Multi-parameter spectral characteristic wavelength combined screening method and device and electronic equipment

Abstract

The invention provides a multi-parameter spectral characteristic wavelength combined screening method and device and electronic equipment, and relates to the technical field of computers. The multi-parameter spectral characteristic wavelength combined screening method comprises the following steps: acquiring respective numerical information of a plurality of associated parameters of a sample of an object to be tested; fusing the numerical information of the plurality of associated parameters to form a fused coding result; according to the fusion coding result, screening out a target wavelength subset from the spectral data of the preset wave band; executing a prediction task for a target parameter of the to-be-tested object by using the spectral data corresponding to the target wavelength subset; wherein the target parameter is related to the multiple associated parameters. According to the method, spectral feature screening of the research object parameter prediction model can be efficiently and accurately carried out so as to meet analysis requirements under a multi-parameter and multi-factor complex scene.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, and electronic device for joint screening of multi-parameter spectral characteristic wavelengths. Background Technology

[0002] In the field of modern spectral analysis, spectral feature selection is a crucial step in improving the accuracy of predicting the parameters of research objects. However, traditional spectral feature selection methods often rely on experience, statistical indicators, or single-variable algorithms, such as genetic algorithms and particle swarm optimization algorithms, which search for the optimal subset of wavelengths through iterative optimization; and stepwise regression methods, which select features by gradually adding or removing wavelengths.

[0003] While traditional spectral feature selection methods have improved the accuracy and robustness of parameter prediction models to some extent, they all have limitations. Especially when dealing with complex scenarios involving multiple parameters or factors, traditional spectral feature selection methods struggle to comprehensively consider all factors, potentially leading to the selection of wavelengths with weak relationships to the target parameter, thus affecting the model's generalization ability. For example, in multi-parameter prediction tasks simultaneously forecasting moisture and particle size, existing methods typically focus only on the characteristic wavelengths of a single parameter, neglecting the impact of moisture content changes on particle size spectral performance, and vice versa, resulting in an inability to accurately capture the complex interactions between parameters.

[0004] Therefore, how to efficiently and accurately screen the spectral features of the research object parameter prediction model to meet the analysis needs of complex scenarios with multiple parameters and factors is an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides a method, apparatus, and electronic device for joint screening of multi-parameter spectral characteristic wavelengths to overcome the above-mentioned deficiencies in the prior art, so as to efficiently and accurately screen the spectral characteristics of the parameter prediction model of the research object, and meet the analysis needs in complex scenarios with multiple parameters and multiple factors.

[0006] This invention provides a method for joint screening of multi-parameter spectral characteristic wavelengths, comprising the following steps.

[0007] The numerical information of multiple associated parameters of the sample of the test object is obtained; the numerical information of the multiple associated parameters is fused to form a fusion coding result; based on the fusion coding result, a target wavelength subset is selected from the spectral data of a preset band; using the spectral data corresponding to the target wavelength subset, a prediction task for the target parameter of the test object is performed; wherein the target parameter is related to the multiple associated parameters.

[0008] According to the multi-parameter spectral characteristic wavelength joint screening method provided by the present invention, the step of fusing the numerical information of the multiple associated parameters to form a fusion coding result includes: discretizing the numerical information of each of the multiple associated parameters to obtain multiple discrete values; and weighting and summing the multiple discrete values ​​to obtain the fusion coding result.

[0009] According to the present invention, a multi-parameter spectral characteristic wavelength joint screening method is provided, wherein the test object is soil, and the multiple associated parameters include soil moisture content and particle size value; the method of discretizing the numerical information of each of the multiple associated parameters to obtain multiple discrete values ​​includes: determining a first discrete value of moisture content based on the interval to which the moisture content of the soil sample belongs; determining a second discrete value of particle size value based on the particle size classification threshold of the soil sample; and the method of weighted summing of the multiple discrete values ​​to obtain the fusion encoding result includes: weighted summing of the first discrete value and the second discrete value to obtain the fusion encoding result of the moisture content and particle size value of the soil sample.

[0010] According to a multi-parameter spectral characteristic wavelength joint screening method provided by the present invention, the step of determining the first discrete value of the moisture content based on the interval to which the moisture content of the soil sample belongs includes: determining the first discrete value of the moisture content based on the median value of the interval to which the moisture content of the soil sample belongs.

