the length is less than the second preset value, the absorption depth is less than the third preset value, and the overall average reflectance is less than the fourth preset value. This specifically includes matching the final first preset value, final second preset value, final third preset value, and final fourth preset value for the corresponding time period according to the time point; and identifying the spectral curve that meets the following conditions as the background spectrum in any image: the standard deviation is less than the final first preset value, the length is less than the final second preset value, the absorption depth is less than the final third preset value, and the overall average reflectance is less than the final fourth preset value.

[0082] S111, obtaining a first ratio spectrum of any pixel according to the background spectrum and the track hyperspectral image data;

[0083] Specifically, the first ratio spectrum, derived by comparing a specific spectral curve with a background spectrum, highlights the potential mineral signature within that pixel. By calculating this first ratio spectrum, the influence of background signatures can be eliminated, allowing for clearer identification of potential mineral signatures. Ratio spectra enhance the recognizability of mineral signature peaks, allowing mineral signatures to be highlighted despite complex background interference.

[0084] S113, using a preset standard spectral library, matching pure endmember spectral blocks in the orbital hyperspectral image data;

[0085] The pre-set standard spectral library is a database containing known mineral spectral signatures, such as the USGS (US Geological Survey) spectral library. This library is used to compare with actual data to identify specific mineral components. A pure endmember spectrum refers to the spectral signature of a specific mineral or feature in orbital hyperspectral imagery data, uninterrupted by other substances.

[0086] Using known mineral spectra from a standard spectral library, the first ratio spectrum can be compared to find matching pure endmember spectral tiles. Tiles can be pixels or regions. During this search, constraints may be applied to ensure the purity of the extracted spectrum, such as the absorption position and depth of the spectral curve. This process involves not only direct spectral comparison but also considers spectral similarity and matching to ensure accurate identification. This matching effectively selects spectra representing specific minerals.

[0087] S115, calculating a second ratio spectrum of the pure end-member spectrum block, and constructing a mineral end-member spectrum library according to the ratio spectrum of the preset standard spectrum library and the second ratio spectrum;

[0088] The second ratio spectrum is a spectrum obtained by performing a ratio operation on the pure end-member spectrum block and the background spectrum, and the mineral end-member spectrum library is a spectrum set containing different mineral characteristics.

[0089] Specifically, the second ratio spectrum of the pure endmember spectrum block can be calculated using the same method as the first ratio spectrum. The ratio spectra in the MICA (Mars-like Inventory of Chemical Alteration) spectral library are used as partial endmember spectra, and the second ratio spectrum is used to supplement the mineral endmember spectral library. In other words, the mineral endmember spectral library includes the ratio spectra of the preset standard spectral library and the second ratio spectrum calculated above.

[0090] S117, constructing a network model, and using a mineral end-member spectrum library to train the network model to obtain a mineral identification model;

[0091] Specifically, a network model containing long and short time series and attention mechanism can be constructed, and the model can be trained using the mineral end-member spectral library.

[0092] S119, inputting the first ratio spectrum into a mineral identification model to obtain a mineral spatial distribution map.

[0093] The mineral spatial distribution map is a visualization of the mineral distribution on the Martian surface, obtained through analysis by the mineral recognition model. It can also be a mineral spatial distribution probability map. The first ratio spectrum is input into the trained mineral recognition model for analysis. The model compares the spectral features with the patterns learned during training, ultimately generating an image showing the spatial distribution of different minerals on the Martian surface.

[0094] The above-mentioned deep learning-based Martian mineral identification method adaptively identifies the background spectrum in the image through four-dimensional constraints of standard deviation, spectral length, absorption depth, and overall reflectivity. It calculates the ratio spectrum based on the dynamic background spectrum, which can effectively eliminate interference from non-target factors such as atmospheric scattering and instrument noise. By directly extracting endmember spectra from hyperspectral images and combining the second ratio spectrum with a preset standard spectral library, a hybrid spectral library containing Martian-specific minerals is constructed. This makes up for the lack of data in the preset standard spectral library and avoids the problem of poor regional adaptability of directly using the preset standard spectral library. In other words, directly using the Earth spectral library may not accurately reflect the surface characteristics of Mars. This improves the model's adaptability to the Martian surface, ensures that the model has higher recognition accuracy in the specific environment of Mars, and solves the problems of spectral noise and insufficient training samples that hinder the spectral identification of Martian minerals, which can improve the accuracy of mineral identification.

