Method and device for detecting enrichment degree of salt lake resources based on hyperspectral recognition technology

Abstract

The present invention provides a method and device for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology. Among them, the method includes: obtaining a feature vector in chronological order by collecting the spectral detection results of the brine area, and then extracting the change law of the spectrum according to the chronological order to predict the enrichment degree of the target element in the brine area. The beneficial effects of the present invention: It realizes the detection of the enrichment degree of the target element, so that it can detect whether the resource enrichment degree of the brine area meets the mining requirements, and reduces the risk of insufficient resources in mining.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and particularly to a method and device for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology. Background Art

[0002] In recent years, remote sensing technology has been gradually used for rapid and low-cost prediction of resources in salt lakes. Its principle is that when electromagnetic waves irradiate on salt lake minerals, reflection phenomena will occur on the salt lake minerals. When electromagnetic waves collide with mineral substances, electrons in the mineral substances will undergo energy level transitions, forming unique spectral characteristics. However, due to many factors such as climate change and human activities, the detection results will have certain variations, and it is impossible to determine whether the resource enrichment degree in the detection area meets the mining requirements. Therefore, there is an urgent need for a method for detecting the enrichment degree of salt lake resources. Summary of the Invention

[0003] The main object of the present invention is to provide a method and device for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology, aiming to solve the problem that it is impossible to determine whether the resource enrichment degree in the detection area meets the mining requirements.

[0004] The present invention provides a method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology, including:

[0005] Obtaining the spectral detection results of n brine areas in the salt lake and the enrichment degree of target elements in each brine area; wherein each of the spectral detection results includes spectral detection data in multiple time periods;

[0006] Calculating the characteristic vector X of each brine area according to the spectral detection data i ={(Q q，q-1 ), (Q q，q-2 ),..., (Q q，z ),...,(Q q，2 ), (Q q，1 )}, where q represents the qth time period, X i represents the characteristic vector of the ith brine area, and (Q q，z ) represents the change rate of spectral detection data between the qth time period and the zth time period, q, z, i belong to positive integers, and q>z, i≤n;

[0007] Performing standardization processing on each of the characteristic vectors according to a preset method to obtain a standard data set;

[0008] Dividing the standard data set into a training data set and a test data set according to a preset ratio;

[0009] Input the training data set and the enrichment degrees of each target element in the brine area corresponding to the training data set into a preset SVM model, and train the preset SVM model according to the optimal hyperparameters;

[0010] Detect the trained model by using the test data set and the enrichment degrees of each target element in the brine area corresponding to the test data set. When the detection result meets the training requirements of the model, obtain the target model;

[0011] Obtain the target spectral detection result of the brine area to be detected, and input the target spectral detection result into the preset target model to obtain the enrichment degree of the target element in the brine area to be detected.

[0012] Further, the step of standardizing each of the feature vectors according to a preset method to obtain a standard data set includes:

[0013] Extract the maximum and minimum points in each feature vector, and arrange them in chronological order to obtain an extreme value sequence;

[0014] Use a cubic spline interpolation function to fit the extreme value sequence to obtain upper and lower envelope lines X max (t) and X min (t);

[0015] Take the mean of the upper and lower envelope lines and denote it as the envelope line mean m(t); where,

[0016] Subtract the envelope line mean from the feature vector to obtain a target sequence;

[0017] Judge whether the target sequence passes the intrinsic mode function test;

[0018] If it passes the test, denote the target sequence as a random oscillation function. Otherwise, denote the target sequence as the first feature vector and recalculate the target sequence until the random oscillation function is obtained;

[0019] Subtract the first feature vector from the feature vector to obtain a second feature vector, and repeat to obtain multiple random oscillation functions until the calculated envelope lines are symmetric and the mean of the envelope lines is 0, thereby obtaining multiple random oscillation functions;

[0020] Aggregate the random oscillation functions of each feature vector to obtain the standard data corresponding to each feature vector, and further obtain a standard data set composed of each standard data.

[0021] Further, before the step of inputting the training data set and the enrichment degrees of various target elements in the brine area corresponding to the training data set into a preset SVM model and training the preset SVM model according to the optimal hyperparameters, the following steps are also included:

[0022] Obtain initial SVM models with multiple different combinations of hyperparameters;

[0023] Divide the training data set into a training set and a validation set according to a preset ratio;

[0024] Input the training set and the enrichment degrees of various target elements in the corresponding brine area into each of the initial SVM models for training to obtain corresponding multiple temporary SVM models;

[0025] Verify each of the temporary SVM models through the validation set to obtain the verification results of each of the temporary SVM models;

[0026] Based on the verification results, select the temporary SVM model with the optimal verification result as the preset SVM model.