[0011] According to the present invention, a multi-parameter spectral feature wavelength joint screening method is provided, wherein the step of screening a target wavelength subset from spectral data of a preset band based on the fusion coding result includes: using a partial least squares regression algorithm to perform principal component decomposition on the fusion coding result and the preprocessed spectral data to extract the principal component score matrix and the loading matrix; calculating the variable importance index for each wavelength feature in the spectral data based on the principal component score matrix and the loading matrix; and selecting the target wavelength subset from the spectral data based on the variable importance index of each wavelength feature in the spectral data.

[0012] According to a multi-parameter spectral feature wavelength joint screening method provided by the present invention, the step of screening the target wavelength subset from the spectral data based on the variable importance index of each wavelength feature in the spectral data includes: screening the first target wavelength subset from the spectral data based on the variable importance index of each wavelength feature in the spectral data; for each of the plurality of correlation parameters, performing the following operations: based on the spectral data corresponding to the first target wavelength subset, using the prediction model corresponding to the correlation parameter to obtain the predicted value of the correlation parameter; if the accuracy of the predicted value of the correlation parameter is lower than a preset threshold, re-screening the second target wavelength subset from the spectral data based on the variable importance index of each wavelength feature in the spectral data, and re-executing the prediction model corresponding to the correlation parameter based on the second target wavelength subset to obtain the predicted value of the correlation parameter and subsequent steps, until the accuracy of the predicted value of the correlation parameter is higher than the preset threshold, and taking the second target wavelength subset at this time as the final target wavelength subset.

[0013] The present invention also provides a multi-parameter spectral characteristic wavelength joint screening device, comprising: The module includes an acquisition module for acquiring numerical information of multiple associated parameters of a sample of the test object; a fusion module for fusing the numerical information of the multiple associated parameters to form a fusion encoding result; a filtering module for filtering a subset of target wavelengths from spectral data of a preset band based on the fusion encoding result; and an execution module for using the spectral data corresponding to the subset of target wavelengths to execute a prediction task for the target parameters of the test object; wherein the target parameters are related to the multiple associated parameters.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-parameter spectral characteristic wavelength joint screening method as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-parameter spectral characteristic wavelength joint screening method as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multi-parameter spectral characteristic wavelength joint screening method as described above.

[0017] The multi-parameter spectral characteristic wavelength joint screening method, apparatus, and electronic device provided by this invention acquires the numerical information of multiple associated parameters of the sample of the test object, fuses the numerical information of multiple associated parameters to form a fusion coding result. Since the fusion coding result accurately characterizes the fusion characteristics of multiple associated parameters, it can effectively avoid the prediction bias caused by a single parameter. Based on the fusion coding result, a subset of target wavelengths that can accurately characterize the fusion characteristics of the multiple associated parameters is screened from a large amount of spectral data. Thus, using the spectral data corresponding to the target wavelength subset, the prediction task of the target parameters of the test object can be performed accurately and efficiently. Therefore, the spectral feature screening of the parameter prediction model of the research object can be performed efficiently and accurately to meet the analysis needs in complex scenarios with multiple parameters and factors. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the multi-parameter spectral characteristic wavelength joint screening method provided by the present invention.

[0020] Figure 2 This is a flowchart illustrating the method for obtaining the fusion encoding result of multiple associated parameters provided by the present invention.

[0021] Figure 3 This is a schematic diagram of the multi-parameter spectral characteristic wavelength joint screening process provided by the present invention.

[0022] Figure 4 This is a numerical distribution diagram of the characteristic wavelengths for screening the variable importance index provided by the present invention.

[0023] Figure 5 This is a schematic diagram of soil particle size classification accuracy obtained using the embodiments provided by the present invention.

[0024] Figure 6 This is a schematic diagram of soil moisture classification accuracy obtained using the embodiments provided by the present invention.

[0025] Figure 7 This is a schematic diagram of the structure of the multi-parameter spectral characteristic wavelength joint screening device provided by the present invention.

[0026] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0028] The following is combined Figures 1-2 The present invention describes a multi-parameter spectral characteristic wavelength joint screening method.

[0029] Figure 1 This is a schematic flowchart of the multi-parameter spectral characteristic wavelength joint screening method provided by the present invention. Figure 1 As shown, the method includes the following: Step 101: Obtain the numerical information of each of the multiple associated parameters of the sample of the object to be tested.

