Soil leaching solution detection method and device based on multi-modal spectral feature fusion

Abstract

The invention provides a soil leaching solution detection method and device based on multi-modal spectral feature fusion, and relates to the technical field of soil detection.The method comprises the steps that in response to a concentration detection request, a mixed light beam is utilized to irradiate a soil leaching solution to obtain multi-modal spectral data, the multi-modal spectral data comprises a plurality of spectral sub-data belonging to a plurality of modals; performing spectral feature extraction on the multiple pieces of spectral sub-data to obtain multiple first spectral features; performing feature fusion on the plurality of first spectral features to obtain a fused feature; concentration detection is carried out based on the fusion features, the nitrogen concentration of the soil leaching solution is obtained, and the detection precision of the nitrogen concentration of the soil leaching solution can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of soil detection, and more specifically, to a soil leachate detection method and device based on multimodal spectral feature fusion. Background Art

[0002] Nitrogen is an important nutrient for crop growth, but excessive nitrogen application will cause nitrogen to enter groundwater or surface water through leaching, causing environmental problems such as eutrophication of water bodies. Soil nitrogen leaching solution detection is an important part of agricultural environmental detection and non-point source pollution prevention and control. The core goal of the detection is to quantify nitrate nitrogen ( ) migration flux, providing a basis for precision agriculture and environmental protection.

[0003] At present, there are at least the following problems in the relevant technologies for soil detection: Due to the complexity and dynamics of the soil-water system, the detection process faces technical challenges such as sampling errors, insufficient detection sensitivity, and poor real-time performance. In addition, the spectral method used in the relevant technology to detect the nitrogen concentration in the soil leachate has low detection accuracy, which directly affects the accuracy and reliability of the detection data, and thus affects the scientific nature of the nitrogen loss assessment and the targeted nature of pollution prevention and control measures. Summary of the invention

[0004] In view of this, the present invention provides a soil leachate detection method and device based on multimodal spectral feature fusion, which can improve detection accuracy.

[0005] One aspect of the present invention provides a soil leaching solution detection method based on multimodal spectral feature fusion, comprising: in response to a concentration detection request, irradiating the soil leaching solution with a mixed light beam to obtain multimodal spectral data, wherein the multimodal spectral data includes multiple spectral sub-data belonging to multiple modalities respectively; performing spectral feature extraction on the multiple spectral sub-data to obtain multiple first spectral features; performing feature fusion on the multiple first spectral features to obtain fused features; and performing concentration detection based on the fused features to obtain the nitrogen concentration of the soil leaching solution.

[0006] According to an embodiment of the present invention, the above-mentioned spectral feature extraction of the above-mentioned multiple spectral sub-data to obtain multiple first spectral features includes: performing feature extraction based on the above-mentioned multiple spectral sub-data respectively to obtain multiple second spectral features; and aligning and matching the above-mentioned multiple second spectral features to obtain the above-mentioned multiple first spectral features.

[0007] According to an embodiment of the present invention, the above-mentioned multiple modes include a first mode and multiple second modes; wherein, the above-mentioned alignment and matching of the above-mentioned multiple second spectral features to obtain the above-mentioned multiple first spectral features includes: using the second spectral features of the above-mentioned first mode as a reference matrix, aligning the second spectral features of each of the above-mentioned multiple second modes based on the above-mentioned reference matrix, to obtain the third spectral features of each of the above-mentioned multiple second modes; and using an iterative nearest point algorithm to optimize and align the third spectral features of each of the above-mentioned multiple second modes with the second spectral features of the above-mentioned first mode, to obtain the first spectral features of each of the above-mentioned multiple second modes, wherein the above-mentioned multiple first spectral features include the second spectral features of the above-mentioned first mode and the first spectral features of each of the above-mentioned multiple second modes.

[0008] According to an embodiment of the present invention, the above-mentioned use of an iterative nearest point algorithm to optimize and align the third spectral features of each of the multiple second modalities with the second spectral features of the first modalities to obtain the first spectral features of each of the multiple second modalities includes: using the above-mentioned iterative nearest point algorithm to calculate the matching error based on the third spectral features of each of the multiple second modalities and the second spectral features of the first modalities; updating the third spectral features based on the matching error to obtain the updated third spectral features of each of the multiple second modalities; and when the matching error is less than or equal to an error threshold, determining the updated third spectral features of each of the multiple second modalities as the first spectral features of each of the multiple second modalities.

[0009] According to an embodiment of the present invention, the method further includes: when the matching error is greater than the error threshold, optimizing the registration based on the updated third spectral features of each of the multiple second modalities and the second spectral features of the first modality.