[0095] In one embodiment, it further includes:

[0096] Preprocessing the orbital hyperspectral image data to obtain a preprocessed spectral image;

[0097] The steps of extracting the spectral curve in any pixel of the orbital hyperspectral image data include:

[0098] Extract the spectral curve in any pixel of the preprocessed spectral image;

[0099] Specifically, the orbital hyperspectral image data needs to be preprocessed before extracting the spectral curve in any pixel. The steps of preprocessing the orbital hyperspectral image data to obtain a preprocessed spectral image include: performing radiometric calibration and atmospheric correction on the orbital hyperspectral image data to obtain a preprocessed spectral image. Radiometric calibration is to convert the CRISM hyperspectral image data into reflectance values ​​to eliminate the influence of sensor response and solar radiation changes. Using a standardized radiometric calibration process usually includes absolute radiometric calibration of the image and correction through a known radiometric standard (such as the solar spectrum). Atmospheric correction is to use an atmospheric correction model to remove the influence of atmospheric scattering and absorption to obtain the true reflectance of the surface. Perform corresponding corrections based on the observation conditions of the image (such as observation angle, meteorological conditions, etc.).

[0100] In one embodiment, it further includes:

[0101] Perform data enhancement processing on the mineral end-member spectral library to obtain a multi-condition training data set;

[0102] The steps for training the network model using the mineral endmember spectral library include:

[0103] The network model is trained using a multi-condition training dataset.

[0104] Specifically, in order to further improve the robustness of the model, a multi-condition training dataset can be generated through data enhancement, and the network model can be trained using the multi-condition training dataset.

[0105] Further, such as Figure 2 As shown in FIG, the steps for data enhancement processing of the mineral end-member spectral library include:

[0106] S210, generating a random number within a preset interval;

[0107] S220, superimposing the random number onto the reflectance of the spectrum of each band in the mineral end-member spectrum library; and / or,

[0108] S230, superimposing linear trend noise on the reflectance of the spectrum of each band in the mineral end-member spectrum library; and / or,

[0109] S240 is the reflectance of the spectrum of each band in the mineral end-member spectrum library, superimposed with multiplicative noise.

[0110] Specifically, all endmember spectra were simulated with noise using the following three methods to form a training dataset containing 100,000 spectra:

[0111] Generate random numbers in the range [-0.05, 0.05] and add them to the reflectance values ​​of each band to simulate the jitter noise of the track spectrum data. This noise simulation can effectively simulate the spectral changes caused by device jitter.

[0112] A linear noise is superimposed on the reflectance values ​​of each band, in the form of y = kx, where x is in the range [0, 1], evenly divided by the number of bands, and k is a random number in the interval [-0.5, 0.5]. This noise simulation can be used to simulate rising or falling trend noise in orbital spectral data, enhancing the model's adaptability to trend changes.

[0113] A multiplicative noise (m) is added to the reflectance values ​​of all bands. m can be a random number in the interval [0.5, 1]. This noise simulation can simulate the tendency of the absorption characteristics of orbital spectral data to weaken, making the model more robust when processing real data.

[0114] In one embodiment, Figure 3 As shown in FIG, the steps of calculating the second ratio spectrum of the clean end member spectrum block include:

[0115] S310, confirming the background spectrum of the pure end member spectrum block;

[0116] S320 , obtaining a second ratio spectrum according to the background spectrum of the pure endmember spectrum block and the pure endmember spectrum block.

[0117] Specifically, the calculation of the second ratio spectrum may refer to the above calculation of the first ratio spectrum.