[0030] The test object refers to the target entity whose specific parameters (such as composition, physical properties, etc.) need to be obtained through spectral analysis. It is usually a complex sample with multi-parameter coupling characteristics. For example, soil samples in agriculture (which need to be analyzed simultaneously for moisture, particle size, total nitrogen content, etc.), grain particles in food processing (which need to be tested for protein, moisture, hardness, etc.), and tablets in pharmaceuticals (which need to be measured for active ingredient content, moisture, density, etc.).

[0031] Correlated parameters refer to multiple parameters that are related to the characteristics of the object under test and have mutual influences, and their numerical changes will jointly affect the spectral characteristics.

[0032] Numerical information refers to the specific measured value or classification code value (such as discrete intervals or continuous values) for each parameter.

[0033] For example, for soil samples, the associated parameters include: moisture content and particle size. The moisture content values ​​are: low (0-4%), medium (5-8%), and high (above 9%); the particle size values ​​are: fine (≤0.3mm), medium (0.45mm), coarse (0.9mm), and very coarse (2mm).

[0034] Step 102: Merge the numerical information of multiple related parameters to form a fusion coding result.

[0035] The fusion coding result refers to a single structured numerical sequence obtained by converting the numerical information of multiple related parameters through a preset algorithm, so as to reflect the coupling relationship between multiple parameters in the form of structured data.

[0036] In practice, multiple methods can be used, such as weighted summation, to fuse the numerical information of multiple related parameters and form a fused encoding result, which is not limited to the description in this specification.

[0037] For an example of fusing numerical information from multiple related parameters to form a fused encoding result, see [link to example]. Figure 2 The relevant content will not be repeated here.

[0038] Step 103: Based on the fusion coding results, select a subset of target wavelengths from the spectral data of the preset bands.

[0039] Preset-band spectral data refers to a continuous reflectance or absorbance data sequence collected within a specific wavelength range using a spectroscopic instrument (such as a near-infrared spectrometer). The wavelength range is pre-set by the instrument's hardware parameters (such as spectral resolution and detector type) and the experimental design, covering the key characteristic absorption peaks of the object being tested (such as soil or agricultural products).

[0040] In the specific implementation process, for soil samples, spectral data of preset bands can be obtained through the following methods.

[0041] like Figure 3 As shown, soil samples were collected from the farm, brought back to the laboratory, dried, and then ground.

[0042] The laboratory soil sample processing included air drying and sieving. To ensure different moisture gradients in the samples, 10 samples were placed in separate sample containers, and the soil was turned over periodically with tweezers, with each turning interval being one week. A precision electronic balance was used to record the actual weight of each sample at each interval, for a total of three weeks. The actual moisture content of each sample was finally determined by the difference in weight before and after each interval. For ease of input and integration, the median moisture content of all samples was taken and divided into three moisture ranges: 2%, 6.5%, and 11%. The laboratory soil samples were sieved using existing sieves with mesh sizes of 10 mesh, 20 mesh, 40 mesh, and 60 mesh. After sieving, each of the 10 samples was divided into four particle size gradients: 2 mm, 0.9 mm, 0.45 mm, and 0.3 mm.

[0043] Spectral data of the samples were acquired using a laboratory hyperspectral analyzer HS1-eSWIR-400-2500H. This hyperspectral analyzer has two internal lenses and can measure spectral data in the 350nm-1100nm range. The built-in moving platform, controlled by the hyperspectral analyzer's acquisition software, was used to position the soil samples directly below the 350nm-1100nm lenses to obtain spectral data values ​​in the 350nm-1100nm band.

[0044] Since the acquired spectral signals may be affected by noise, baseline drift and other factors, it is necessary to preprocess the acquired spectral data.

[0045] In the specific implementation process, the best preprocessing method for spectral data can be selected through the following methods.

[0046] In the embodiments provided by this invention, Savitzky-Golay Smoothing, Standard Normal Variate Transform, Multiplicative Scatter Correction, and First Derivative, as well as their pairwise combinations, are used to preprocess the spectral data. The preprocessed spectral data is then modeled and analyzed (see the process below of using partial least squares regression to select a subset of target wavelengths from the spectral data). The modeling results of the preprocessed spectral data are compared with those of the original spectral data. Finally, Savitzky-Golay Smoothing and Standard Normal Variate Transform are selected as the optimal preprocessing methods.