[0010] According to an embodiment of the present invention, the above-mentioned feature extraction is performed based on the above-mentioned multiple spectral sub-data to obtain multiple second spectral features, including: for each spectral sub-data, the above-mentioned spectral sub-data is decentralized to obtain pre-processed spectral sub-data; principal component analysis is performed on the above-mentioned pre-processed spectral sub-data to obtain a principal component matrix; and the above-mentioned second spectral features are extracted from the above-mentioned principal component matrix.

[0011] According to an embodiment of the present invention, the above-mentioned use of a mixed light beam to irradiate the soil leaching solution to obtain multimodal spectral data includes: using the above-mentioned mixed light beam to irradiate the above-mentioned soil leaching solution to obtain a mixed spectrum, wherein the above-mentioned mixed spectrum includes multiple initial spectra respectively belonging to the above-mentioned multiple modes; performing spatial light modulation on the above-mentioned multiple initial spectra to obtain multiple target spectra; and using a detector to perform photoelectric detection on the above-mentioned multiple target spectra to obtain the above-mentioned multimodal spectral data.

[0012] Another aspect of the present invention provides a soil leaching solution detection device based on multimodal spectral feature fusion, comprising: a fluorescent excitation light source for exciting a first light beam; a broadband light source for exciting a second light beam; a beam combiner for combining the first light beam and the second light beam to obtain a mixed light beam, and outputting the mixed light beam to the soil leaching solution; a detector for photoelectrically detecting the spectrum generated by the soil leaching solution due to the action of the mixed light beam to obtain multimodal spectral data, wherein the multimodal spectral data includes multiple spectral sub-data belonging to multiple modes respectively; and a processor for performing spectral feature extraction on the multiple spectral sub-data to obtain multiple first spectral features, performing feature fusion on the multiple first spectral features to obtain fused features, and performing concentration detection based on the fused features to obtain the nitrogen concentration of the soil leaching solution.

[0013] According to an embodiment of the present invention, the soil leaching solution is used to output a mixed spectrum based on the mixed light beam; the device also includes: a light modulation module, used to receive the mixed spectrum, perform spatial light modulation on the mixed spectrum to obtain multiple target spectra, and project the multiple target spectra to a detector; wherein the detector is also used to perform photoelectric detection on the multiple target spectra to obtain the multimodal spectral data.

[0014] According to an embodiment of the present invention, the above-mentioned light modulation module includes: a grating, used to perform spectroscopic processing on the above-mentioned mixed spectrum to obtain multiple initial spectra belonging to the above-mentioned multiple modes respectively; a first lens, used to converge the above-mentioned multiple initial spectra to a spatial light modulator; the above-mentioned spatial light modulator is used to perform spectral spatial distribution modulation on the above-mentioned multiple initial spectra respectively to obtain the above-mentioned multiple target spectra; and a second lens, used to project the above-mentioned multiple target spectra respectively to the multiple arrays of the above-mentioned detector.

[0015] According to an embodiment of the present invention, multimodal spectral technology is used to obtain multimodal spectral data. Based on the different detection principles of multimodal spectral technology, the detection information of soil leaching solution can be enriched. Spectral feature extraction of multiple spectral sub-data can make up for the information differences between multiple modalities. Feature fusion of the first spectral features corresponding to multiple modalities can reduce the interference of other components in the soil leaching solution, improve the detection accuracy of the nitrogen concentration of the soil leaching solution, and improve the accuracy and reliability of nitrogen detection data, so that the nitrogen loss status of the soil can be more accurately evaluated, crop yield and quality can be improved, and targeted pollution prevention and control measures can be taken to reduce environmental governance costs and protect aquatic ecosystems. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0017] Figure 1 The operation flow chart of the soil leachate detection method based on multimodal spectral feature fusion according to an embodiment of the present invention is shown;

[0018] Figure 2 A schematic diagram of a soil leachate detection method based on multimodal spectral feature fusion according to an embodiment of the present invention is shown;

[0019] Figure 3 A schematic diagram of a soil leaching solution detection device based on multimodal spectral feature fusion according to an embodiment of the present invention is shown;

[0020] Figure 4 A schematic diagram of a soil leaching solution detection device based on multimodal spectral feature fusion according to another embodiment of the present invention is shown;

[0021] Figure 5 A schematic diagram of a soil leaching solution detection device based on multimodal spectral feature fusion according to another embodiment of the present invention is shown;

[0022] Figure 6 A schematic diagram of a soil leachate detection device based on multimodal spectral feature fusion according to yet another embodiment of the present invention is shown. DETAILED DESCRIPTION

[0023] Below, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of concepts of the present invention.

