A method and system for inversion of lithium concentration in salt lakes based on LightGBM

Abstract

The present invention relates to a method and system for inverting lithium concentration in a salt lake based on LightGBM. The method comprises: obtaining measured data of lithium concentration in a salt lake in a test area; obtaining spectral data of sampling points in the test area; constructing a LightGBM regression model of lithium concentration based on spectral characteristics based on the measured data of lithium concentration in the salt lake in the test area and the spectral data; obtaining spectral data of the entire salt lake range; inputting the spectral data of the entire salt lake range into the LightGBM regression model of lithium concentration based on spectral characteristics, and inverting the lithium concentration data of the salt lake in the entire salt lake range. The present invention can improve the inversion speed and accuracy of lithium concentration in a salt lake.

Description

Technical Field

[0001] The present invention relates to the research field of lithium concentration in salt lakes, and in particular to a method and system for inverting lithium concentration in salt lakes based on LightGBM. Background Art

[0002] At present, the main methods used for the inversion of lithium concentration in salt lakes are principal component analysis, correlation analysis, linear regression, machine learning, etc. The lithium concentration inversion methods based on principal component analysis, correlation analysis and linear regression were produced earlier and have been the main methods for the inversion of lithium concentration in salt lakes for a long time. In recent years, with the rise of big data, machine learning algorithms have begun to be widely used. The lithium concentration inversion based on machine learning algorithms currently only uses BP (Back Propagation) neural network and RF (Random Forest) algorithm. Summary of the invention

[0003] The purpose of the present invention is to provide a salt lake lithium concentration inversion method and system based on LightGBM to improve the inversion accuracy and speed of the salt lake lithium concentration.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A method for inverting lithium concentration in salt lakes based on LightGBM, including:

[0006] Obtain measured data on lithium concentration in salt lakes in the test area;

[0007] Acquire the spectral data of the sampling points in the test area; the spectral data of the sampling points are acquired from remote sensing images, and the remote sensing images are remote sensing images taken at the same sampling period as the measured data of the lithium concentration of the salt lake;

[0008] Based on the measured data of lithium concentration in the salt lake and the spectral data in the test area, a LightGBM regression model of lithium concentration based on spectral characteristics is constructed;

[0009] Obtain spectral data for the entire salt lake;

[0010] The spectral data of the entire salt lake range is input into the lithium concentration LightGBM regression model based on spectral characteristics to invert the salt lake lithium concentration data of the entire salt lake range.

[0011] Optionally, acquiring the spectrum data of the sampling points of the test area specifically includes:

[0012] Acquire remote sensing image data from Landsat or Sentinel-2, etc.; the remote sensing image data is remote sensing image data of the same sampling period as the measured data of lithium concentration in the salt lake;

[0013] According to the sampling point coordinates of the measured data of lithium concentration in the salt lake, pixel values ​​of all bands of each sampling point coordinate are extracted from the remote sensing image data to obtain pixel value data;

[0014] The pixel values ​​of different bands are used as different features, the pixel values ​​of all bands of the same pixel point are used as a group of spectral data, and each group of spectral data is extracted from the pixel value data to obtain the spectral data of the sampling point.

[0015] Optionally, the lithium concentration LightGBM regression model based on spectral characteristics is constructed based on the measured data of lithium concentration in the salt lake of the test area and the spectral data, specifically including:

[0016] The measured data of the salt lake lithium concentration and the spectrum data of the test area are divided into a training sample and a verification sample; the training sample includes part of the measured data of the salt lake lithium concentration and part of the spectrum data, and the verification sample also includes part of the measured data of the salt lake lithium concentration and part of the spectrum data;

[0017] Build a LightGBM regression model and initialize the parameters of the LightGBM regression model;

[0018] Using the training samples to train the LightGBM regression model;

[0019] Using the validation sample to evaluate the accuracy of the LightGBM regression model;

[0020] If the accuracy of the LightGBM regression model does not meet the accuracy standard, modify the parameters of the LightGBM regression model and retrain;

[0021] If the accuracy of the LightGBM regression model reaches the accuracy standard or the accuracy of the LightGBM regression model no longer improves, the training is stopped to obtain the trained LightGBM regression model of lithium concentration based on spectral features.