[0027] Further, after the step of obtaining the target spectral detection result of the brine area to be detected and inputting the target spectral detection result into a preset target model to obtain the enrichment degree of the target element in the brine area to be detected, the following steps are also included:

[0028] Conduct chemical detection on the water samples in the brine area to be detected for multiple time periods;

[0029] Judge whether the chemical detection results of the water samples in each time period meet the preset target;

[0030] If it meets the preset target, determine that the brine area to be detected is an area to be exploited.

[0031] Further, after the step of obtaining the spectral detection results of n brine areas of the salt lake, the following steps are also included:

[0032] Select one spectral detection result from the spectral detection results of the n brine areas as the first spectral detection result;

[0033] Calculate the difference value between the first spectral detection result and other spectral detection results;

[0034] Construct a difference value set for the spectral detection results with the difference value less than or equal to the preset difference value;

[0035] Define the difference value between each element in the difference value set and the first spectral detection result as the preset difference value;

[0036] According to the formula Calculate the density of the first spectral detection result, where ρ c (O) represents the density of the first spectral detection result, N(O) represents the set of differences, P represents an element in the set of differences, O represents the first spectral detection result, and d c (O, P) represents the difference value between element P and O, and c represents the preset difference;

[0037] According to the formula Calculate the discrete score of the first spectral detection result; where LOF c (O) represents the discrete score of the first spectral detection result, and ρ c (P) represents the density of the spectral detection result corresponding to element P;

[0038] Determine whether the discrete score is greater than the preset discrete score;

[0039] If so, remove the first spectral detection result from the n spectral detection results.

[0040] The present invention also provides a detection device for the enrichment degree of salt lake resources based on hyperspectral recognition technology, including:

[0041] A first acquisition module, configured to acquire the spectral detection results of n brine areas of the salt lake and the enrichment degree of the target elements in each brine area; where each spectral detection result includes spectral detection data in multiple time periods;

[0042] A calculation module, configured to calculate the feature vector X of each brine area according to the spectral detection data i =

[0043] {(Q q，q-1 ), (Q q，q-2 ),..., (Q q，z ),..., (Q q，2 ), (Q q，1 )}, where q represents the qth time period, X i represents the feature vector of the ith brine area, and (Q q，z ) represents the change rate of the spectral detection data between the qth time period and the zth time period, q, z, i belong to positive integers, and q > z, i ≤ n;

[0044] A processing module, configured to perform standardization processing on each of the feature vectors according to a preset method to obtain a standard data set;

[0045] A splitting module, configured to divide the standard data set into a training data set and a test data set according to a preset ratio;

[0046] An input module, configured to input the training data set and the enrichment degrees of various target elements in the brine area corresponding to the training data set into a preset SVM model, and train the preset SVM model according to optimal hyperparameters;

[0047] A detection module, configured to detect the trained model by using the test data set and the enrichment degrees of various target elements in the brine area corresponding to the test data set, and obtain a target model when the detection result meets the training requirements of the model;

[0048] A second acquisition module, configured to acquire a target spectral detection result of a brine area to be detected, and input the target spectral detection result into the preset target model to obtain the enrichment degree of the target element in the brine area to be detected.

[0049] Further, the processing module includes:

[0050] An extraction sub-module, configured to extract the maximum points and minimum points in each of the feature vectors, and arrange them in chronological order to obtain an extreme value sequence;

[0051] A fitting sub-module, configured to fit the extreme value sequence by using a cubic spline interpolation function to obtain upper and lower envelope lines X max (t) and X min (t);

[0052] A value-taking sub-module, configured to take the mean value of the upper and lower envelope lines and denote it as the envelope line mean m(t);

[0053] Wherein,

[0054] A first calculation sub-module, configured to subtract the envelope line mean from the feature vector to obtain a target sequence;

[0055] A judgment sub-module, configured to judge whether the target sequence passes the intrinsic mode function test;

[0056] A recording sub-module, configured to, if passing the test, record the target sequence as a random oscillation function, otherwise, record the target sequence as a first feature vector and recalculate the target sequence until the random oscillation function is obtained;

[0057] A second calculation sub-module, configured to subtract the first feature vector from the feature vector to obtain a second feature vector, and repeat to obtain multiple random oscillation functions until the calculated envelope line is symmetric and the mean value of the envelope line is 0, so as to obtain multiple random oscillation functions;

[0058] An aggregation sub-module, configured to aggregate the random oscillation functions of each feature vector, so as to obtain standard data corresponding to each feature vector, and further obtain a standard data set composed of each standard data.