[0024] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0025] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0026] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0027] Nitrogen in soil is mainly in the form of ammonium nitrogen ( ) and nitrate nitrogen ( ) exists in two forms. Ammonium nitrogen has a positive charge and is easily adsorbed by soil colloids, making it relatively stable; while nitrate nitrogen has a negative charge and is not easily adsorbed by soil, so it easily moves with water and enters groundwater or surface water through leaching. Detecting the migration flux of nitrate nitrogen is the key to assessing the risk of nitrogen loss. By understanding the migration path and rate of nitrate nitrogen in the soil, the possibility and extent of its entry into groundwater or surface water can be predicted. Quantifying the migration flux of nitrate nitrogen can provide a basis for precision agriculture and environmental protection.

[0028] Many factors need to be considered in the sampling of soil nitrogen leaching solutions. There is spatial variability in soil nitrogen leaching solutions. The soil itself has a high degree of spatial heterogeneity, and the unevenness of farmland management measures (such as fertilization and irrigation) leads to significant differences in the intensity of nitrogen leaching solutions in different locations. Soil nitrogen leaching solutions are also temporally dynamic. In the short term after rainfall or irrigation events, the nitrogen concentration in the soil nitrogen leaching solution often increases in a pulsed manner, and the increase may last for hours to days. If the sampling frequency is insufficient and the nitrogen change period is missed, errors will occur in the assessment of nitrogen loss. In addition, the sampling depth will also affect the assessment results. Different crops have different root distribution depths, and soil nitrogen leaching solutions mainly occur in areas below the root layer. It is necessary to scientifically design the sampling depth in combination with specific crops and soil conditions to have ecological significance.

[0029] Detection errors can also occur during the sample storage and pretreatment of soil nitrogen leaching solutions. The nitrogen forms (especially nitrate nitrogen) in soil nitrogen leaching solution samples may continue to transform after sampling. Microbial activity can cause nitrate nitrogen to be reduced to nitrite nitrogen or even nitrogen gas. At the same time, light and temperature changes can also accelerate the transformation. Suspended particles and colloidal substances in soil nitrogen leaching solution samples may interfere with subsequent analysis and need to be removed by filtration or centrifugation, but the filter material will affect the evaluation results, especially the evaluation of dissolved organic nitrogen.

[0030] Different analytical instruments have different detection sensitivities for the migration flux of nitrate nitrogen. The concentration of nitrate nitrogen in groundwater is usually low, and the test paper method or reflectance photometry method cannot meet the requirements of detection sensitivity. Although ion chromatographs can measure multiple ions at the same time, the analysis time for each sample is long, which makes it difficult to meet the needs of rapid detection of large quantities of samples.

[0031] The spectral method is based on the characteristic absorption of nitrate nitrogen in the ultraviolet and near-infrared bands, and uses a chemometric algorithm for qualitative and quantitative analysis. It has the advantages of being fast and non-destructive, and can also detect multiple components at the same time, making it suitable for online detection. Since the composition of soil nitrogen leaching solution is complex, in addition to nitrate nitrogen, it also contains a large number of interfering substances.

[0032] By using near-infrared light to measure the absorption characteristics of soil leaching solutions to near-infrared light and analyzing spectral data, rapid detection of components in soil leaching solutions can be achieved. The near-infrared spectrum range is usually between 700nm and 2500nm. The spectral information in this band is related to the chemical bond vibrations of organic and inorganic substances in the soil. Nitrogen (for example, nitrate nitrogen) in soil leaching solutions has characteristic absorption peaks in the near-infrared spectrum. By analyzing the intensity and position of these absorption peaks, qualitative and quantitative analysis of nitrogen can be achieved. However, within the near-infrared band, although the spectral absorption of organic matter is low, the absorption of water molecules is high, which can easily cause the absorption characteristics of the target detection substance nitrate nitrogen to be covered by strong absorption peaks, resulting in a decrease in the quantitative prediction accuracy of the model.

[0033] When ultraviolet-visible light (wavelength range is about 190nm to 780nm) passes through the soil leaching solution, the molecules or ions in the soil leaching solution will absorb light of specific wavelengths. The absorption intensity is proportional to the concentration of the substance in the soil leaching solution, which conforms to the Beer-Lambert law. By measuring the absorbance, the concentration of the substance in the soil leaching solution can be calculated. However, dissolved organic matter has strong absorption at 200nm-220nm, which will seriously interfere with the determination of nitrate nitrogen. Although correction can be made by deducting the background value at 275nm, the influence of organic matter cannot be completely eliminated for samples with high organic matter content. In addition, chloride ion is another important interferent. High concentration chloride ion peaks may mask adjacent nitrate nitrogen peaks, affecting the accuracy of integration.

[0034] Therefore, the detection accuracy of nitrogen concentration in soil leaching solution is low when using a single-mode spectrum.

[0035] The embodiments of the present invention provide a soil leachate detection method and device based on multimodal spectral feature fusion, in order to solve the above technical problems.