[0118] In one embodiment, a deep learning-based Martian mineral spectrum identification device is provided, comprising:

[0119] A first acquisition module is used to acquire orbital hyperspectral image data;

[0120] An extraction module, used to extract the spectral curve in any pixel of the orbital hyperspectral image data;

[0121] The second acquisition module is used to obtain the spectral characteristics of each spectral curve and calculate the standard deviation of each spectral curve based on the spectral characteristics;

[0122] A third acquisition module is used to obtain the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve;

[0123] a confirmation module, configured to confirm a spectral curve that meets the following conditions as a background spectrum in any image: a standard deviation less than a first preset value, a length less than a second preset value, an absorption depth less than a third preset value, and an overall average reflectivity less than a fourth preset value;

[0124] A ratio spectrum acquisition module is used to obtain a first ratio spectrum of any pixel based on the background spectrum and track hyperspectral image data;

[0125] A screening module is used to match pure endmember spectral blocks in orbital hyperspectral image data using a preset standard spectral library;

[0126] A mineral end-member spectrum library construction module is used to calculate the second ratio spectrum of the pure end-member spectrum block and construct the mineral end-member spectrum library according to the ratio spectrum and the second ratio spectrum of the preset standard spectrum library;

[0127] The network model building module is used to build a network model and train the network model using the mineral end-member spectrum library to obtain a mineral identification model;

[0128] The identification module is used to input the first ratio spectrum into the mineral identification model to obtain a mineral spatial distribution map.

[0129] For the specific limitations of the Mars mineral spectrum identification device based on deep learning, please refer to the limitations of the Mars mineral spectrum identification method based on deep learning above, which will not be repeated here. The various modules in the above-mentioned Mars mineral spectrum identification device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. It should be noted that the division of modules 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.

[0130] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0131] Acquire orbital hyperspectral image data;

[0132] Extract the spectral curve of any pixel in orbital hyperspectral image data;

[0133] Obtaining the spectral characteristics of each spectral curve, and calculating the standard deviation of each spectral curve based on the spectral characteristics;

[0134] Obtaining the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve;

[0135] Identify the spectral curve that meets the following conditions as the background spectrum in any image: the standard deviation is less than a first preset value, the length is less than a second preset value, the absorption depth is less than a third preset value, and the overall average reflectivity is less than a fourth preset value;

[0136] According to the background spectrum and the track hyperspectral image data, the first ratio spectrum of any pixel is obtained;

[0137] Using a pre-set standard spectral library, pure endmember spectral patches are matched in orbital hyperspectral imagery data;

[0138] Calculating the second ratio spectrum of the pure end-member spectrum block, and constructing a mineral end-member spectrum library according to the ratio spectrum and the second ratio spectrum of the preset standard spectrum library;

[0139] Construct a network model and use the mineral end-member spectrum library to train the network model to obtain a mineral identification model;

[0140] The first ratio spectrum is input into the mineral identification model to obtain a mineral spatial distribution map.

[0141] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0142] Acquire orbital hyperspectral image data;

[0143] Extract the spectral curve of any pixel in orbital hyperspectral image data;

[0144] Obtaining the spectral characteristics of each spectral curve, and calculating the standard deviation of each spectral curve based on the spectral characteristics;

[0145] Obtaining the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve;

[0146] Identify the spectral curve that meets the following conditions as the background spectrum in any image: the standard deviation is less than a first preset value, the length is less than a second preset value, the absorption depth is less than a third preset value, and the overall average reflectivity is less than a fourth preset value;

[0147] According to the background spectrum and the track hyperspectral image data, the first ratio spectrum of any pixel is obtained;

[0148] Using a pre-set standard spectral library, pure endmember spectral patches are matched in orbital hyperspectral imagery data;

[0149] Calculating the second ratio spectrum of the pure end-member spectrum block, and constructing a mineral end-member spectrum library according to the ratio spectrum and the second ratio spectrum of the preset standard spectrum library;

[0150] Construct a network model and use the mineral end-member spectrum library to train the network model to obtain a mineral identification model;

[0151] The first ratio spectrum is input into the mineral identification model to obtain a mineral spatial distribution map.

[0152] It is understood that the embodiments described herein may be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or a combination thereof.

[0153] For software implementation, the technology described herein can be implemented by a unit that performs the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0154] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software 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 beyond the scope of this application.

[0155] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0156] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple modules 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 an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0157] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0158] In addition, the functional units in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk. It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0159] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.