[0059] Furthermore, the detection device for the enrichment degree of salt lake resources based on hyperspectral recognition technology further includes:

[0060] A model acquisition module, configured to acquire initial SVM models with multiple different combinations of hyperparameters;

[0061] A dataset division module, configured to divide the training dataset into a training set and a validation set according to a preset ratio;

[0062] A data input module, configured to input the training set and the enrichment degrees of each target element in the corresponding brine area into each of the initial SVM models for training to obtain corresponding multiple temporary SVM models;

[0063] A model verification module, configured to verify each of the temporary SVM models through the validation set to obtain the verification results of each of the temporary SVM models;

[0064] A marking module, configured to select the temporary SVM model with the best verification result based on the verification results as the preset SVM model.

[0065] The present invention also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0066] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0067] The beneficial effects of the present invention: By collecting the spectral detection results of the brine area, characteristic vectors in chronological order are obtained, and then the change law of the spectrum is extracted according to the chronological order, so as to realize the detection of the enrichment degree of the target element, and thus it can be detected whether the resource enrichment degree of the brine area meets the mining requirements, greatly reducing the risk of insufficient resources in mining. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 is a flowchart of a method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology according to an embodiment of the present invention;

[0069] Figure 2 is a structural schematic block diagram of a detection device for the enrichment degree of salt lake resources based on hyperspectral recognition technology according to an embodiment of the present invention;

[0070] Figure 3 is a structural schematic block diagram of a computer device according to an embodiment of the present application.

[0071] The implementation, functional features, and advantages of the object of the present invention will be further described in conjunction with embodiments with reference to the accompanying drawings. Specific Embodiments

[0072] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts belong to the scope of protection of the present invention.

[0073] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly. The connection described may be a direct connection or an indirect connection.

[0074] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and B may represent: A exists alone, A and B exist simultaneously, and B exists alone.

[0075] In addition, in the present invention, descriptions such as "first" and "second" are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0076] Referring to Figure 1 , the present invention proposes a method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology, including:

[0077] S1: Obtain the spectral detection results of n brine areas of the salt lake and the enrichment degree of target elements in each brine area; where each spectral detection result includes spectral detection data in multiple time periods;

[0078] S2: Calculate the feature vector X of each brine area according to the spectral detection data i ={(Q q，q-1 ), (Q q，q-2 ),..., (Q q，z ),...,(Q q，2), (Q q，1 )}, where q represents the q-th time period, and X i represents the feature vector of the i-th brine area, and (Q q，z ) represents the change rate of the spectral detection data between the q-th time period and the z-th time period. q, z, and i are positive integers, and q > z, i ≤ n;

[0079] S3: Standardize each of the said feature vectors according to a preset method to obtain a standard data set;

[0080] S4: Divide the said standard data set into a training data set and a test data set according to a preset ratio;

[0081] S5: Input the said training data set and the enrichment degrees of the respective target elements of the brine areas corresponding to the training data set into a preset SVM model, and train the preset SVM model according to the optimal hyperparameters;

[0082] S6: Detect the trained model through the said test data set and the enrichment degrees of the respective target elements of the brine areas corresponding to the test data set. When the detection result meets the training requirements of the model, obtain the target model;

[0083] S7: Obtain the target spectral detection result of the brine area to be detected, input the said target spectral detection result into the preset target model, and obtain the enrichment degree of the target element of the brine area to be detected.

[0084] As described in the above step S1, obtain the spectral detection results of n brine regions in the salt lake, as well as the enrichment degrees of target elements in each brine region; each of the spectral detection results includes spectral detection data in multiple time periods; among them, the spectral detection results can be detected by a drone carrying a multispectral sensor, or can be obtained by satellite remote sensing data. In order to reduce the error caused by the atmosphere, it is preferably detected by a drone carrying a multispectral sensor, so as to obtain a remote sensing image, and then obtain the spectral detection results of the brine region according to the remote sensing image. The brine region refers to a region of the salt lake, and its area or volume size can be set artificially. In addition, the brine region is a known brine region, that is, accurate data detected by chemical detection methods, that is, the enrichment degree of the target element can be obtained. In addition, the enrichment degree is the content of the target element per unit volume, and whether it is enriched needs to be obtained according to the type of target element. It should be noted that the target element is a mineral element, such as lithium, uranium, magnesium and other elements. In addition, the obtained spectral detection results include spectral detection data in multiple time periods. The multiple time periods are generally the detection results in a certain period closest to the current time, so as to form a time series. Specifically, it can be a set of data collected in chronological order according to a constant sampling frequency. This can include variable traces of the dynamic evolution of the brine region, and hidden information can be extracted from it. In addition, this application only performs data statistics on a single target element, and different target elements need to train different models for prediction.

