Training method, detection method, device and electronic equipment for predicting model of trace element content

Abstract

The present invention provides a training method, a detection method, a device and an electronic device for a trace element content prediction model, and applies the fields of computer technology and artificial intelligence technology. The invention comprises: obtaining the sample trace element content and the sample laser induced breakdown spectrum; resampling the sample laser induced breakdown spectrum to obtain a normalized sample laser induced breakdown spectrum; determining the correlation coefficient according to the set of pixel point intensity values ​​in the sample trace element content and the normalized sample laser induced breakdown spectrum; determining the set of correlation coefficient pixel point intensity values ​​that meet different preset correlation coefficient thresholds as the target sample spectrum data; inputting the target sample spectrum data into K sub-network structures respectively to obtain the predicted sample trace element content; adjusting the parameters of the K sub-network structures respectively according to the predicted sample trace element content and the sample trace element content, and determining the target trace element content prediction model from the K sub-network structures.

Description

Technical Field

[0001] The present invention relates to the field of computer technology and artificial intelligence technology, and specifically to a training method, a detection method, a device and an electronic device for a trace element content prediction model. Background Art

[0002] Among the surface material composition analyzers carried by the Mars rover, the Laser-Induced Breakdown Spectroscopy (LIBS) instrument is used to obtain elemental composition information of Martian surface materials. It has technical advantages such as simultaneous detection of multiple elements, single-point high-precision analysis, and remote detection. LIBS spectroscopy is a type of atomic emission spectroscopy. It usually consists of a beam of high-power, short-pulse laser focused on the target surface, causing part of the surface material to vaporize into plasma, and then the plasma radiation is received by the spectrometer. LIBS spectra contain emission lines of atoms, ions, and a small number of molecules, which can reveal the elemental composition information of the target. However, due to the complexity of the composition of Martian materials, the Mars LIBS spectrum is usually affected by physical and chemical matrix effects, that is, the characteristic spectral lines in the spectrum may be related to multiple elements or affected by the physical properties of the material, which brings difficulties to the quantification of elements in the Mars LIBS spectrum.

[0003] In the Martian LIBS spectra, the characteristic spectral lines of the two trace elements Mn and Ba usually show multi-peak overlap and low intensity, which are difficult to accurately identify. At present, the signal extraction of Mn and Ba in the Martian LIBS instrument is based on manual visual inspection and comparison with the ground spectral library, which is subjective and can only extract strong features. In the quantitative inversion of Mn and Ba by the Martian LIBS instrument, either traditional single variable methods or some linear regression machine learning methods such as the Least Absolute Shrinkage and Selection Operator are used, and some more advanced integrated machine learning algorithms are not applied. Summary of the invention

[0004] In view of this, the present invention provides a training method, a detection method, a device and an electronic device for a trace element content prediction model.

[0005] One aspect of the present invention provides a method for training a trace element content prediction model, which is applicable to element content prediction in a Mars surface composition detector, comprising:

[0006] The sample trace element content contained in the sample and N sample laser induced breakdown spectra corresponding to the sample are obtained, wherein N is an integer ≥ 1; each sample laser induced breakdown spectrum in the N sample laser induced breakdown spectra is resampled to obtain normalized N sample laser induced breakdown spectra, wherein the wavelength values ​​of the normalized N sample laser induced breakdown spectra are consistent; according to the sample trace element content and the set of pixel point intensity values ​​in the normalized N sample laser induced breakdown spectra, the correlation coefficient between the sample trace element content and the set of pixel point intensity values ​​of the same wavelength is determined; based on different preset correlation coefficient thresholds, The set of pixel intensity values ​​of the same wavelength corresponding to correlation coefficients that meet different preset correlation coefficient thresholds is determined as the target sample spectral data; for the target sample spectral data corresponding to different preset correlation coefficient thresholds, the target sample spectral data are respectively input into the K sub-network structures of the machine learning model to obtain the predicted sample trace element contents corresponding to the K sub-network structures; according to the predicted sample trace element contents and the sample trace element contents corresponding to the K sub-network structures, the parameters of the K sub-network structures of the machine learning model are adjusted respectively, and the target trace element content prediction model is determined from the K sub-network structures.

[0007] Another aspect of the present invention provides a method for detecting the content of an element, which is applicable to the detection of the content of an element in a Mars surface composition detector, comprising:

[0008] The target spectral data is input into the target trace element content prediction model to obtain the target trace element content corresponding to the target spectral data; wherein the target trace element content prediction model is trained according to the above method.

[0009] Another aspect of the present invention provides a training device for a trace element content prediction model, which is suitable for predicting the element content in a Mars surface composition detector, including: an acquisition module, used to obtain the sample trace element content contained in the sample and N sample laser induced breakdown spectra corresponding to the sample, wherein N is an integer ≥ 1; a sampling module, used to resample each sample laser induced breakdown spectrum in the N sample laser induced breakdown spectra to obtain normalized N sample laser induced breakdown spectra, wherein the wavelength values ​​of the normalized N sample laser induced breakdown spectra are consistent; a first determination module, used to determine the correlation coefficient between the sample trace element content and a set of pixel point intensity values ​​at the same wavelength in the sample trace element content and the normalized N sample laser induced breakdown spectra; a second determination module, used to determine the correlation coefficient between the sample trace element content and a set of pixel point intensity values ​​at the same wavelength in the sample trace element content and the normalized N sample laser induced breakdown spectra; Different preset correlation coefficient thresholds are used to determine the set of pixel intensity values ​​of the same wavelength corresponding to the correlation coefficients that meet the different preset correlation coefficient thresholds as the target sample spectral data; an acquisition module is used to input the target sample spectral data corresponding to different preset correlation coefficient thresholds into the K sub-network structures of the machine learning model respectively, and obtain the predicted sample trace element contents corresponding to the K sub-network structures respectively; an adjustment module is used to adjust the parameters of the K sub-network structures of the machine learning model according to the predicted sample trace element contents and the sample trace element contents corresponding to the K sub-network structures respectively, and determine the target trace element content prediction model from the K sub-network structures.

[0010] Another aspect of the present invention provides a trace element content detection device, which is suitable for detecting element content in a Mars surface composition detector, comprising:

[0011] The input module is used to input the target spectrum data into the target trace element content prediction model to obtain the target trace element content corresponding to the target spectrum data; wherein the target trace element content prediction model is obtained by training according to the above-mentioned device.

[0012] Another aspect of the present invention provides an electronic device, comprising:

[0013] one or more processors;

[0014] a memory for storing one or more programs,

[0015] When one or more programs are executed by one or more processors, the one or more processors implement the above method.

[0016] Another aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the above method when executed.

[0017] Another aspect of the present invention provides a computer program product, the computer program product comprising computer executable instructions, and the instructions are used to implement the above method when being executed.