[0047] The target wavelength subset refers to the set of wavelength points that are significantly related to the fusion coding result, selected from the spectral data of a preset band through feature filtering algorithms (such as VIP value, PLSR).

[0048] In some embodiments, partial least squares regression algorithm can be used to perform principal component decomposition on the fused coding results and preprocessed spectral data to extract the principal component score matrix and loading matrix.

[0049] In the specific implementation process, the preprocessed spectral data is used as the independent variable X, and the fusion encoding result is used as the dependent variable Y. A partial least squares regression (PLSR) model is adopted for modeling. PLSR extracts principal components iteratively, maximizes the covariance between the independent and dependent variables, and establishes a linear relationship model between the two. Finally, the score matrix T (the projection of the sample into the principal component space) and the loading matrix P (the contribution weights of the independent variables to the principal components) are extracted from the trained PLSR model.

[0050] Based on the principal component score matrix and loading matrix, the VIP (Variable Importance in the Projection) is calculated for each wavelength feature in the spectral data. The VIP value measures the contribution of each independent variable to the model's prediction, and the specific calculation formula is as follows: (1) in, Let be the VIP value of the j-th wavelength feature; p is the total number of variables; A is the number of principal components used in the model; The variance explained by the a-th principal component (or the contribution determined by the model). This is the loading of the j-th variable on the a-th principal component.

[0051] Based on the importance index of the variables of each wavelength feature in the spectral data, a subset of target wavelengths is obtained by screening from the spectral data.

[0052] In the specific implementation process, the following screening criteria can be used as a basis, such as... Figure 4 The importance index of each wavelength feature in the spectral data shown was used to filter and obtain a subset of target wavelengths: VIP threshold screening retains wavelength features with VIP values ​​greater than a preset threshold (e.g., 1.0) to ensure they have significant predictive capabilities.

[0053] Cross-validation standard error (cvSE) screening retains only wavelength features with cvSE less than a preset error threshold (e.g., 0.075).

[0054] cvSE is used to quantify the importance of variables (such as...) Figure 4 As shown in the figure, it reflects the uncertainty of the VIP value in repeated modeling: Error range: the greater the fluctuation of the VIP value of the variable in cross-validation, the less stable its importance; stability: the more reliable the VIP value estimate, the more stable the variable's importance.

[0055] In some embodiments, to obtain more accurate wavelength features from spectral data, a first target wavelength subset can be obtained from the spectral data based on the variable importance index of each wavelength feature in the spectral data; for each of the multiple correlation parameters, the following operations are performed: Based on the spectral data corresponding to the first target wavelength subset, the prediction model corresponding to the correlation parameter is used to obtain the predicted value of the correlation parameter. If the accuracy of the predicted value of the correlation parameter is lower than the preset threshold, the second target wavelength subset is obtained from the spectral data by re-selecting the variable importance index of each wavelength feature in the spectral data. Based on the second target wavelength subset, the prediction model corresponding to the correlation parameter is re-executed to obtain the predicted value of the correlation parameter and subsequent steps, until the accuracy of the predicted value of the correlation parameter is higher than the preset threshold. The second target wavelength subset at this time is taken as the final target wavelength subset.

[0056] In some embodiments, for soil samples, a soil particle size prediction model can be constructed based on an SVM (Support Vector Machine) model to further validate the selected target wavelength subset.

[0057] In the specific implementation process, the spectral data of soil samples can be used as sample data, and the particle size values ​​of soil samples can be used as labels to establish a training dataset. The initial SVM model is then trained using the training dataset to obtain a soil particle size prediction model.

[0058] During the selection of target wavelength subsets, the first or second target wavelength subset can be transformed into a feature vector suitable for processing by the soil particle size prediction model. This vector is then input into the soil particle size prediction model to obtain the predicted soil particle size output by the model. The predicted value is then compared with the actual soil particle size value to calculate the prediction accuracy.

[0059] In some embodiments, for soil samples, a moisture content prediction model can be constructed based on an SVM (Support Vector Machine) model to further validate the selected target wavelength subset.

[0060] In the specific implementation process, the spectral data of soil samples can be used as sample data, and the moisture content of soil samples can be used as the label of the sample data to establish a training dataset. The initial SVM model is then trained using the training dataset to obtain a moisture content prediction model.