[0085] As described in the above step S2, calculate the feature vectors of each brine region according to the spectral detection data. Specifically, it can be defined that the spectral detection data at each moment is related to the spectral detection data in the previous q time periods, that is, the feature vector X is obtained. i ={(Q q，q-1 ),(Q q，q-2 ),...,(Q q，z )...,(Q q，2 ),(Q q，1 )}, where q represents the qth time period, X i represents the feature vector of the ith brine region, and (Q q，z ) represents the change rate of the spectral detection data between the qth time period and the zth time period. q, z, and i belong to positive integers, and q>z, i≤n.

[0086] As described in the above step S3, perform standardization processing on each of the feature vectors according to a preset method to obtain a standard data set. Among them, the standardization processing can be to reduce the dimension of the data, remove unnecessary data, etc. For example, it can adopt a sliding window method, or can be a method of reducing the dimension and transforming the data. There will be a detailed description later, so it will not be elaborated here.

[0087] As described in the above steps S4 - S6, the standard data set is divided into a training data set and a test data set according to a preset ratio. The training data set and the enrichment degrees of each target element in the brine area corresponding to the training data set are input into a preset SVM model, and the preset SVM model is trained according to the optimal hyperparameters. The trained model is detected by the test data set and the enrichment degrees of each target element in the brine area corresponding to the test data set. When the detection result meets the training requirements of the model, a target model is obtained. Among them, SVM (Support Vector Machine) is a most popular machine learning technology, which is an approximate representation of minimizing structural risk. It seeks the best combination point between the model generalization performance and the fitting performance, rather than the traditional empirical risk minimization, and cleverly solves the dimensionality problem, and can process the data after standardization. Specifically, its training method is the same as the existing method for training SVM, which will not be elaborated here. The preset SVM model is a pre - set SVM model.

[0088] As described in the above step S7, obtain the target spectral detection result of the brine area to be detected, and input the target spectral detection result into the preset target model to obtain the enrichment degree of the target element in the brine area to be detected. That is, inputting the target spectral detection result of the brine area to be detected into the preset target model can obtain the enrichment degree of the corresponding target element. It should be noted that the target spectral detection result corresponds to the spectral detection result, including spectral detection data of multiple time periods. When inputting into the preset target model, it is still necessary to calculate the corresponding feature vectors and make predictions to obtain the enrichment degree of the target element in the brine area to be detected. By collecting the spectral detection results of the brine area, the feature vectors in time sequence are obtained, and then the change law of the spectrum is extracted according to the time sequence, so as to realize the detection of the enrichment degree of the target element, and thus it can be detected whether the resource enrichment degree of the brine area meets the mining requirements, greatly reducing the risk of insufficient resources in mining.

[0089] In one embodiment, the step S3 of standardizing each of the feature vectors according to a preset method to obtain a standard data set includes:

[0090] S301: Extract the maximum value points and minimum value points in each of the feature vectors, and arrange them in time sequence to obtain an extreme value sequence;

[0091] S302: Fit the extreme value sequence using a cubic spline interpolation function to obtain the upper and lower envelope lines X max (t) and X min (t);

[0092] S303: Take the mean of the upper and lower envelope lines and denote it as the envelope line mean m(t); where,

[0093] S304: Subtract the envelope mean from the feature vector to obtain a target sequence;

[0094] S305: Determine whether the target sequence passes the intrinsic mode function test;

[0095] S306: If it passes the test, record the target sequence as a random oscillation function; otherwise, record the target sequence as the first feature vector and recalculate the target sequence until the random oscillation function is obtained;

[0096] S307: Subtract the first feature vector from the feature vector to obtain a second feature vector, and repeat to obtain multiple random oscillation functions until the calculated envelope is symmetric and the mean of the envelope is 0, thereby obtaining multiple random oscillation functions;

[0097] S308: Aggregate the random oscillation functions of each feature vector to obtain the standard data corresponding to each feature vector, and further obtain a standard data set composed of each standard data.