[0018] According to an embodiment of the present invention, since a correlation coefficient between the sample trace element content and a set of pixel intensity values ​​of the same wavelength in the normalized sample laser induced breakdown spectrum is adopted, the two can be determined, and then based on different preset correlation coefficient thresholds, a set of pixel intensity values ​​of the same wavelength corresponding to each preset correlation coefficient threshold is determined from the correlation coefficients as the target sample spectral data; each sub-network structure in the machine learning model is trained separately using the target sample spectrum corresponding to each preset correlation coefficient threshold, so that the sub-network structure with the smallest error is determined as the target trace element content prediction model. Therefore, the technical problem in the prior art that the extracted target sample spectral data is inaccurate due to comparison with the ground spectral library based on manual visual inspection, thereby affecting the accuracy of trace element detection, is at least partially overcome, and the technical effect of being able to effectively extract trace element signals from the Mars LIBS spectrum and accurately give the trace element content in the Martian surface material with a mass fraction accuracy of up to a few ten-thousandths is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0020] Figure 1 An exemplary system architecture of a training method, a detection method and a device that can apply a trace element content prediction model according to an embodiment of the present invention is shown;

[0021] Figure 2 A flow chart of a method for training a trace element content prediction model according to an embodiment of the present invention is shown;

[0022] Figure 3 A flow chart of a method for determining a correlation coefficient according to an embodiment of the present invention is shown;

[0023] Figure 4 A schematic diagram showing a method for training a trace element content prediction model according to an embodiment of the present invention is shown;

[0024] Figure 5 A block diagram of a training device for a trace element content prediction model according to an embodiment of the present invention is shown;

[0025] Figure 6 A block diagram of an electronic device suitable for implementing a training method and a detection method for a trace element content prediction model according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0026] Below, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of concepts of the present invention.

[0027] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0028] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0029] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0030] In the embodiments of the present invention, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage and other aspects of the data involved (for example, including but not limited to user personal information) are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information and network security.

[0031] An embodiment of the present invention provides a method for training a trace element content prediction model, which is applicable to element content prediction in a Mars surface composition detector, and includes:

[0032] The sample trace element content contained in the sample and N sample laser induced breakdown spectra corresponding to the sample are obtained, wherein N is an integer ≥ 1; each sample laser induced breakdown spectrum in the N sample laser induced breakdown spectra is resampled to obtain normalized N sample laser induced breakdown spectra, wherein each sample laser induced breakdown spectrum contains M pixel points; according to the sample trace element content and the intensity value of the M pixel points of the normalized laser induced breakdown spectrum of each sample, the correlation coefficient between the sample trace element content and the intensity value of each pixel point is determined; based on different preset correlation coefficient thresholds, The intensity value of the pixel point corresponding to the correlation coefficient that meets different preset correlation coefficient thresholds is determined as the target sample spectral data; for the target sample spectral data corresponding to different preset correlation coefficient thresholds, the target sample spectral data are respectively input into the K sub-network structures of the machine learning model to obtain the predicted sample trace element contents corresponding to the K sub-network structures; according to the predicted sample trace element contents and the sample trace element contents corresponding to the K sub-network structures, the parameters of the K sub-network structures of the machine learning model are adjusted respectively, and the target trace element content prediction model is determined from the K sub-network structures.

[0033] Figure 1 The exemplary system architecture of the training method, detection method and device that can apply the trace element content prediction model according to the embodiment of the present invention is shown. It should be noted that Figure 1 The examples shown are merely examples of system architectures to which the embodiments of the present invention may be applied, to help those skilled in the art understand the technical content of the present invention, but do not mean that the embodiments of the present invention may not be used in other devices, systems, environments or scenarios.

[0034] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0035] The user may use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only for example).

[0036] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0037] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0038] It should be noted that the training method and detection method of the trace element content prediction model provided in the embodiment of the present invention can generally be executed by the server 105. Accordingly, the training device and detection device of the trace element content prediction model provided in the embodiment of the present invention can generally be set in the server 105. The training method and detection method of the trace element content prediction model provided in the embodiment of the present invention can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Correspondingly, the training device and detection device of the trace element content prediction model provided in the embodiment of the present invention can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Alternatively, the training method and detection method of the trace element content prediction model provided in the embodiment of the present invention may also be performed by the first terminal device 101, the second terminal device 102 or the third terminal device 103, or may also be performed by other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103. Accordingly, the training device and detection device of the trace element content prediction model provided in the embodiment of the present invention may also be set in the first terminal device 101, the second terminal device 102 or the third terminal device 103, or may be set in other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0039] For example, the sample laser induced breakdown spectrum and the sample trace element content may be originally stored in any one of the first terminal device 101, the second terminal device 102 or the third terminal device 103 (for example, the first terminal device 101, but not limited thereto), or stored in an external storage device and may be imported into the first terminal device 101. Then, the first terminal device 101 may locally execute the training method and detection method of the trace element content prediction model provided in the embodiment of the present invention, or the sample laser induced breakdown spectrum and the sample trace element content may be sent to other terminal devices, servers, or server clusters, and the training method and detection method of the trace element content prediction model provided in the embodiment of the present invention may be executed by other terminal devices, servers, or server clusters that receive the sample laser induced breakdown spectrum and the sample trace element content.

[0040] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0041] Figure 2 A flow chart of a method for training a trace element content prediction model according to an embodiment of the present invention is shown.

[0042] like Figure 2 As shown, the method 200 includes operations S210 to S260.

[0043] In operation S210, the content of trace elements in the sample and N laser induced breakdown spectra of the samples corresponding to the sample are obtained, where N is an integer ≥1.

[0044] According to an embodiment of the present invention, the method can be applied to the prediction of element content in a Mars surface composition detector. In an existing ground sample library, the trace element content of samples in the ground sample library can be measured in a laboratory using an inductively coupled plasma mass spectrometer. For example, the content of Mn and Ba in the sample can be measured by taking the target sample trace elements manganese (Mn) and barium (Ba) in the sample as an example.

[0045] According to an embodiment of the present invention, since the content of trace elements in Martian materials is usually low, it is necessary to eliminate samples with relatively high trace element contents in the ground sample library and retain samples with relatively low trace element contents.

[0046] According to an embodiment of the present invention, a laser induced breakdown spectroscopy (LIBS) instrument may be used to collect laser induced breakdown spectra of samples containing low sample trace element contents in a Mars simulation environment.

[0047] According to an embodiment of the present invention, a laser induced breakdown spectrometer can be used to collect N sample laser induced breakdown spectra of each sample, from which the composition information of the sample trace elements can be obtained.

[0048] In operation S220, each of the N sample laser induced breakdown spectra is resampled to obtain normalized N sample laser induced breakdown spectra, wherein the wavelength values ​​of the normalized N sample laser induced breakdown spectra are consistent.