[0061] During the selection of the target wavelength subset, the first or second target wavelength subset can be transformed into a feature vector suitable for the moisture content prediction model, and then input into the moisture content prediction model to obtain the predicted value of moisture content output by the moisture content prediction model. The predicted value is then compared with the actual moisture content to calculate the prediction accuracy.

[0062] In the embodiments provided by this invention, the fusion encoding results of multiple correlation parameters of the sample of the test object are used to screen out a subset of target wavelengths from a large amount of spectral data that can accurately characterize the fusion characteristics of these multiple correlation parameters. A support vector machine (SVM) classification model is then applied as an inversion verification model for the screening results, significantly improving the prediction accuracy of the target parameters of the test object predicted using the target wavelength subset. For example, as... Figure 5 , Figure 6 As shown, the correlation coefficient (R) of soil moisture content predicted using the target wavelength subset reaches 0.808, and the R value for particle size classification is as high as 0.942.

[0063] Furthermore, since the target wavelength subset obtained by utilizing the embodiments provided by the present invention can simultaneously predict multiple parameters of the object under test (e.g., simultaneously predict soil moisture content and particle size), the utilization efficiency of computing resources can be effectively improved.

[0064] Step 104: Using the spectral data corresponding to the target wavelength subset, perform a prediction task for the target parameters of the object under test; wherein, the target parameters are related to multiple associated parameters.

[0065] The target parameter is a predictive index that is coupled with multiple related parameters (such as soil moisture content and particle size).

[0066] As an example only, if the object to be tested is soil in a certain physical area, the target parameter can be the total nitrogen content of the soil, which is affected by the soil moisture content and particle size.

[0067] In the specific implementation process, steps 201-202 can be used to obtain the fusion encoding result based on the moisture content and particle size of the soil sample. Then, in step 103, based on the fusion encoding result, a subset of target wavelengths closely related to the soil moisture content and particle size is selected from the spectral data of the preset bands. Subsequently, based on the partial spectral data corresponding to the target wavelength subset contained in the spectral data collected from the soil to be tested, the total nitrogen content of the soil to be tested is predicted using a soil total nitrogen content model. For a detailed description of the spectral data collected from the soil to be tested, please refer to the relevant content in step 103; it will not be repeated here.

[0068] Among them, the soil total nitrogen content model is a machine learning model trained using a training dataset, such as the SVM model.

[0069] The training dataset can consist of partial spectral data corresponding to a subset of target wavelengths from spectral data collected from the soil to be tested (as samples) and the total nitrogen content of the soil to be tested (as labels). In practice, a total of 120 samples can be collected, and the collected dataset can be divided into a training dataset and a prediction dataset in a 4:1 ratio for comparison of subsequent modeling results.

[0070] Figure 2 This is a flowchart illustrating the method for obtaining the fusion encoding result of multiple associated parameters provided by the present invention. For example... Figure 2 As shown, the method includes the following: Step 201: Discretize the numerical information of each of the multiple associated parameters to obtain multiple discrete values.

[0071] In the specific implementation process, the values ​​of the parameters can be discretized by combining the classification information or interval information of the parameters.

[0072] In some embodiments, the test object is soil, and multiple associated parameters include soil moisture content and particle size. A first discrete value of moisture content can be determined based on the interval to which the moisture content of the soil sample belongs. For example, the median or boundary value of the interval to which the moisture content belongs can be used as the first discrete value of moisture content, or the median or boundary value of the interval to which the moisture content belongs can be linearly transformed and used as the first discrete value of moisture content. A second discrete value of particle size can be determined based on the particle size classification threshold of the soil sample. For example, the particle size classification threshold corresponding to the particle size value can be used as the second discrete value of particle size value, or the particle size classification threshold corresponding to the particle size value can be linearly transformed and used as the second discrete value of particle size value.

[0073] For example, moisture content can be categorized into the following ranges: low moisture (0-4%), medium moisture (5-8%), and high moisture (above 9%). If the moisture content is within (0-4%), the first discrete value of the moisture content can be encoded as any one of 2%, 6.5%, or 11%.