[0098] As described in the above steps S301 - S308, in order to extract the implicit features therein, the feature vector can be decomposed. Specifically, extract the maximum and minimum points in each feature vector and arrange them in chronological order to obtain an extreme value sequence, and use a cubic spline interpolation function to fit the extreme value sequence to obtain the upper and lower envelope lines X max (t) and X min (t), and take the mean of the upper and lower envelope lines as the envelope mean m(t); where Subtract the envelope mean from the feature vector to obtain a target sequence. At this time, if the target sequence does not pass the intrinsic mode function test, the target sequence can be recorded as the first feature vector and the target sequence can be recalculated until the random oscillation function is obtained. If it passes the test, the target sequence is recorded as the random oscillation function, and the random oscillation functions of each feature vector are aggregated to obtain the standard data corresponding to each feature vector, and further obtain a standard data set composed of each standard data. Among them, cubic spline interpolation is abbreviated as Spline interpolation, which is a smooth curve passing through a series of shape value points. Mathematically, it is a process of solving a system of three-moment equations to obtain a set of curve functions. The intrinsic mode function is a pre-set function, which can be set artificially. By decomposing the feature vector as described above, the prediction performance of the machine learning model can be improved.

[0099] In one embodiment, before the step S5 of inputting the training data set and the enrichment degrees of the respective target elements in the brine areas corresponding to the training data set into a preset SVM model and training the preset SVM model according to the optimal hyperparameters, the following steps are further included:

[0100] S401: Obtain initial SVM models with multiple different combinations of hyperparameters;

[0101] S402: Divide the training data set into a training set and a validation set according to a preset ratio;

[0102] S403: Input the training set and the enrichment degrees of the respective target elements in the corresponding brine areas into each of the initial SVM models for training to obtain corresponding multiple temporary SVM models;

[0103] S404: Verify each of the temporary SVM models through the validation set to obtain the verification results of each of the temporary SVM models;

[0104] S405: Select the temporary SVM model with the optimal verification result based on the verification results as the preset SVM model.

[0105] As described in the above steps S401 - S405, the selection of the preset SVM model is realized. The generalization ability of the SVM model depends to a great extent on the penalty coefficient, the insensitive loss parameter, and the kernel function width parameter. Among them, the penalty coefficient adjusts the confidence range and the empirical risk ratio of the learning machine by controlling the penalty degree of the out-of-bounds samples. For example, if the penalty coefficient is too large, the model training is difficult and overfitting occurs. If the penalty coefficient is too small, the model is prone to underfitting. The confidence range mainly controls the size of the regression error of the decision function. If it is too large, the number of support vectors is too small, reducing the prediction accuracy. The empirical risk ratio affects the complexity of the distribution of samples in the high-dimensional feature space. Therefore, initial SVM models with multiple different combinations of hyperparameters can be used, and the temporary SVM model with the optimal verification result is selected as the preset SVM model, thereby further improving the prediction accuracy of the model.

[0106] In one embodiment, after the step S7 of obtaining the target spectral detection result of the brine area to be detected and inputting the target spectral detection result into a preset target model to obtain the enrichment degree of the target element in the brine area to be detected, the following steps are further included:

[0107] S801: Conduct chemical detection on the water samples in the brine area to be detected for multiple time periods;

[0108] S802: Determine whether the chemical detection results of the water samples in each time period meet the preset target;

[0109] S803: If the preset target is met, it is determined that the brine area to be detected is an area to be exploited.

[0110] As described in the above steps S801 - S803, the determination of whether the brine area to be detected is an area to be exploited is realized. Chemical detection is performed on water samples in multiple time periods of the brine area to be detected. The detection method is not limited, as long as it can be used to detect the enrichment degree of target elements. It is judged whether the chemical detection results of the water samples in each time period meet the preset target; if the preset target is met, it is determined that the brine area to be detected is an area to be exploited, thereby improving the accuracy and determining the exploitation range.

[0111] In one embodiment, it is characterized in that after the step S1 of obtaining the spectral detection results of n brine areas of the salt lake, the following steps are further included:

[0112] S201: Select one spectral detection result from the spectral detection results of n brine areas as the first spectral detection result;

[0113] S202: Calculate the difference value between the first spectral detection result and other spectral detection results;

[0114] S203: Construct a difference value set with the spectral detection results whose difference values are less than or equal to the preset difference value;

[0115] S204: Define the difference value between each element in the difference value set and the first spectral detection result as the preset difference value;

[0116] S205: According to the formula Calculate the density of the first spectral detection result, where ρ c (O) represents the density of the first spectral detection result, N(O) represents the difference value set, P represents the element in the difference value set, O represents the first spectral detection result, d c (O, P) represents the difference value between element P and O, and c represents the preset difference value;

[0117] S206: According to the formula Calculate the discrete score of the first spectral detection result; where LOF c (O) represents the discrete score of the first spectral detection result, ρ c (P) represents the density of the spectral detection result corresponding to element P;

[0118] S207: Judge whether the discrete score is greater than the preset discrete score;

[0119] S208: If so, remove the first spectral detection result from the n spectral detection results.