[0049] According to an embodiment of the present invention, after resampling N sample laser induced breakdown spectra, N normalized sample laser induced breakdown spectra can be obtained. Resampling can make the sampling wavelength values ​​of the N sample laser induced breakdown spectra consistent, and the spectral resolution after resampling can be consistent with the highest resolution of the spectrometer.

[0050] In operation S230, a correlation coefficient between the sample trace element content and the set of pixel intensity values ​​at the same wavelength in the normalized N sample laser induced breakdown spectra is determined according to the sample trace element content and the set of pixel intensity values ​​at the same wavelength.

[0051] According to an embodiment of the present invention, each sample laser induced breakdown spectrum includes M pixels, and each pixel includes a wavelength and an intensity value at the wavelength.

[0052] According to an embodiment of the present invention, taking the set of pixel intensity values ​​of the same wavelength in N sample laser induced breakdown spectra as a dimension, the N sample laser induced breakdown spectra may contain M sets of pixel intensity values ​​of the same wavelength.

[0053] According to an embodiment of the present invention, for each sample trace element in the sample, a correlation coefficient between the sample trace element and a set of pixel intensity values ​​at each of the M same wavelengths may be calculated.

[0054] According to an embodiment of the present invention, for example, for the trace element Mn in the sample, the content of the trace element Mn in the sample and the set of intensity values ​​of the pixel points at the same wavelength in the laser induced breakdown spectra of N samples can be used to calculate the correlation coefficient between the set of intensity values ​​of the pixel points at the wavelength in the laser induced breakdown spectra of N samples and the content of the trace element Mn in the sample, so that the correlation coefficients between the sets of intensity values ​​of M pixel points at the same wavelength in the laser induced breakdown spectra of N samples and the content of the trace element Mn in the sample can be calculated, so that M correlation coefficients can be obtained for the Mn element in the sample.

[0055] According to an embodiment of the present invention, the above method can be used to calculate the correlation coefficient between the content of other trace elements in the sample (for example, Ba) and the set of M pixel intensity values ​​of the same wavelength in the laser induced breakdown spectra of N samples, so that M correlation coefficients can be obtained for the Ba element in the sample.

[0056] In operation S240, based on different preset correlation coefficient thresholds, a set of pixel intensity values ​​at the same wavelength corresponding to correlation coefficients that meet different preset correlation coefficient thresholds is determined as target sample spectral data.

[0057] According to an embodiment of the present invention, for each trace element content in the sample, for example, the Mn element, different preset correlation coefficient thresholds corresponding to the Mn element can be set. For each preset correlation coefficient threshold, a correlation coefficient higher than the preset correlation threshold can be selected from the M correlation coefficients obtained for the Mn element in the sample, and the intensity value in the set of pixel point intensity values ​​corresponding to the correlation coefficient is used as the target sample spectral data corresponding to the Mn element.

[0058] According to an embodiment of the present invention, the above method can be utilized. For example, for the Ba element, different preset correlation coefficient thresholds corresponding to the Ba element can be set. For each preset correlation coefficient threshold, a correlation coefficient higher than the preset correlation threshold can be selected from the M correlation coefficients obtained for the Ba element in the sample, and the intensity value in the set of pixel point intensity values ​​corresponding to the correlation coefficient is used as the target sample spectral data corresponding to the Ba element.

[0059] In operation S250, for target sample spectral data corresponding to different preset correlation coefficient thresholds, the target sample spectral data are respectively input into K sub-network structures of the machine learning model to obtain predicted sample trace element contents corresponding to the K sub-network structures.

[0060] According to an embodiment of the present invention, for each sample trace element, for example, for the Mn element, and for the target sample spectral data related to each preset correlation coefficient threshold corresponding to the Mn element, the target sample spectral data corresponding to each preset correlation coefficient threshold is respectively input into the K sub-network structures of the machine learning model, so that the predicted trace element content output by each sub-network structure corresponding to each preset correlation coefficient threshold can be obtained.

[0061] In operation S260, according to the predicted sample trace element contents and the sample trace element contents corresponding to the K sub-network structures, the parameters of the K sub-network structures of the machine learning model are adjusted respectively, and a target trace element content prediction model is determined from the K sub-network structures.

[0062] According to an embodiment of the present invention, for each preset correlation coefficient threshold and each sub-network structure, a prediction loss value can be calculated based on the output predicted sample trace element content and sample trace element content, and then the parameters of each sub-network structure are adjusted based on the predicted loss value. When the training preset conditions are met, the training is stopped to obtain the trained K sub-network structures. The trained K sub-network structures are then verified respectively, and the model with the smallest error can be determined as the intermediate trace element content prediction model corresponding to each preset correlation threshold.

[0063] According to an embodiment of the present invention, for each preset correlation coefficient threshold, the intermediate trace element content prediction model is tested and verified, and the model with the smallest prediction error of the intermediate trace element content prediction model is determined as the target trace element content prediction model, and its corresponding preset correlation coefficient threshold is determined as the target correlation coefficient threshold corresponding to the target sample trace element that needs to be tested.

[0064] According to an embodiment of the present invention, by adopting a set of pixel intensity values ​​of the same wavelength based on the sample trace element content and the normalized sample laser induced breakdown spectrum, the correlation coefficient between the two can be determined, and then based on different preset correlation coefficient thresholds, a set of pixel intensity values ​​corresponding to each preset correlation coefficient threshold is determined from the correlation coefficients as the target sample spectral data; the target sample spectrum corresponding to each preset correlation coefficient threshold is used to train each sub-network structure in the machine learning model respectively, so as to determine the sub-network structure with the smallest error as the target trace element content prediction model. Therefore, the technical problem in the prior art that the extracted target sample spectral data is inaccurate due to comparison with the ground spectral library based on manual visual inspection, thereby affecting the accuracy of trace element detection, is at least partially overcome, and the technical effect of being able to effectively extract trace element signals from the Mars LIBS spectrum and accurately give the trace element content in the Martian surface material with a mass fraction accuracy of up to a few ten-thousandths is achieved.

[0065] According to an embodiment of the present invention, each sample laser induced breakdown spectrum of N sample laser induced breakdown spectra is resampled to obtain normalized N sample laser induced breakdown spectra, including:

[0066] For the m-th pixel point among the M pixels points in each sample laser induced breakdown spectrum, the average wavelength of the m-th pixel point is calculated to obtain the average wavelength of the M pixels points; the intensity value of the average wavelength of each sample laser induced breakdown spectrum in the N sample laser induced breakdown spectra is calculated by linear interpolation to obtain the intensity value of the average wavelength of the M pixels points; based on the intensity value of the average wavelength of the M pixels points and the average wavelength, the normalized N sample laser induced breakdown spectra are obtained, wherein M is a positive integer ≥1, and 1≤m≤M.