[0074] For example, based on particle size classification thresholds, soil sample particle sizes can be classified as: fine (0.3 mm), medium (0.45 mm), coarse (0.9 mm), and very coarse (2 mm). If a soil sample has a particle size of 1.9 mm, it belongs to very coarse particles, and the corresponding particle size classification threshold is 2 mm. Therefore, the second discrete value of the particle size value can be determined to be 2.

[0075] In some embodiments, a first discrete value of moisture content can be determined based on the median value of the interval to which the moisture content of the soil sample belongs; and a second discrete value of particle size can be determined based on the particle size classification threshold of the soil sample.

[0076] Step 202: Weighted summation of multiple discrete values ​​to obtain the fused coding result.

[0077] In some embodiments, the test object is soil. After obtaining the first discrete value of moisture content and the second discrete value of particle size in step 201, the first discrete value and the second discrete value can be weighted and summed to obtain the fused encoding result of the soil sample's moisture content and particle size. The specific formula is as follows: The fusion coding result = a × the first discrete value + b × the second discrete value (2) Where 'a' and 'b' are the weights of moisture content and particle size, respectively, and can be determined based on experience or experimental results. For example, a=1, b=10. If the median moisture content is 2% and the particle size is 2mm, the corresponding integrated column value is 22.

[0078] Based on the moisture content range and particle size classification threshold in step 201, the three discrete moisture content values ​​and four discrete particle size values ​​are combined in pairs using formula (2) to obtain fusion coding results under 12 gradients. The fusion coding results under different gradients reflect the different combinations of moisture content and particle size values ​​of the soil sample, thus characterizing the different coupling characteristics of moisture content and particle size values ​​of the soil sample.

[0079] In the embodiments provided by this invention, the numerical information of each of the multiple correlation parameters is discretized to obtain multiple discrete values; the multiple discrete values ​​are then weighted and summed to obtain a fusion coding result. Subsequently, based on this fusion coding result, a subset of target wavelengths that can accurately characterize the coupling properties of the multiple correlation parameters can be selected from the spectral data.

[0080] The multi-parameter spectral characteristic wavelength joint screening device provided by the present invention is described below. The multi-parameter spectral characteristic wavelength joint screening device described below can be referred to in correspondence with the multi-parameter spectral characteristic wavelength joint screening method described above.

[0081] Figure 7 This is a schematic diagram of the structure of the multi-parameter spectral characteristic wavelength joint screening device provided by the present invention.

[0082] like Figure 7 As shown, the multi-parameter spectral characteristic wavelength joint screening device includes the following modules: The acquisition module 710 is used to acquire the numerical information of each of the multiple associated parameters of the sample of the object to be tested.

[0083] The fusion module 720 is used to fuse the numerical information of the multiple associated parameters to form a fusion encoding result.

[0084] The filtering module 730 is used to filter out a subset of target wavelengths from the spectral data of a preset band based on the fusion coding result.

[0085] The execution module 740 is used to perform a prediction task of target parameters for the object under test using the spectral data corresponding to the target wavelength subset; wherein the target parameters are related to the plurality of associated parameters.

[0086] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840, wherein the processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute a multi-parameter spectral feature wavelength joint screening method. This method includes: acquiring numerical information of multiple associated parameters of the sample to be tested; fusing the numerical information of the multiple associated parameters to form a fusion encoding result; selecting a subset of target wavelengths from spectral data of a preset band based on the fusion encoding result; and using the spectral data corresponding to the subset of target wavelengths to perform a prediction task for the target parameters of the sample to be tested; wherein the target parameters are related to the multiple associated parameters.

[0087] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0088] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multi-parameter spectral feature wavelength joint screening method provided by the above methods. The method includes: acquiring numerical information of multiple associated parameters of a sample of a test object; fusing the numerical information of the multiple associated parameters to form a fusion encoding result; selecting a subset of target wavelengths from spectral data of a preset band based on the fusion encoding result; and using the spectral data corresponding to the subset of target wavelengths to perform a prediction task for the target parameters of the test object; wherein the target parameters are related to the multiple associated parameters.

[0089] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a multi-parameter spectral feature wavelength joint screening method provided by the above methods. The method includes: acquiring numerical information of multiple associated parameters of a sample of a test object; fusing the numerical information of the multiple associated parameters to form a fusion encoding result; selecting a subset of target wavelengths from spectral data of a preset band based on the fusion encoding result; and using the spectral data corresponding to the subset of target wavelengths to perform a prediction task for a target parameter of the test object; wherein the target parameter is related to the multiple associated parameters.