[0022] Optionally, the parameters of the LightGBM regression model include: learning_rate (learning rate), num_boost_round (number of training rounds), num_leaves (maximum number of leaf nodes in the tree), max_depth (maximum depth of the tree), min_gain_to_split (minimum split value), min_data_in_leaf (minimum amount of data on a leaf), feature_fraction (parameter ratio during iteration), bagging_fraction (data ratio during iteration), reg_alpha (L1 regularization coefficient) and reg_lambda (L2 regularization coefficient).

[0023] Optionally, the spectral data of the entire salt lake range is input into the lithium concentration LightGBM regression model based on spectral characteristics to invert the salt lake lithium concentration data of the entire salt lake range, specifically including:

[0024] Inputting the spectral data into the LightGBM regression model of lithium concentration based on spectral characteristics to invert the lithium concentration value of each pixel point;

[0025] The obtained lithium concentration values ​​are rearranged according to the arrangement of the spectral data in the remote sensing image to obtain the inversion result of the salt lake lithium concentration data of the entire salt lake range.

[0026] The present invention also provides a salt lake lithium concentration inversion system based on LightGBM, comprising:

[0027] A module for acquiring measured data of lithium concentration in salt lakes in the test area, used to acquire measured data of lithium concentration in salt lakes in the test area;

[0028] A spectral data acquisition module for sampling points in the test area, used to acquire spectral data of the sampling points in the test area; the spectral data of the sampling points are acquired from remote sensing images, and the remote sensing images are remote sensing images of the same sampling period as the measured data of lithium concentration in the salt lake;

[0029] A module for constructing a lithium concentration LightGBM regression model based on spectral features, which is used to construct a lithium concentration LightGBM regression model based on spectral features based on the measured data of lithium concentration in salt lakes in the test area and spectral data;

[0030] The spectrum data acquisition module of the entire salt lake is used to acquire the spectrum data of the entire salt lake;

[0031] The inversion module is used to input the spectral data of the entire salt lake range into the lithium concentration LightGBM regression model based on spectral characteristics to invert the salt lake lithium concentration data of the entire salt lake range.

[0032] Optionally, the spectrum data acquisition module of the sampling points in the test area specifically includes:

[0033] A remote sensing image data acquisition unit, used to acquire remote sensing image data from Landsat or Sentinel-2, etc.; the remote sensing image data is remote sensing image data of the same sampling period as the measured data of lithium concentration in the salt lake;

[0034] A pixel value extraction unit is used to extract the pixel values ​​of all bands of each sampling point coordinate from the remote sensing image data according to the sampling point coordinates of the measured data of the lithium concentration of the salt lake, so as to obtain pixel value data;

[0035] The spectral data extraction unit is used to take the pixel values ​​of different bands as different features, take the pixel values ​​of all bands of the same pixel point as a group of spectral data, extract each group of spectral data from the pixel value data, and obtain the spectral data of the sampling point.

[0036] Optionally, the lithium concentration LightGBM regression model construction module based on spectral characteristics specifically includes:

[0037] A sample generating unit is used to divide the measured data of the salt lake lithium concentration and the spectrum data of the test area into a training sample and a verification sample; the training sample includes part of the measured data of the salt lake lithium concentration and part of the spectrum data, and the verification sample also includes part of the measured data of the salt lake lithium concentration and part of the spectrum data;

[0038] LightGBM regression model construction unit, used to build the LightGBM regression model and initialize the parameters of the LightGBM regression model;

[0039] A training unit, used to train the LightGBM regression model using the training samples;

[0040] An accuracy evaluation unit, used to evaluate the accuracy of the LightGBM regression model using the verification sample;

[0041] A parameter updating unit, used for modifying the parameters of the LightGBM regression model and retraining the model when the accuracy of the LightGBM regression model does not meet the accuracy standard;

[0042] The iteration stop unit is used to stop training when the accuracy of the LightGBM regression model reaches the accuracy standard or the accuracy of the LightGBM regression model is no longer improved, so as to obtain the trained LightGBM regression model of lithium concentration based on spectral features.