[0067] According to an embodiment of the present invention, before resampling the sample laser induced breakdown spectrum, data preprocessing can also be performed on each sample laser induced breakdown spectrum of the N sample laser induced breakdown spectra. Specifically, it may include: deducting the dark background of the spectrometer, converting the response value of the spectrometer into a radiant brightness value, converting the number of pixel bits of the spectrometer into a wavelength value, removing Gaussian white noise in the data, removing the continuous spectrum background signal in the data and splicing each spectral channel, and dividing the radiant brightness value of each pixel in the spectrum by the total area of ​​the corresponding spectral interval to achieve normalization. One or more of the following.

[0068] According to an embodiment of the present invention, each sample laser induced breakdown spectrum includes M pixels. The wavelength corresponding to each pixel in the N sample laser induced breakdown spectra can be averaged to obtain the average wavelength of each pixel. For example, taking the m-th pixel in each sample laser induced breakdown spectrum as an example, the N sample laser induced breakdown spectra have N m-th pixels, and the wavelengths corresponding to the N m-th pixels are averaged to obtain the average wavelength for the m-th pixel. Based on the above method, the average wavelength of the M pixels can be obtained.

[0069] According to an embodiment of the present invention, the N sample laser induced breakdown spectra can be linearly interpolated to the average wavelength using a linear interpolation method, and the intensity value corresponding to the average wavelength is calculated using the same pixel point on adjacent laser induced breakdown spectra, thereby obtaining the intensity value corresponding to the average wavelength on each sample laser induced breakdown spectrum.

[0070] According to an embodiment of the present invention, based on the above resampling, the wavelengths of N sample laser induced breakdown spectra are made consistent, and the average wavelength and intensity value corresponding to the pixel point on each sample laser induced breakdown spectrum can be used to obtain normalized N sample laser induced breakdown spectra.

[0071] Figure 3 A flow chart of a method for determining a correlation coefficient according to an embodiment of the present invention is shown.

[0072] like Figure 3 As shown, the method 300 may include: operations S310 to S330.

[0073] In operation S310, a covariance operation is performed on the trace element content of the sample and a set of pixel intensity values ​​at the same wavelength to obtain a first operation result.

[0074] In operation S320, a product operation is performed on the trace element content of the sample and the variance value of the set of pixel intensity values ​​at the same wavelength to obtain a second operation result.

[0075] In operation S330, a correlation coefficient between the trace element content of the sample and a set of pixel intensity values ​​at the same wavelength is determined according to the first operation result and the second operation result.

[0076] According to an embodiment of the present invention, for example, the wavelength values ​​of N normalized sample laser induced breakdown spectra are unified, each normalized sample laser induced breakdown spectrum contains M pixel points, each pixel point contains a wavelength and an intensity value corresponding to the wavelength, that is, each normalized sample laser induced breakdown spectrum contains M wavelength values.

[0077] According to an embodiment of the present invention, for pixels of the same wavelength in N normalized sample laser induced breakdown spectra, for example, for pixels corresponding to the first wavelength value with a pixel number, that is, the first pixel in each normalized sample laser induced breakdown spectrum, that is, N first pixels, the set of intensity values ​​corresponding to the N first pixels (hereinafter referred to as the first pixel) is recorded as I(1); a covariance operation is performed on the sample trace element Mn content and the set of intensity values ​​of the first pixel to obtain a first operation result; then a first variance of the sample trace element Mn content and a second variance of the set of intensity values ​​of the first pixel are calculated, and the first variance and the first variance are added together. The two variance operations are performed to obtain the second operation result; the first operation result and the second operation result can be used to obtain the first correlation coefficient between the trace element Mn content of the sample and the intensity value set corresponding to the first pixel point; based on the above method, the second correlation coefficient between the trace element Mn content of the sample and the intensity value set corresponding to the N second pixel points is calculated; in this way, the Mth correlation coefficient between the trace element Mn content of the sample and the intensity value set corresponding to the N Mth pixel points can be obtained, thereby obtaining the correlation coefficients between the trace element Mn content of the sample and the intensity value sets corresponding to the M pixel points, that is, obtaining M correlation coefficients for the trace element Mn.

[0078] According to an embodiment of the present invention, for example, the above method can be referred to to calculate the correlation coefficient between the content of trace element Ba and the set of intensity values ​​corresponding to M pixel points, that is, to obtain M correlation coefficients for trace element Ba; and to obtain M correlation coefficients for other trace elements.

[0079] According to an embodiment of the present invention, the correlation coefficient Corr(·) between the sample trace element content and the set of pixel intensity values ​​can be obtained by the following formula (1), that is:

[0080] (1);

[0081] Wherein, Corr(x) is the xth correlation coefficient between the trace element content of the sample and the set of intensity values ​​corresponding to the xth pixel; x is the pixel number; I(x) is the set of intensity values ​​corresponding to the xth pixel; C is the trace element content of the sample; Cov(·) is the covariance calculation; Var(·) is the variance calculation.

[0082] According to an embodiment of the present invention, different preset correlation coefficient thresholds include H, wherein based on different preset correlation coefficient thresholds, intensity values ​​of pixel points corresponding to correlation coefficients satisfying different preset correlation coefficient thresholds are used as target sample spectral data, including:

[0083] For the hth preset correlation coefficient threshold, the correlation coefficient is compared with the hth preset correlation coefficient threshold, the intensity value of the pixel corresponding to the hth preset correlation coefficient threshold is determined from the correlation coefficient, and the intensity value is used as the target sample spectrum data of the hth group to obtain the target sample spectrum data of the Hth group. Wherein, H is a positive integer ≥1, 1≤h≤H.

[0084] According to an embodiment of the present invention, H different preset correlation coefficient thresholds may be set for each sample trace element content. For example, for the sample trace element Mn content, and for the hth different preset correlation coefficient threshold, at least one correlation coefficient of the hth group greater than the hth different preset correlation coefficient threshold may be determined from the M correlation coefficients obtained for the sample trace element Mn content.

[0085] According to an embodiment of the present invention, the set of pixel intensity values ​​of the same wavelength corresponding to each correlation coefficient can be inferred from at least one correlation coefficient of the hth group. For example, the at least one correlation coefficient of the hth group can include 3, and then it can be determined that each of the 3 correlation coefficients corresponds to a set of pixel intensity values ​​in the N sample laser induced breakdown spectra, for example, the 1st pixel, the 5th pixel, and the Mth pixel, respectively, and each intensity value in the above set of pixel intensity values ​​is used as the target sample spectrum data of the hth group, that is, the target sample spectrum data of the hth group corresponding to the hth preset correlation coefficient threshold.