[0120] As described in the above steps S201 - S208, the standardization process of the data is achieved. Since the individual spectral detection results may deviate too much from other spectral detection results and are quite different from most of the spectral detection results, it is necessary to eliminate these spectral detection results. Specifically, first calculate the difference score between the first spectral detection result and other spectral detection results, and obtain other spectral detection results with a difference score less than a preset difference score, where the preset difference score is a fixed value set in advance. Then calculate the density of the first spectral detection result. Calculate the discrete score based on the density. If this discrete score is large, it indicates that the difference between the first spectral detection result and other spectral detection results is large. At this time, the first spectral detection result can be eliminated. Calculate each spectral detection result as the first spectral detection result in this way and eliminate them one by one, so as to make a better prediction based on the final spectral detection results.

[0121] Refer to Figure 2 , the present invention also provides a device for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology, including:

[0122] The first acquisition module 10 is used to acquire the spectral detection results of n brine areas of the salt lake and the enrichment degree of target elements in each brine area, where each spectral detection result includes spectral detection data in multiple time periods;

[0123] The calculation module 20 is used to calculate the feature vector X of each brine area according to the spectral detection data i =

[0124] {(Q q，q-1 ), (Q q，q-2 ),..., (Q q，z ),...,(Q q，2 ), (Q q，1 )}, where q represents the q-th time period, X i represents the feature vector of the i-th brine area, and (Q q，z ) represents the change rate of spectral detection data between the q-th time period and the z-th time period. q, z, and i are positive integers, and q > z, i ≤ n;

[0125] The processing module 30 is used to perform standardization processing on each of the feature vectors according to a preset method to obtain a standard data set;

[0126] The splitting module 40 is used to divide the standard data set into a training data set and a test data set according to a preset ratio;

[0127] An input module 50 for inputting the training data set and the enrichment degrees of various target elements in the brine area corresponding to the training data set into a preset SVM model, and training the preset SVM model according to the optimal hyperparameters;

[0128] A detection module 60 for detecting the trained model by using the test data set and the enrichment degrees of various target elements in the brine area corresponding to the test data set, and obtaining a target model when the detection result meets the training requirements of the model;

[0129] A second acquisition module 70 for acquiring the target spectral detection result of the brine area to be detected, and inputting the target spectral detection result into the preset target model to obtain the enrichment degree of the target element in the brine area to be detected.

[0130] In one embodiment, the processing module 30 includes:

[0131] An extraction sub-module for extracting the maximum and minimum points in each of the feature vectors and arranging them in chronological order to obtain an extreme value sequence;

[0132] A fitting sub-module for fitting the extreme value sequence by using a cubic spline interpolation function to obtain upper and lower envelope lines X max (t) and X min (t);

[0133] A value-taking sub-module for taking the mean of the upper and lower envelope lines and denoting it as the envelope line mean m(t); where

[0134] A first calculation sub-module for subtracting the envelope line mean from the feature vector to obtain a target sequence;

[0135] A judgment sub-module for judging whether the target sequence passes the intrinsic mode function test;

[0136] A recording sub-module for, if passing the test, denoting the target sequence as a random oscillation function, otherwise, denoting the target sequence as a first feature vector and recalculating the target sequence until the random oscillation function is obtained;

[0137] A second calculation sub-module for subtracting the first feature vector from the feature vector to obtain a second feature vector, and repeating to obtain multiple random oscillation functions until the calculated envelope line is symmetric and the mean of the envelope line is 0, thereby obtaining multiple random oscillation functions;

[0138] An aggregation sub-module for aggregating the random oscillation functions of each feature vector, thereby obtaining the standard data corresponding to each feature vector, and further obtaining a standard data set composed of each standard data.

[0139] In one embodiment, the detection device for the enrichment degree of salt lake resources based on hyperspectral recognition technology further includes:

[0140] A model acquisition module, configured to acquire initial SVM models with multiple different combinations of hyperparameters;

[0141] A dataset division module, configured to divide the training dataset into a training set and a validation set according to a preset ratio;

[0142] A data input module, configured to input the training set and the enrichment degrees of each target element in the corresponding brine area into each of the initial SVM models for training, to obtain corresponding multiple temporary SVM models;

[0143] A model verification module, configured to verify each of the temporary SVM models through the validation set, to obtain the verification results of each of the temporary SVM models;

[0144] A marking module, configured to select the temporary SVM model with the optimal verification result as the preset SVM model based on the verification results.