[0043] Optionally, the parameters of the LightGBM regression model include: learning_rate (learning rate), num_boost_round (number of training rounds), num_leaves (maximum number of leaf nodes in the tree), max_depth (maximum depth of the tree), min_gain_to_split (minimum split value), min_data_in_leaf (minimum amount of data on a leaf), feature_fraction (parameter ratio during iteration), bagging_fraction (data ratio during iteration), reg_alpha (L1 regularization coefficient) and reg_lambda (L2 regularization coefficient).

[0044] Optionally, the inversion module specifically includes:

[0045] A lithium concentration value inversion unit at a pixel point, used for inputting the spectral data into the lithium concentration LightGBM regression model based on spectral features to invert the lithium concentration value of each pixel point;

[0046] The arrangement unit is used to rearrange the obtained lithium concentration values ​​according to the arrangement mode of the spectral data in the remote sensing image, so as to obtain the inversion result of the salt lake lithium concentration data of the entire salt lake range.

[0047] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0048] The present invention takes the remote sensing spectral data of salt lakes as the research object, uses the fast and accurate regression prediction capability of LightGBM to find the complex mapping relationship between each band of remote sensing images and the lithium concentration of salt lakes, and performs remote sensing inversion of lithium concentration of salt lakes, thereby providing fast and effective technical support for the inversion of lithium concentration of salt lakes, thereby improving the accuracy and speed of inversion of lithium concentration of salt lakes, and providing technical inspiration for other types of remote sensing inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 It is a schematic diagram of the process of the salt lake lithium concentration inversion method based on LightGBM of the present invention;

[0051] Figure 2A schematic diagram of the process of constructing a LightGBM regression model of lithium concentration based on spectral characteristics in the present invention;

[0052] Figure 3 It is a schematic diagram of the inversion results of the salt lake lithium concentration data of the entire salt lake range of the present invention;

[0053] Figure 4 It is a structural schematic diagram of the salt lake lithium concentration inversion system based on LightGBM of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] As a GBDT (GradientBoosting Decision Tree) algorithm framework, LightGBM has higher accuracy and smaller memory consumption compared with traditional GBDT, RF, XGBoost (eXtreme Gradient Boosting), and SVM (Support Vector Machines). Therefore, the present invention applies LightGBM to the inversion of lithium concentration in salt lakes based on spectral features, which can reduce memory consumption, increase inversion speed, and improve inversion accuracy.

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Figure 1 FIG. 1 is a flow chart of the method for inverting lithium concentration in salt lakes based on LightGBM of the present invention. Figure 1 As shown, the salt lake lithium concentration inversion method based on LightGBM of the present invention includes the following steps:

[0058] Step 100: Obtaining the measured data of lithium concentration in the salt lake in the test area. In this step, the test area is a part of the entire salt lake, and the measured data of lithium concentration in the salt lake is the accurate data actually measured.

[0059] Step 200: Obtain the spectral data of the sampling points in the test area. The spectral data of the sampling points are obtained from remote sensing images, which are remote sensing images of the same sampling period as the measured data of the lithium concentration of the salt lake, that is, the sampling time is roughly the same, and the time difference does not exceed 10 days; the image band range includes visible light to near infrared; the image has been previously processed by radiation calibration and atmospheric correction. The remote sensing image data comes from Landsat or Sentinel-2, etc. First, according to the longitude and latitude of the sampling points in the measured data of the lithium concentration of the salt lake, all the band pixel values ​​of the same sampling coordinate point are extracted from the remote sensing image; then the pixel value of each band of the remote sensing image is used as a feature, and the pixel values ​​of all bands of a pixel point are used as a set of spectral data. In the subsequent training of the lithium concentration LightGBM regression model based on spectral features and the inversion of the model, all the bands of the remote sensing image are input, and there is no need to perform spectral analysis in advance.