[0086] According to an embodiment of the present invention, the above method can be used to calculate H groups of target sample spectral data for the trace element Mn content of the sample, and each group of target sample spectral data corresponds to a different preset correlation coefficient threshold.

[0087] According to an embodiment of the present invention, the target sample spectral data are respectively input into the K sub-network structures of the machine learning model to obtain the predicted sample trace element content corresponding to the K sub-network structures, including:

[0088] For the target sample spectral data of the hth group and the kth sub-network structure among the K sub-network structures, the training set of the target sample spectral data of the hth group is processed using the kth sub-network structure to obtain the predicted sample trace element content corresponding to the kth sub-network structure, and obtain the predicted sample trace element content corresponding to the K sub-network structures. Wherein, K is a positive integer ≥1.

[0089] According to an embodiment of the present invention, the machine learning model includes K sub-network structures, and each sub-network structure can be a Gaussian process regression network structure, an extreme random tree network structure, a random forest network structure, or the like.

[0090] According to an embodiment of the present invention, the target sample spectral data of the hth group can be divided into a training set and a validation set according to a certain ratio. For each sample trace element, for example, for the sample trace element Mn, the training set of the target sample spectral data of the hth group can be input into the kth sub-network structure among the K sub-network structures of the machine learning model, and the predicted sample trace element content corresponding to the kth sub-network structure can be obtained, that is, the predicted Mn element content corresponding to the kth sub-network structure.

[0091] According to an embodiment of the present invention, based on the above method, K predicted sample trace element contents corresponding to the K sub-network structures in the machine learning model can be obtained, that is, the predicted Mn element contents corresponding to the K sub-network structures.

[0092] According to an embodiment of the present invention, based on the above method, the predicted Ba element content corresponding to the K sub-network structures can be obtained; for other sample trace elements, the predicted other trace element contents corresponding to the K sub-network structures can be obtained.

[0093] According to an embodiment of the present invention, according to the predicted sample trace element content and the sample trace element content corresponding to the K sub-network structures, the parameters of the K sub-network structures of the machine learning model are adjusted respectively, and the target trace element content prediction model is determined from the K sub-network structures, including:

[0094] For the target sample spectral data of the hth group, according to the predicted sample trace element content and the sample trace element content corresponding to the kth sub-network structure, the kth difference value is obtained, and based on the kth difference value, the parameters of the kth sub-network structure are adjusted to obtain the trained kth sub-network structure, and the trained K sub-network structures are obtained; for the target sample spectral data of the hth group and the trained K sub-network structures, the trained kth sub-network structure is verified using the validation set of the target sample spectral data of the hth group, and the target sample trace element content corresponding to the kth sub-network structure is obtained, and the target sample trace element content corresponding to the K sub-network structure is obtained. content; for the h-th group of target sample spectral data and the trained K sub-network structures, the k-th error value is obtained according to the target sample trace element content and the sample trace element content corresponding to the k-th sub-network structure, and K error values ​​are obtained; the sub-network structure corresponding to the k-th error value among the K error values ​​that meet the preset conditions is determined as the h-th intermediate trace element content prediction model corresponding to the h-th preset correlation coefficient threshold, and H intermediate trace element content prediction models corresponding to the K preset correlation coefficient thresholds are obtained; the target trace element content prediction model is determined from the H intermediate trace element content prediction models corresponding to the K preset correlation coefficient thresholds.

[0095] According to an embodiment of the present invention, for example, for the trace element Mn in the sample, the difference value between the predicted Mn element content corresponding to the kth sub-network structure and the trace element Mn content in the sample can be calculated to obtain the kth difference value of the kth sub-network structure. The difference value can be used to adjust the parameters of the kth sub-network structure to obtain the trained kth sub-network structure, thereby obtaining K trained sub-network structures.

[0096] According to an embodiment of the present invention, the validation set of the hth group of target sample spectral data can be input into the trained kth sub-network structure, and the trace element content of the target sample corresponding to the kth sub-network structure can be obtained.

[0097] According to an embodiment of the present invention, the target sample trace element content is a predicted result of the sample trace element content obtained in the process of verifying the sub-network result using the verification set.

[0098] According to an embodiment of the present invention, the error value of the kth sub-network may be determined using the target sample trace element content and the sample trace element content, thereby obtaining K error values.

[0099] According to an embodiment of the present invention, the preset condition may be to determine the largest error value from the K error values. The K error values ​​may be sorted in descending order, and the subnetwork structure corresponding to the largest error value may be determined as the hth intermediate trace element content prediction model corresponding to the hth preset correlation coefficient threshold.

[0100] According to an embodiment of the present invention, H preset correlation coefficient thresholds correspond to H groups of target sample spectral data. Using the above method, for the sample trace element Mn, H intermediate trace element content prediction models corresponding to the H preset correlation coefficient thresholds can be obtained.

[0101] According to an embodiment of the present invention, an intermediate trace element content prediction model with the largest error value is selected from the error values ​​corresponding to the H intermediate trace element content prediction models as the target trace element content prediction model for measuring the trace element Mn content.

[0102] According to an embodiment of the present invention, the above method may be used to determine a target trace element content prediction model for measuring the content of the trace element Ba, or a prediction model for measuring the content of other trace elements.

[0103] The present invention also provides a method for detecting trace element content, which is suitable for detecting element content in a Mars surface composition detector. The method includes: inputting target spectral data into a target trace element content prediction model to obtain a target trace element content corresponding to the target spectral data; wherein the target trace element content prediction model is trained according to any of the above methods.

[0104] According to an embodiment of the present invention, a plurality of spectral data of Martian surface materials collected by a laser induced breakdown spectrometer carried by a Mars rover in a Martian environment are preprocessed, that is, specifically, the preprocessing may include: deducting the dark background of the spectrometer, converting the response value of the spectrometer into a radiant brightness value, converting the number of pixel bits of the spectrometer into a wavelength value, removing Gaussian white noise in the data, removing the continuous spectrum background signal in the data, splicing each spectral channel, and dividing the radiant brightness value of each pixel in the spectrum by the total area of ​​the corresponding spectral interval to achieve normalization. The preprocessed laser induced breakdown spectrum is obtained.

[0105] According to an embodiment of the present invention, for example, if it is necessary to detect the content of trace element Mn in the surface material of Mars, the multiple laser induced breakdown spectra after preprocessing can be resampled so that the wavelengths of the multiple laser induced breakdown spectra after preprocessing are consistent, and the set of pixel intensity values ​​corresponding to the correlation coefficient threshold for extracting the Mn element characteristics in the laser induced breakdown spectrum determined in the above training method can be used to extract the set of pixel intensity values ​​of the same wavelength from the multiple laser induced breakdown spectra as the target spectrum data.