[0145] Advantages of the present invention: By collecting the spectral detection results of the brine area, characteristic vectors in chronological order are obtained, and then the variation law of the spectrum is extracted according to the chronological order, so as to realize the detection of the enrichment degree of the target element, and thus it can be detected whether the resource enrichment degree of the brine area meets the mining requirements, greatly reducing the risk of insufficient resources in mining.

[0146] Referring to Figure 3 , in the embodiments of the present application, a computer device is further provided. The computer device may be a server, and its internal structure may be as shown in Figure 3 . The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store various spectral detection results, etc. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it can implement the method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology described in any of the above embodiments.

[0147] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.

[0148] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology described in any of the above embodiments can be implemented.

[0149] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided in the present application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0150] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, apparatus, article, or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such a process, apparatus, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, apparatus, article, or method including that element.

[0151] The embodiment of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is a theory, method, technology, and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results.

[0152] The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, robotics, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0153] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology, characterized in that, Including: S1: Obtain the spectral detection results of n brine regions in the salt lake and the enrichment degrees of target elements in each brine region; where each of the spectral detection results includes spectral detection data in multiple time periods; S2: Calculate the eigenvector X of each brine area according to the spectral detection data i ={(Q q，q-1 ), (Q q，q-2 ),..., (Q q，z ),...,(Q q，2 ), (Q q，1 )}, where q represents the q-th time period, X i represents the eigenvector of the i-th brine area, (Q q，z ) represents the change rate of the spectral detection data between the q-th time period and the z-th time period, q, z, i belong to positive integers, and q>z, i≤n; S3: Standardize each of the feature vectors according to a preset method to obtain a standard data set; S4: Divide the standard data set into a training data set and a test data set according to a preset ratio; S5: Input the training data set and the enrichment degrees of the respective target elements in the brine regions corresponding to the training data set into a preset SVM model, and train the preset SVM model according to the optimal hyperparameters; S6: Detect the trained model through the test data set and the enrichment degrees of the respective target elements in the brine regions corresponding to the test data set. When the detection result meets the training requirements of the model, obtain the target model; S7: Obtain the target spectral detection result of the brine region to be detected, input the target spectral detection result into the preset target model, and obtain the enrichment degree of the target element in the brine region to be detected; The step S3 of standardizing each of the feature vectors according to a preset method to obtain a standard data set includes: S301: Extract the maximum value points and minimum value points in each of the feature vectors, and arrange them in chronological order to obtain an extreme value sequence; S302: Fit the extreme value sequence using a cubic spline interpolation function to obtain the upper and lower envelopes X max (t) and X min (t); where t represents time; S303: Take the mean value of the upper and lower envelope lines and denote it as the envelope line mean value m(t); where, S304: Subtract the envelope line mean from the feature vector to obtain a target sequence; S305: Determine whether the target sequence passes the intrinsic mode function test; S306: If it passes the test, record the target sequence as a random oscillation function. Otherwise, record the target sequence as the first feature vector and recalculate the target sequence until the random oscillation function is obtained; S307: Subtract the first feature vector from the feature vector to obtain a second feature vector, and repeat to obtain multiple random oscillation functions until the calculated envelope line is symmetric and the mean of the envelope line is 0, thereby obtaining multiple random oscillation functions; S308: Aggregate the random oscillation functions of each feature vector to obtain the standard data corresponding to each feature vector, and further obtain a standard data set composed of each standard data.

2. The method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology according to claim 1, characterized in that, Before the step S5 of inputting the training data set and the enrichment degrees of the respective target elements in the brine regions corresponding to the training data set into a preset SVM model and training the preset SVM model according to the optimal hyperparameters, it further includes: S401: Obtain initial SVM models with multiple different hyperparameter combinations; S402: Divide the training data set into a training set and a validation set according to a preset ratio; S403: Input the training set and the enrichment degrees of the respective target elements in the corresponding brine regions into each of the initial SVM models for training to obtain corresponding multiple temporary SVM models; S404: Verify each of the temporary SVM models through the validation set to obtain the verification results of each of the temporary SVM models; S405: Based on the verification results, select the temporary SVM model with the best verification result as the preset SVM model.

3. The method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology according to claim 1, characterized in that, After the step S7 of obtaining the target spectrum detection result of the brine area to be detected, inputting the target spectrum detection result into a preset target model, and obtaining the enrichment of the target element in the brine area to be detected, the method further includes: S801: Perform chemical testing on water samples in multiple time periods in the brine area to be tested; S802: Determine whether the chemical test results of the water samples in each time period meet the preset targets; S803: If the preset target is met, the brine area to be detected is determined to be an area to be mined.