[0060] For example, taking Landsat-8 data as an example, the remote sensing image has been processed by radiometric calibration, atmospheric correction and cropping. Radiometric calibration, atmospheric correction and cropping are all completed in ENVI, and the lake surface vector is used as a mask for cropping. After completing the image preprocessing, the sampling point vector data is generated according to the latitude and longitude of the salt lake sampling point. The multi-value extraction function of ArcGIS is used to extract the spectral values ​​of each band of the remote sensing image of the same period into the sampling point vector. Finally, the vector attributes are exported to the Excel table. The generated spectral data example is shown in Table 1.

[0061] Table 1 Spectral characteristic data of salt lake sampling points

[0062]

[0063]

[0064]

[0065] Step 300: Construct a LightGBM regression model of lithium concentration based on spectral characteristics based on the measured data of lithium concentration in salt lakes in the test area and the spectral data. Figure 2 The schematic diagram of the process of constructing the LightGBM regression model of lithium concentration based on spectral characteristics in the present invention is as follows: Figure 2 As shown, the following steps are included:

[0066] Step 1: Divide the salt lake lithium concentration data and spectrum data into training samples and verification samples. For example, taking the data in Table 1 as an example, divide the 27 sample data into 24 training samples and 3 verification samples in a ratio of 9:1. The training samples are used to train the LightGBM model, and the verification samples are used to verify the accuracy of the LightGBM model, so as to obtain the optimal LightGBM model.

[0067] Step 2: Build a LightGBM regression model and initialize the parameters of the LightGBM regression model. The construction of the LightGBM regression model is completed in Python. The parameters of the LightGBM regression model include learning_rate (learning rate), num_boost_round (number of training rounds), num_leaves (the maximum number of leaf nodes in the tree), max_depth (the maximum depth of the tree), min_gain_to_split (minimum split value), min_data_in_leaf (minimum amount of data on a leaf), feature_fraction (parameter ratio during iteration), bagging_fraction (data ratio during iteration), reg_alpha (L1 regularization coefficient) and reg_lambda (L2 regularization coefficient). During initialization, you can set learning_rate to 0.5, num_boost_round to 50, num_leaves to 20, max_depth to 4, min_gain_to_split to 0.01, min_data_in_leaf to 16, feature_fraction to 1, bagging_fraction to 1, reg_alpha to 0, and reg_lambda to 0.

[0068] Step 3: Use training samples to train the LightGBM regression model.

[0069] Step 4: Use the validation samples to evaluate the accuracy of the training model. For example, RMSE and R 2 Conduct accuracy evaluation.

[0070] Step 5: If the accuracy standard is not met, modify the parameters and retrain. The modification order is: modify learning_rate and num_boost_round, num_leaves and max_depth, min_gain_to_split and min_data_in_leaf, feature_fraction and bagging_fraction, and reg_alpha and reg_lambda in sequence.

[0071] Step 6: Repeated training until the accuracy reaches the standard or cannot be improved, and a trained LightGBM regression model of lithium concentration based on spectral features is obtained. Taking the data set in Table 1 as an example and the above initialization process, the parameters of the final LightGBM regression model of lithium concentration based on spectral features are: learning_rate is set to 0.15, num_boost_round is set to 90, num_leaves is set to 20, max_depth is set to 3, min_gain_to_split is set to 0.01, min_data_in_leaf is set to 10, feature_fraction is set to 1, bagging_fraction is set to 1, reg_alpha is set to 0, and reg_lambda is set to 0. The actual and predicted values ​​of the validation set are: 496.70715, 565.421820, 537.202686; 494.88453, 551.95363659, 525.68021378. The accuracy verification results are: RMSE: 10.2869; R 2 :0.867.