[0090] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0091] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for joint screening of multi-parameter spectral characteristic wavelengths, characterized in that, include: Obtain the numerical information of each of the multiple associated parameters of the sample of the object to be tested; The numerical information of the multiple associated parameters is fused to form a fusion coding result; Based on the fusion coding results, a subset of target wavelengths is selected from the spectral data of the preset band; Using the spectral data corresponding to the subset of target wavelengths, a task is performed to predict the target parameters of the object under test; wherein the target parameters are related to the plurality of associated parameters.

2. The multi-parameter spectral characteristic wavelength joint screening method according to claim 1, characterized in that, The process of fusing the numerical information of the multiple associated parameters to form a fusion encoding result includes: The numerical information of each of the multiple associated parameters is discretized to obtain multiple discrete values; The weighted summation of the multiple discrete values ​​yields the fusion coding result.

3. The multi-parameter spectral characteristic wavelength joint screening method according to claim 1, characterized in that, The object to be tested is soil, and the multiple associated parameters include soil moisture content and particle size. The step of discretizing the numerical information of each of the multiple associated parameters to obtain multiple discrete values ​​includes: The first discrete value of the moisture content is determined based on the range to which the moisture content of the soil sample belongs; Based on the particle size classification threshold of the soil sample, a second discrete value of the particle size is determined; The step of weighted summing of the multiple discrete values ​​to obtain the fused encoding result includes: The first discrete value and the second discrete value are weighted and summed to obtain the fusion encoding result of the moisture content and particle size value of the soil sample.

4. The multi-parameter spectral characteristic wavelength joint screening method according to claim 3, characterized in that, Determining the first discrete value of the moisture content based on the range to which the soil sample's moisture content belongs includes: The first discrete value of the moisture content is determined based on the median value of the interval to which the moisture content of the soil sample belongs.

5. The multi-parameter spectral characteristic wavelength joint screening method according to any one of claims 1 to 4, characterized in that, The step of selecting a subset of target wavelengths from the spectral data of a preset band based on the fusion coding result includes: Using partial least squares regression, principal component decomposition is performed on the fused coding result and the preprocessed spectral data to extract the principal component score matrix and loading matrix. Based on the principal component score matrix and the loading matrix, the variable importance index is calculated for each wavelength feature in the spectral data; The target wavelength subset is obtained by filtering from the spectral data based on the importance index of each wavelength feature variable in the spectral data.

6. The multi-parameter spectral characteristic wavelength joint screening method according to claim 5, characterized in that, The step of selecting the target wavelength subset from the spectral data based on the importance index of variables representing each wavelength characteristic in the spectral data includes: Based on the importance index of each wavelength feature in the spectral data, a first target wavelength subset is obtained by filtering from the spectral data; For each of the plurality of associated parameters, perform the following operation: Based on the spectral data corresponding to the first target wavelength subset, the predicted value of the correlation parameter is obtained using the prediction model corresponding to the correlation parameter; If the accuracy of the predicted value of the correlation parameter is lower than a preset threshold, a second target wavelength subset is obtained from the spectral data based on the importance index of each wavelength feature variable in the spectral data. Based on the second target wavelength subset, the prediction model corresponding to the correlation parameter is re-executed to obtain the predicted value of the correlation parameter and subsequent steps, until the accuracy of the predicted value of the correlation parameter is higher than the preset threshold. The second target wavelength subset at this time is then taken as the final target wavelength subset.

7. A multi-parameter spectral characteristic wavelength joint screening device, characterized in that, include: The acquisition module is used to acquire the numerical information of multiple associated parameters of the sample of the object to be tested; The fusion module is used to fuse the numerical information of the multiple associated parameters to form a fusion encoding result; The filtering module is used to filter out a subset of target wavelengths from the spectral data of a preset band based on the fusion coding result. An execution module is used to perform a prediction task for the target parameters of the object under test using the spectral data corresponding to the subset of target wavelengths; wherein the target parameters are related to the plurality of associated parameters.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the multi-parameter spectral characteristic wavelength joint screening method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-parameter spectral characteristic wavelength joint screening method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-parameter spectral characteristic wavelength joint screening method as described in any one of claims 1 to 6.

Images