4. The method for detecting the enrichment degree of salt lake resources based on hyperspectral recognition technology according to claim 1, wherein, After the step S1 of obtaining the spectrum detection results of n brine areas of the salt lake, the method further includes: S201: Selecting a spectrum detection result from the spectrum detection results of n brine areas as a first spectrum detection result; S202: Calculating the difference between the first spectrum detection result and the other spectrum detection results; S203: constructing a difference set from the spectrum detection results whose difference values ​​are less than or equal to a preset difference; S204: defining the difference between each element in the difference set and the first spectrum detection result as the preset difference; S205: Calculate the density of the first spectral detection result according to the formula where ρ c (O) represents the density of the first spectral detection result, N(O) represents the difference set, P represents an element in the difference set, O represents the first spectral detection result, and d c (O, P) represents the difference value between element P and O, and c represents the preset difference; S206: Calculate the discrete score of the first spectral detection result according to the formula ; where LOF c (O) represents the discrete score of the first spectral detection result, ρ c (P) represents the density of the spectral detection result corresponding to element P; S207: Determine whether the discrete score is greater than a preset discrete score; S208: If yes, then remove the first spectrum detection result from the n spectrum detection results.

5. A detection device for the enrichment degree of salt lake resources based on hyperspectral recognition technology, characterized in that, include: The first acquisition module is used to obtain the spectrum detection results of n brine areas of the salt lake and the enrichment of the target elements in each brine area; Each of the spectrum detection results includes spectrum detection data within multiple time periods; A calculation module, configured to calculate a feature vector X of each brine area according to the spectral detection data i = {(Q q，q-1 ), (Q q，q-2 ),..., (Q q，z ),..., (Q q，2 ), (Q q，1 )}, where q represents the q-th time period, X i represents the feature vector of the i-th brine area, (Q q，z ) represents the change rate of the spectral detection data between the q-th time period and the z-th time period, q, z, i belong to positive integers, and q > z, i ≤ n; A processing module, used for performing standardization processing on each of the feature vectors according to a preset method to obtain a standard data set; A splitting module, used to divide the standard data set into a training data set and a test data set according to a preset ratio; An input module, used to input the training data set and the enrichment of each target element in the brine area corresponding to the training data set into a preset SVM model, and train the preset SVM model according to the optimal hyperparameters; A detection module, used to detect the trained model through the test data set and the enrichment of each target element in the brine area corresponding to the test data set, and obtain the target model when the detection result meets the training requirements of the model; The second acquisition module is used to obtain the target spectrum detection result of the brine area to be detected, and input the target spectrum detection result into a preset target model to obtain the enrichment of the target element in the brine area to be detected; The processing module comprises: An extraction submodule, used to extract the maximum and minimum points in each of the feature vectors, and arrange them in chronological order to obtain an extreme value sequence; A fitting sub-module, which is used to fit the extreme value sequence by using a cubic spline interpolation function to obtain upper and lower envelope lines X max (t) and X min (t); where t represents time; A value-taking sub-module is used to take the average value of the upper and lower envelope lines and denote it as the envelope line average value m(t); where, A first calculation submodule is used to subtract the envelope mean from the feature vector to obtain a target sequence; A judgment submodule, used to judge whether the target sequence passes the intrinsic mode function test; Recorded as a submodule, for recording the target sequence as a random oscillation function if the test is passed, otherwise, recording the target sequence as the first eigenvector and recalculating the target sequence until the random oscillation function is obtained; A second computing sub-module, configured to subtract the first feature vector from the feature vector to obtain a second feature vector, and repeat to obtain a plurality of random oscillation functions until the calculated envelope is symmetric and the mean value of the envelope is 0, thereby obtaining a plurality of random oscillation functions; An aggregation sub-module, configured to aggregate the random oscillation functions of each feature vector, thereby obtaining the standard data corresponding to each feature vector, and further obtaining a standard data set composed of each standard data.

6. The detection device for the enrichment degree of salt lake resources based on hyperspectral recognition technology according to claim 5, characterized in that, The salin lake resource enrichment degree detection device based on the hyperspectral identification technology further includes: A model acquisition module, configured to acquire initial SVM models with multiple different hyperparameter combinations; A data set division module, configured to divide the training data set into a training set and a validation set according to a preset ratio; A data input module, configured to input the training set and the enrichment degrees of each target element in the corresponding brine area into each of the initial SVM models for training to obtain corresponding temporary SVM models; A model validation module, configured to validate each of the temporary SVM models through the validation set to obtain the validation results of each of the temporary SVM models; A marking module, configured to select the temporary SVM model with the optimal validation result as the preset SVM model based on the validation results.

7. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images