[0072] Step 400: Obtain spectral data of the entire salt lake. The present invention uses Python GDAL library to read remote sensing images and organize the remote sensing data into training data form (the pixel value of each band of the remote sensing image is used as a feature, and the pixel values ​​of all bands of a pixel point are used as a set of spectral data).

[0073] Step 500: Input the spectral data of the entire salt lake into the LightGBM regression model of lithium concentration based on spectral features to invert the lithium concentration data of the entire salt lake. Input the spectral data of the entire salt lake into the trained LightGBM regression model of lithium concentration based on spectral features to invert the lithium concentration value of each pixel point, and then rearrange the obtained lithium concentration values ​​according to the arrangement of the spectral feature data in the remote sensing image to obtain the inversion result of the lithium concentration data of the entire salt lake. An example of the inversion result is as follows: Figure 3 shown.

[0074] Based on the above scheme, the present invention also provides a salt lake lithium concentration inversion system based on LightGBM. Figure 4 FIG. 1 is a schematic diagram of the structure of the salt lake lithium concentration inversion system based on LightGBM of the present invention. Figure 4 As shown, the salt lake lithium concentration inversion system based on LightGBM of the present invention includes:

[0075] The test area salt lake lithium concentration measured data acquisition module 401 is used to obtain the test area salt lake lithium concentration measured data.

[0076] The spectral data acquisition module 402 of the sampling points in the test area is used to acquire the spectral data of the sampling points in the test area; the spectral data of the sampling points are acquired from remote sensing images, and the remote sensing images are remote sensing images of the same sampling period as the measured data of the lithium concentration of the salt lake.

[0077] The module 403 for constructing a LightGBM regression model of lithium concentration based on spectral features is used to construct a LightGBM regression model of lithium concentration based on spectral features based on the measured data of lithium concentration in salt lakes in the test area and the spectral data.

[0078] The spectral data acquisition module 404 for the entire salt lake is used to acquire the spectral data for the entire salt lake.

[0079] The inversion module 405 is used to input the spectral data of the entire salt lake range into the lithium concentration LightGBM regression model based on spectral characteristics to invert the salt lake lithium concentration data of the entire salt lake range.

[0080] As a specific embodiment, in the salt lake lithium concentration inversion system based on LightGBM of the present invention, the spectrum data acquisition module 402 of the sampling point in the test area specifically includes:

[0081] The remote sensing image data acquisition unit is used to acquire remote sensing image data from Landsat or Sentinel-2; the remote sensing image data is remote sensing image data of the same sampling period as the measured data of lithium concentration in the salt lake.

[0082] The pixel value extraction unit is used to extract the pixel values ​​of all bands of each sampling point coordinate from the remote sensing image data according to the sampling point coordinates of the measured data of the lithium concentration of the salt lake, so as to obtain pixel value data.

[0083] The spectral data extraction unit is used to take the pixel values ​​of different bands as different features, take the pixel values ​​of all bands of the same pixel point as a group of spectral data, extract each group of spectral data from the pixel value data, and obtain the spectral data of the sampling point.

[0084] As a specific embodiment, in the salt lake lithium concentration inversion system based on LightGBM of the present invention, the lithium concentration LightGBM regression model construction module 403 based on spectral characteristics specifically includes:

[0085] The sample generation unit is used to divide the measured data of the salt lake lithium concentration and the spectral data of the test area into training samples and verification samples; the training samples include part of the measured data of the salt lake lithium concentration and part of the spectral data, and the verification samples also include part of the measured data of the salt lake lithium concentration and part of the spectral data.

[0086] The LightGBM regression model construction unit is used to build the LightGBM regression model and initialize the parameters of the LightGBM regression model.

[0087] A training unit is used to train the LightGBM regression model using the training samples.