[0106] According to an embodiment of the present invention, the target spectral data can be input into a module for detecting the Mn element content in a machine learning model, and the module is used to preload a target trace element content prediction model to detect the Mn element content in Martian surface materials.

[0107] Figure 4 A schematic diagram of a method for training a trace element content prediction model according to an embodiment of the present invention is shown.

[0108] like Figure 4 As shown, in the schematic diagram 400, a spectrum library 401 of ground laser induced breakdown spectroscopy (LIBS) is established, and the spectrum library includes sample trace element contents (e.g., Mn element content and Ba element content) 401-1 contained in ground samples, and collected ground LIBS spectra 401-2; for example, taking the sample trace element Mn as an example, the ground LIBS spectrum 401-2 is subjected to data preprocessing 402, and the preprocessed LIBS spectrum data is subjected to LIBS spectrum resampling 403, and the correlation coefficient between the resampled LIBS spectrum and the content of the sample trace element Mn is calculated, and the correlation coefficient and different preset correlation coefficient thresholds are used to determine the set of pixel intensity values ​​corresponding to different preset correlation coefficient thresholds related to the sample trace element Mn in the ground LIBS spectrum, that is, the target sample spectrum data 404 corresponding to different preset correlation coefficient thresholds; the target sample spectrum data 404 are respectively input into a plurality of sub-network structures 405 in the machine learning model for training, for example, using a Gaussian process regression (GPR) network structure, a random forest (RF) network structure, and a random forest (RNF) network structure. The network structure and the extreme random tree (XT) network structure are taken as examples to obtain a prediction model for each different preset correlation coefficient threshold, and a target trace element content prediction model 406 for detecting the trace element Mn is determined from the prediction model of each different preset correlation coefficient threshold; the Mars on-orbit detection LIBS spectrum data 407 is processed through data 402 to 404 to obtain target spectrum data, and the target spectrum data is input into the target trace element content prediction model to detect the trace element Mn content 408.

[0109] According to an embodiment of the present invention, Table 1 shows an example of the application of the target trace element content prediction model trained by the training method proposed by the present invention to the detection of trace element content in the LIBS observation of the Mars rover.

[0110] Table 1

[0111]

[0112] Among them, the predicted values ​​of trace element contents in Table 1 are the average values ​​of the predicted results of each trace element content in multiple Martian days. The actual values ​​of the contents in Table 1 are measured by an inductively coupled plasma mass spectrometer in a simulated Martian environment on the ground. It can be seen from Table 1 that in the prediction of the on-orbit calibration samples with known composition, the absolute error of the mass fraction of Mn and Ba contents is usually less than 1 / 10,000, and the maximum absolute error does not exceed 4 / 10,000. It can be seen that the applicability of this method has been verified in the actual Martian environment.

[0113] Figure 5 A block diagram of a training device for a trace element content prediction model according to an embodiment of the present invention is shown.

[0114] like Figure 5 As shown, the device 500 is suitable for predicting the element content in a Mars surface composition detector, and includes: an acquisition module 510, a sampling module 520, a first determination module 530, a second determination module 540, an acquisition module 550 and an adjustment module 560.

[0115] The acquisition module 510 is used to acquire the content of trace elements in the sample and N laser induced breakdown spectra of the samples corresponding to the sample, where N is an integer ≥1.

[0116] The sampling module 520 is used to resample each sample laser induced breakdown spectrum of the N sample laser induced breakdown spectra to obtain normalized N sample laser induced breakdown spectra, wherein the wavelength values ​​of the normalized N sample laser induced breakdown spectra are consistent.

[0117] The first determination module 530 is used to determine the correlation coefficient between the sample trace element content and the set of pixel intensity values ​​at the same wavelength in the normalized N sample laser induced breakdown spectra.

[0118] The second determination module 540 is used to determine, based on different preset correlation coefficient thresholds, a set of pixel intensity values ​​of the same wavelength corresponding to correlation coefficients that meet different preset correlation coefficient thresholds as target sample spectral data.

[0119] The acquisition module 550 is used to input the target sample spectral data corresponding to different preset correlation coefficient thresholds into the K sub-network structures of the machine learning model respectively, and obtain the predicted sample trace element content corresponding to the K sub-network structures respectively.

[0120] The adjustment module 560 is used to adjust the parameters of the K sub-network structures of the machine learning model according to the predicted sample trace element content and the sample trace element content corresponding to the K sub-network structures, and determine the target trace element content prediction model from the K sub-network structures.

[0121] According to an embodiment of the present invention, each sample laser induced breakdown spectrum includes M pixels.

[0122] According to an embodiment of the present invention, the sampling module 520 includes: a wavelength calculation submodule, an intensity value calculation submodule and a spectrum acquisition submodule.

[0123] The wavelength calculation submodule is used to calculate the average wavelength of the m-th pixel point among the M pixel points in the laser induced breakdown spectrum of each sample, so as to obtain the average wavelength of the M pixel points.

[0124] The intensity value calculation submodule is used to calculate the intensity value of the average wavelength of the mth pixel point of each sample laser induced breakdown spectrum in the N sample laser induced breakdown spectra by linear interpolation to obtain the intensity value of the average wavelength of M pixel points.

[0125] The spectrum acquisition submodule is used to obtain normalized laser-induced breakdown spectra of N samples based on the intensity value and average wavelength of the average wavelength of M pixel points.

[0126] According to an embodiment of the present invention, the first determination module 530 includes: a covariance operation submodule, a product operation submodule and a correlation coefficient determination submodule.

[0127] The covariance operation submodule is used to perform a covariance operation on the trace element content of the sample and a set of pixel intensity values ​​of the same wavelength to obtain a first operation result.

[0128] The product operation submodule is used to perform a product operation on the trace element content of the sample and the variance value of the set of pixel intensity values ​​of the same wavelength to obtain a second operation result.

[0129] The correlation coefficient determination submodule is used to determine the correlation coefficient between the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength according to the first operation result and the second operation result.

[0130] According to an embodiment of the present invention, the different preset correlation coefficient thresholds include H ones.

[0131] According to an embodiment of the present invention, the second determination module 540 includes: a threshold comparison submodule.

[0132] The threshold comparison submodule is used to compare the correlation coefficient with the h-th preset correlation coefficient threshold for the h-th preset correlation coefficient threshold, determine the intensity value of the pixel point corresponding to the h-th preset correlation coefficient threshold from the correlation coefficient, and use the intensity value as the target sample spectral data of the h-th group to obtain the target sample spectral data of the H-th group.

[0133] According to an embodiment of the present invention, the obtaining module 550 includes: an element content obtaining submodule.

[0134] The element content acquisition submodule is used to process the training set of the target sample spectral data of the hth group and the kth sub-network structure among the K sub-network structures using the kth sub-network structure to obtain the predicted sample trace element content corresponding to the kth sub-network structure, and obtain the predicted sample trace element content corresponding to the K sub-network structures.