Claims

1. A method for identifying Martian mineral spectra based on deep learning, characterized in that: include: Acquire orbital hyperspectral image data; Extracting a spectral curve in any pixel of the track hyperspectral image data; Acquiring spectral characteristics of each of the spectral curves, and calculating a standard deviation of each of the spectral curves based on the spectral characteristics; Obtaining the length of each spectral curve, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve; Identify a spectral curve that meets the following conditions as a background spectrum in any image: the standard deviation is less than a first preset value, the length is less than a second preset value, the absorption depth is less than a third preset value, and the overall average reflectivity is less than a fourth preset value; Obtaining a first ratio spectrum of any pixel according to the background spectrum and the track hyperspectral image data; Using a preset standard spectral library, matching pure endmember spectral blocks in the orbital hyperspectral image data; Calculating a second ratio spectrum of the pure end-member spectrum block, and constructing a mineral end-member spectrum library according to the ratio spectrum of the preset standard spectrum library and the second ratio spectrum; Constructing a network model, and using the mineral end-member spectral library to train the network model to obtain a mineral identification model; The first ratio spectrum is input into the mineral identification model to obtain a mineral spatial distribution map.

2. The method for identifying Martian minerals based on deep learning according to claim 1, characterized in that: Also includes: Preprocessing the track hyperspectral image data to obtain a preprocessed spectral image; The step of extracting a spectral curve in any pixel of the track hyperspectral image data comprises: A spectral curve in any pixel of the preprocessed spectral image is extracted.

3. The method for identifying Martian minerals based on deep learning according to claim 2, characterized in that: The step of preprocessing the track hyperspectral image data to obtain a preprocessed spectral image comprises: The orbital hyperspectral image data is subjected to radiometric calibration and atmospheric correction processing to obtain the preprocessed spectral image.

4. The method for identifying Martian mineral spectra based on deep learning according to claim 1, characterized in that: Also includes: Performing data enhancement processing on the mineral end-member spectral library to obtain a multi-condition training data set; The step of training the network model using the mineral end-member spectral library comprises: The network model is trained using the multi-operating condition training data set.

5. The method for identifying Martian mineral spectra based on deep learning according to claim 4, characterized in that: The step of performing data enhancement processing on the mineral end-member spectral library comprises: Generate a random number within a preset range; superimposing the random number onto the reflectance of the spectrum of each band in the mineral end-member spectrum library; and / or, The reflectance of the spectrum of each band in the mineral end-member spectrum library is superimposed with linear trend noise; and / or, is the reflectance of the spectrum of each band in the mineral end-member spectral library, superimposed with multiplicative noise.

6. The method for identifying Martian mineral spectra based on deep learning according to claim 1, characterized in that: The step of calculating the second ratio spectrum of the pure end member spectrum block comprises: confirming the background spectrum of the pure end member spectrum block; The second ratio spectrum is obtained according to the background spectrum of the pure endmember spectrum block and the pure endmember spectrum block.

7. The method for identifying Martian minerals based on deep learning according to claim 1, characterized in that: The preset wavelength bands are 4 μm, 1.9 μm, 2.2 μm, and 2.3 μm.

8. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.

9. A Martian mineral spectrum identification device based on deep learning, characterized in that: include: A first acquisition module is used to acquire orbital hyperspectral image data; An extraction module, configured to extract a spectral curve in any pixel of the track hyperspectral image data; a second acquisition module, configured to acquire spectral features of each of the spectral curves, and calculate a standard deviation of each of the spectral curves based on the spectral features; a third acquisition module, configured to acquire the length of each of the spectral curves, the absorption depth of a preset wavelength band of the spectral curve, and the overall average reflectivity of the spectral curve; a confirmation module, configured to confirm a spectral curve that meets the following conditions as a background spectrum in any image: the standard deviation is less than a first preset value, the length is less than a second preset value, the absorption depth is less than a third preset value, and the overall average reflectivity is less than a fourth preset value; a ratio spectrum acquisition module, configured to obtain a first ratio spectrum of any pixel according to the background spectrum and the track hyperspectral image data; A screening module, configured to match pure end-member spectral blocks in the track hyperspectral image data using a preset standard spectral library; a mineral end-member spectrum library construction module, configured to calculate a second ratio spectrum of the pure end-member spectrum block, and construct a mineral end-member spectrum library based on the ratio spectrum of the preset standard spectrum library and the second ratio spectrum; A network model construction module is used to construct a network model and train the network model using the mineral end-member spectrum library to obtain a mineral identification model; The identification module is used to input the first ratio spectrum into the mineral identification model to obtain a mineral spatial distribution map.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the steps of the method according to any one of claims 1 to 7 when executed.

Images