[0088] The accuracy evaluation unit is used to evaluate the accuracy of the LightGBM regression model using the verification sample.

[0089] A parameter updating unit is used to modify the parameters of the LightGBM regression model and retrain it when the accuracy of the LightGBM regression model does not meet the accuracy standard.

[0090] The iteration stop unit is used to stop training when the accuracy of the LightGBM regression model reaches the accuracy standard or the accuracy of the LightGBM regression model is no longer improved, so as to obtain the trained LightGBM regression model of lithium concentration based on spectral features.

[0091] As a specific embodiment, in the salt lake lithium concentration inversion system based on LightGBM of the present invention, the inversion module 405 specifically includes:

[0092] The lithium concentration value inversion unit of the pixel point is used to input the spectral data into the lithium concentration LightGBM regression model based on spectral characteristics to invert the lithium concentration value of each pixel point.

[0093] The arrangement unit is used to rearrange the obtained lithium concentration values ​​according to the arrangement mode of the spectral data in the remote sensing image, so as to obtain the inversion result of the salt lake lithium concentration data of the entire salt lake range.

[0094] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0095] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for inverting lithium concentration in salt lakes based on LightGBM, characterized in that: include: Obtain measured data on lithium concentration in salt lakes in the test area; Acquire spectral data of sampling points in the test area; The spectral data of the sampling point is obtained from a remote sensing image, and the remote sensing image is a remote sensing image of the same sampling period as the measured data of the lithium concentration of the salt lake, specifically including: obtaining remote sensing image data from Landsat or Sentinel-2; the remote sensing image data is a remote sensing image data of the same sampling period as the measured data of the lithium concentration of the salt lake; according to the sampling point coordinates of the measured data of the lithium concentration of the salt lake, extracting the pixel values ​​of all bands of each sampling point coordinate from the remote sensing image data to obtain pixel value data; taking the pixel values ​​of different bands as different features, taking the pixel values ​​of all bands of the same pixel point as a group of spectral data, extracting each group of spectral data from the pixel value data to obtain the spectral data of the sampling point; Based on the measured data of lithium concentration in the salt lake and the spectral data in the test area, a LightGBM regression model of lithium concentration based on spectral characteristics is constructed; Obtain spectral data for the entire salt lake; Input the spectral data of the entire salt lake range into the LightGBM regression model of lithium concentration based on spectral characteristics, and invert the salt lake lithium concentration data of the entire salt lake range, specifically including: input the spectral data into the LightGBM regression model of lithium concentration based on spectral characteristics, and invert the lithium concentration value of each pixel point; rearrange the obtained lithium concentration values ​​according to the arrangement of the spectral data in the remote sensing image, and obtain the inversion result of the salt lake lithium concentration data of the entire salt lake range.

2. The method for inversion of salt lake lithium concentration based on LightGBM according to claim 1, characterized in that: The lithium concentration LightGBM regression model based on the measured data of the salt lake lithium concentration in the test area and the spectral data is constructed based on the spectral characteristics, specifically including: The measured data of the salt lake lithium concentration and the spectrum data of the test area are divided into a training sample and a verification sample; the training sample includes part of the measured data of the salt lake lithium concentration and part of the spectrum data, and the verification sample also includes part of the measured data of the salt lake lithium concentration and part of the spectrum data; Build a LightGBM regression model and initialize the parameters of the LightGBM regression model; Using the training samples to train the LightGBM regression model; Using the validation sample to evaluate the accuracy of the LightGBM regression model; If the accuracy of the LightGBM regression model does not meet the accuracy standard, modify the parameters of the LightGBM regression model and retrain the LightGBM regression model; If the accuracy of the LightGBM regression model reaches the accuracy standard or the accuracy of the LightGBM regression model no longer improves, the training is stopped to obtain the trained LightGBM regression model of lithium concentration based on spectral features.