[0135] According to an embodiment of the present invention, the adjustment module 560 includes: an adjustment submodule, a verification submodule, an error value acquisition submodule, a prediction model first determination submodule and a prediction model second determination submodule.

[0136] The self-adjusting module is used to obtain the kth difference value for the target sample spectral data of the hth group according to the predicted sample trace element content and the sample trace element content corresponding to the kth sub-network structure, and based on the kth difference value, adjust the parameters of the kth sub-network structure to obtain the trained kth sub-network structure, and obtain the trained K sub-network structures.

[0137] The verification submodule is used to verify the trained kth subnetwork structure using the verification set of the hth group of target sample spectral data and the trained K subnetwork structures, and obtain the target sample trace element content corresponding to the kth subnetwork structure, and obtain the target sample trace element content corresponding to the K subnetwork structures.

[0138] The error value obtaining submodule is used to obtain the kth error value according to the target sample spectral data of the hth group and the trained K sub-network structures, and obtain K error values ​​according to the target sample trace element content and the sample trace element content corresponding to the kth sub-network structure;

[0139] The first determination submodule of the prediction model is used to determine the subnetwork structure corresponding to the kth error value among the K error values ​​that meet the preset conditions as the hth intermediate trace element content prediction model corresponding to the hth preset correlation coefficient threshold, and obtain H intermediate trace element content prediction models corresponding to the K preset correlation coefficient thresholds.

[0140] The second prediction model determination submodule determines a target trace element content prediction model from H intermediate trace element content prediction models corresponding to K preset correlation coefficient thresholds.

[0141] The present invention also proposes a trace element content detection device, which is suitable for element content detection in a Mars surface composition detector. The device includes: an input module, used to input target spectral data into a target trace element content prediction model to obtain a target trace element content corresponding to the target spectral data; wherein the target trace element content prediction model is trained according to the above-mentioned device.

[0142] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems on chips, systems on substrates, systems on packages, application specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging the circuit, or by any one of the three implementation methods of software, hardware, and firmware, or by a proper combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be executed.

[0143] For example, any of the acquisition module 510, the sampling module 520, the first determination module 530, the second determination module 540, the acquisition module 550, and the adjustment module 560 or the input module can be combined into one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of the modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present invention, at least one of the acquisition module 510, the sampling module 520, the first determination module 530, the second determination module 540, the acquisition module 550 and the adjustment module 560 or the input module may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging the circuit, or implemented in any one of the three implementation modes of software, hardware and firmware or in an appropriate combination of any of them. Alternatively, at least one of the acquisition module 510, the sampling module 520, the first determination module 530, the second determination module 540, the acquisition module 550 and the adjustment module 560 or the input module may be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function may be executed.

[0144] It should be noted that the training device part and the trace element content detection device part of the trace element content prediction model in the embodiment of the present invention correspond to the training method part and the trace element content detection method part of the trace element content prediction model in the embodiment of the present invention. The description of the training device part and the trace element content detection device part of the trace element content prediction model specifically refers to the training method part and the trace element content detection method part of the trace element content prediction model, and will not be repeated here.

[0145] Figure 6 A block diagram of an electronic device suitable for implementing a training method and a detection method for a trace element content prediction model according to an embodiment of the present invention is shown. Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0146] like Figure 6As shown, the electronic device 600 according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 to a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include an onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0147] In RAM 603, various programs and data required for the operation of electronic device 600 are stored. Processor 601, ROM 602 and RAM 603 are connected to each other via bus 604. Processor 601 performs various operations of the method flow according to the embodiment of the present invention by executing the programs in ROM 602 and / or RAM 603. It should be noted that the program can also be stored in one or more memories other than ROM 602 and RAM 603. Processor 601 can also perform various operations of the method flow according to the embodiment of the present invention by executing the programs stored in one or more memories.

[0148] According to an embodiment of the present invention, the electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to the bus 604. The electronic device 600 may further include one or more of the following components connected to the input / output (I / O) interface 605: an input portion 606 including a keyboard, a mouse, etc.; an output portion 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 608 including a hard disk, etc.; and a communication portion 609 including a network interface card such as a LAN card, a modem, etc. The communication portion 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output (I / O) interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage portion 608 as needed.

[0149] According to an embodiment of the present invention, the method flow according to an embodiment of the present invention can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, the above-mentioned functions defined in the system of the embodiment of the present invention are executed. According to an embodiment of the present invention, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0150] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiment; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0151] According to an embodiment of the present invention, the computer-readable storage medium may be non-volatile. For example, it may include, but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0152] For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 602 and / or the RAM 603 described above and / or one or more memories other than the ROM 602 and the RAM 603 .

[0153] An embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method provided by the embodiment of the present invention. When the computer program product runs on an electronic device, the program code is used to enable the electronic device to implement the method provided by the embodiment of the present invention.

[0154] When the computer program is executed by the processor 601, the above functions defined in the system / device of the embodiment of the present invention are executed. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0155] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0156] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0157] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It can be understood by those skilled in the art that the features recorded in the various embodiments of the present invention can be combined and / or combined in various ways, even if such a combination or combination is not explicitly recorded in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features recorded in the various embodiments of the present invention can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present invention.

[0158] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination advantageously. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A method for training a trace element content prediction model, suitable for predicting element content in a Mars surface composition detector, characterized in that: include: Obtaining the sample trace element content contained in the sample and N sample laser induced breakdown spectra corresponding to the sample, wherein N is an integer ≥1; Resampling each of the N sample laser induced breakdown spectra to obtain normalized N sample laser induced breakdown spectra, wherein the wavelength values ​​of the normalized N sample laser induced breakdown spectra are consistent; Determine the correlation coefficient between the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength in the normalized N laser induced breakdown spectra of the sample according to the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength; Based on different preset correlation coefficient thresholds, determining a set of pixel intensity values ​​of the same wavelength corresponding to the correlation coefficients that meet the different preset correlation coefficient thresholds as target sample spectral data; For the target sample spectral data corresponding to different preset correlation coefficient thresholds, the target sample spectral data are respectively input into K sub-network structures of the machine learning model to obtain the predicted sample trace element contents corresponding to the K sub-network structures, wherein K is a positive integer ≥ 1; According to the predicted sample trace element content and the sample trace element content respectively corresponding to the K sub-network structures, respectively, adjusting the parameters of the K sub-network structures of the machine learning model, and determining a target trace element content prediction model from the K sub-network structures; Wherein, each of the sample laser induced breakdown spectra contains M pixels; Resampling each of the N sample laser induced breakdown spectra to obtain normalized N sample laser induced breakdown spectra includes: For each of the sample laser induced breakdown spectra, the average wavelength of the m-th pixel is calculated to obtain the average wavelength of the M pixel points, where M is a positive integer ≥ 1, and 1≤m≤M; Calculating the intensity value of the average wavelength of the m-th pixel point of each of the N sample laser induced breakdown spectra by linear interpolation to obtain the intensity value of the average wavelength of M pixel points; Based on the intensity value and the average wavelength of the M pixel points, obtaining the normalized N laser-induced breakdown spectra of the samples; Wherein, determining the correlation coefficient between the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength in the normalized N laser induced breakdown spectra of the sample, comprises: Performing a covariance operation on the trace element content of the sample and the set of pixel intensity values ​​at the same wavelength to obtain a first operation result; Performing a product operation on the trace element content of the sample and the variance value of the set of pixel intensity values ​​at the same wavelength to obtain a second operation result; According to the first operation result and the second operation result, a correlation coefficient between the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength is determined.