3. The method for inversion of salt lake lithium concentration based on LightGBM according to claim 2, characterized in that: The parameters of the LightGBM regression model include: learning rate, number of training rounds, maximum number of leaf nodes in the tree, maximum depth of the tree, minimum split value, minimum amount of data on a leaf, parameter ratio during iteration, data ratio during iteration, L1 regularization coefficient and L2 regularization coefficient.

4. A salt lake lithium concentration inversion system based on LightGBM, characterized in that: include: A module for acquiring measured data of lithium concentration in salt lakes in the test area, used to acquire measured data of lithium concentration in salt lakes in the test area; The spectrum data acquisition module of the sampling points in the test area is used to acquire the spectrum data of the sampling points in the test area; the spectrum data of the sampling points are acquired from remote sensing images, and the remote sensing images are remote sensing images of the same sampling period as the measured data of the lithium concentration of the salt lake, specifically including: A remote sensing image data acquisition unit, used to acquire remote sensing image data from Landsat or Sentinel-2; the remote sensing image data is remote sensing image data of the same sampling period as the measured data of lithium concentration in the salt lake; A pixel value extraction unit is used to extract the pixel values ​​of all bands of each sampling point coordinate from the remote sensing image data according to the sampling point coordinates of the measured data of the lithium concentration of the salt lake, so as to obtain pixel value data; A spectral data extraction unit, used to take pixel values ​​of different bands as different features, take pixel values ​​of all bands of the same pixel point as a group of spectral data, extract each group of spectral data from the pixel value data, and obtain the spectral data of the sampling point; A module for constructing a lithium concentration LightGBM regression model based on spectral features, which is used to construct a lithium concentration LightGBM regression model based on spectral features based on the measured data of lithium concentration in salt lakes in the test area and spectral data; The spectrum data acquisition module of the entire salt lake is used to acquire the spectrum data of the entire salt lake; The inversion module is used to input the spectral data of the entire salt lake range into the lithium concentration LightGBM regression model based on spectral characteristics to invert the salt lake lithium concentration data of the entire salt lake range, specifically including: A lithium concentration value inversion unit at a pixel point, used for inputting the spectral data into the lithium concentration LightGBM regression model based on spectral features to invert the lithium concentration value of each pixel point; The arrangement unit is used to rearrange the obtained lithium concentration values ​​according to the arrangement mode of the spectral data in the remote sensing image, so as to obtain the inversion result of the salt lake lithium concentration data of the entire salt lake range.

5. The salt lake lithium concentration inversion system based on LightGBM according to claim 4 is characterized in that: The lithium concentration LightGBM regression model construction module based on spectral characteristics specifically includes: A sample generating unit is used to divide the measured data of the salt lake lithium concentration and the spectrum data of the test area into a training sample and a verification sample; the training sample includes part of the measured data of the salt lake lithium concentration and part of the spectrum data, and the verification sample also includes part of the measured data of the salt lake lithium concentration and part of the spectrum data; LightGBM regression model construction unit, used to build the LightGBM regression model and initialize the parameters of the LightGBM regression model; A training unit, used to train the LightGBM regression model using the training samples; An accuracy evaluation unit, used to evaluate the accuracy of the LightGBM regression model using the verification sample; A parameter updating unit, used for modifying the parameters of the LightGBM regression model and retraining the model when the accuracy of the LightGBM regression model does not meet the accuracy standard; The iteration stop unit is used to stop training when the accuracy of the LightGBM regression model reaches the accuracy standard or the accuracy of the LightGBM regression model is no longer improved, so as to obtain the trained LightGBM regression model of lithium concentration based on spectral features.

6. The salt lake lithium concentration inversion system based on LightGBM according to claim 5 is characterized in that: The parameters of the LightGBM regression model include: learning rate, number of training rounds, maximum number of leaf nodes in the tree, maximum depth of the tree, minimum split value, minimum amount of data on a leaf, parameter ratio during iteration, data ratio during iteration, L1 regularization coefficient and L2 regularization coefficient.

Images