2. The method according to claim 1, characterized in that The different preset correlation coefficient thresholds include H, wherein based on the different preset correlation coefficient thresholds, the intensity value of the pixel point corresponding to the correlation coefficient that satisfies the different preset correlation coefficient thresholds is used as the target sample spectral data, including: For the h-th preset correlation coefficient threshold, the correlation coefficient is compared with the h-th preset correlation coefficient threshold, the intensity value of the pixel point corresponding to the h-th preset correlation coefficient threshold is determined from the correlation coefficient, and the intensity value is used as the target sample spectral data of the h-th group to obtain the target sample spectral data of the H-th group, wherein H is a positive integer ≥1, and 1≤h≤H.

3. The method according to claim 2, characterized in that Inputting the target sample spectral data into the K sub-network structures of the machine learning model respectively, and obtaining the predicted sample trace element contents corresponding to the K sub-network structures respectively, including: For the target sample spectral data of the hth group and the kth sub-network structure among the K sub-network structures, the kth sub-network structure is used to perform data processing on the training set of the target sample spectral data of the hth group to obtain the predicted sample trace element content corresponding to the kth sub-network structure, and to obtain the predicted sample trace element content corresponding to the K sub-network structures.

4. The method according to claim 3, characterized in that According to the predicted sample trace element content and the sample trace element content respectively corresponding to the K sub-network structures, the parameters of the K sub-network structures of the machine learning model are adjusted respectively, and a target trace element content prediction model is determined from the K sub-network structures, including: For the target sample spectral data of the hth group, according to the predicted sample trace element content corresponding to the kth sub-network structure and the sample trace element content, a kth difference value is obtained, and based on the kth difference value, a parameter of the kth sub-network structure is adjusted to obtain a trained kth sub-network structure, and K trained sub-network structures are obtained; For the h-th group of target sample spectral data and the K trained sub-network structures, the trained k-th sub-network structure is verified using the validation set of the h-th group of target sample spectral data to obtain the target sample trace element content corresponding to the k-th sub-network structure, and obtain the target sample trace element content corresponding to the K sub-network structures; For the h-th group of target sample spectral data and the K trained sub-network structures, obtain the k-th error value according to the target sample trace element content corresponding to the k-th sub-network structure and the sample trace element content, and obtain K error values; Determine the subnetwork structure corresponding to the kth error value among the K error values ​​that meet the preset conditions as the hth intermediate trace element content prediction model corresponding to the hth preset correlation coefficient threshold, and obtain H intermediate trace element content prediction models corresponding to the K preset correlation coefficient thresholds; A target trace element content prediction model is determined from H intermediate trace element content prediction models corresponding to the K preset correlation coefficient thresholds.

5. A method for detecting trace element content, suitable for detecting element content in a Mars surface composition detector, characterized in that: include: The target spectral data is input into a target trace element content prediction model to obtain a target trace element content corresponding to the target spectral data; wherein the target trace element content prediction model is trained according to the method according to any one of claims 1-4.

6. A training device for a trace element content prediction model, suitable for predicting element content in a Mars surface composition detector, characterized in that: include: An acquisition module, used to acquire the sample trace element content contained in the sample and N sample laser induced breakdown spectra corresponding to the sample, wherein N is an integer ≥1; A sampling module, used for resampling each of the N sample laser induced breakdown spectra to obtain normalized N sample laser induced breakdown spectra, wherein the wavelength values ​​of the normalized N sample laser induced breakdown spectra are consistent; A first determination module is used to determine the correlation coefficient between the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength in the normalized N laser induced breakdown spectra of the sample according to the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength; A second determination module is used to determine, based on different preset correlation coefficient thresholds, a set of pixel intensity values ​​of the same wavelength corresponding to the correlation coefficients that meet the different preset correlation coefficient thresholds as target sample spectral data; An acquisition module is used to input the target sample spectral data corresponding to different preset correlation coefficient thresholds into K sub-network structures of the machine learning model respectively, and obtain the predicted sample trace element content corresponding to the K sub-network structures respectively; An adjustment module, for adjusting parameters of the K sub-network structures of the machine learning model according to the predicted sample trace element contents and the sample trace element contents respectively corresponding to the K sub-network structures, and determining a target trace element content prediction model from the K sub-network structures; Wherein, each of the sample laser induced breakdown spectra contains M pixel points, and the sampling module includes: A wavelength calculation submodule, for calculating the average wavelength of the m-th pixel point among the M pixel points in each of the sample laser induced breakdown spectra, to obtain the average wavelength of the M pixel points, wherein M is a positive integer ≥1, and 1≤m≤M; an intensity value calculation submodule, used to calculate the intensity value of the average wavelength of the m-th pixel point of each of the N sample laser induced breakdown spectra by linear interpolation to obtain the intensity value of the average wavelength of M pixel points; A spectrum acquisition submodule, used for obtaining normalized N laser-induced breakdown spectra of the samples based on the intensity value and the average wavelength of the average wavelength of the M pixel points; Wherein, the first determining module includes: A covariance operation submodule, used for performing a covariance operation on the sample trace element content and the set of pixel intensity values ​​of the same wavelength to obtain a first operation result; A product operation submodule, used for performing a product operation on the trace element content of the sample and the variance value of the set of pixel intensity values ​​of the same wavelength to obtain a second operation result; The correlation coefficient determination submodule is used to determine the correlation coefficient between the trace element content of the sample and the set of pixel intensity values ​​of the same wavelength according to the first operation result and the second operation result.

7. A trace element content detection device, suitable for element content detection in a Mars surface composition detector, characterized in that: include: An input module is used to input target spectral data into a target trace element content prediction model to obtain a target trace element content corresponding to the target spectral data; wherein the target trace element content prediction model is trained according to the device according to claim 6.

8. An electronic device comprising: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

Images