[0036] Figure 1The operational flow chart of the soil leachate detection method based on multimodal spectral feature fusion according to an embodiment of the present invention is shown.

[0037] like Figure 1 As shown, the soil leachate detection method based on multimodal spectral feature fusion includes operations S110 to S140.

[0038] In operation S110 , in response to a concentration detection request, the soil leaching solution is irradiated with a mixed light beam to acquire multimodal spectral data.

[0039] According to an embodiment of the present invention, the multimodal spectral data includes a plurality of spectral sub-data respectively belonging to a plurality of modalities.

[0040] In operation S120, spectral features are extracted from the plurality of spectral sub-data to obtain a plurality of first spectral features.

[0041] In operation S130, feature fusion is performed on the plurality of first spectral features to obtain a fused feature.

[0042] In operation S140, concentration detection is performed based on the fused features to obtain the nitrogen concentration of the soil leaching solution.

[0043] In one example, a partial least squares regression method can be used to establish a prediction model, and soil leaching solution samples with known nitrogen concentrations are used as training sets to obtain regression parameters. Based on the fused features and regression parameters, the nitrogen concentration of the soil leaching solution is obtained.

[0044] In one example, a mixed light beam may include a near-infrared light beam, a fluorescent light beam, and an ultraviolet-visible light beam. The mixed light beam irradiates the soil leaching solution, and the corresponding multi-modalities obtained include a near-infrared light modality, a fluorescent modality, and an ultraviolet-visible light modality. The multiple spectral sub-data of the multiple modalities may be a near-infrared light absorption spectrum, a fluorescent spectrum, and an ultraviolet-visible light absorption spectrum.

[0045] Fluorescence spectroscopy technology can achieve qualitative and quantitative analysis of the components of soil leaching solutions by stimulating fluorescent substances (such as organic matter, humic acid, microbial metabolites, etc.) in soil leaching solutions and measuring the intensity and wavelength characteristics of the emission spectrum.

[0046] Fluorescence spectroscopy, UV-visible spectroscopy and near-infrared spectroscopy are three analytical techniques based on the interaction between matter and electromagnetic radiation, but their mechanisms of action and spectral ranges are significantly different. Fluorescence spectroscopy involves the relaxation luminescence of excited state electrons, UV-visible spectroscopy is directly related to the fundamental frequency absorption of electronic transitions, and near-infrared spectroscopy focuses on the low-energy frequency absorption of molecular vibrations.

[0047] According to an embodiment of the present invention, multimodal spectral technology is used to obtain multimodal spectral data. Based on the different detection principles of multimodal spectral technology, the detection information of soil leaching solution can be enriched. Spectral feature extraction of multiple spectral sub-data can make up for the information differences between multiple modalities. Feature fusion of the first spectral features corresponding to multiple modalities can reduce the interference of other components in the soil leaching solution, improve the detection accuracy of the nitrogen concentration of the soil leaching solution, and improve the accuracy and reliability of nitrogen detection data, so that the nitrogen loss status of the soil can be more accurately evaluated, crop yield and quality can be improved, and targeted pollution prevention and control measures can be taken to reduce environmental governance costs and protect aquatic ecosystems.

[0048] According to an embodiment of the present invention, a mixed light beam is used to irradiate a soil leaching solution to obtain a mixed spectrum, wherein the mixed spectrum includes multiple initial spectra belonging to multiple modes respectively; spatial light modulation is performed on the multiple initial spectra to obtain multiple target spectra; and a detector is used to perform photoelectric detection on the multiple target spectra to obtain multimodal spectral data.

[0049] According to the embodiment of the present invention, feature extraction is performed based on the multiple spectral sub-data to obtain multiple second spectral features.

[0050] According to an embodiment of the present invention, for each spectral sub-data, the spectral sub-data is decentralized to obtain pre-processed spectral sub-data.

[0051] In one example, the spectral subdata can be mean-centered using Decentralize the data to obtain the preprocessed spectral sub-data ,in, , Is a positive integer.

[0052] (1);

[0053] (2);

[0054] in, Represents the average spectral data of the spectral sub-data of the training set during the prediction model training process.

[0055] According to an embodiment of the present invention, for each spectral sub-data, principal component analysis is performed on the preprocessed spectral sub-data to obtain a principal component matrix.

[0056] In one example, a principal component analysis (PCA) method may be used to perform principal component analysis on the preprocessed spectral sub-data.

[0057] The preprocessed spectral sub-data Divide the spectrum of multimodal spectral data by the standard deviation , get the spectral sub-data after standardization , which can be expressed as:

[0058] (3);

[0059] (4);

[0060] in, express The The spectral sub-data The absorbance at each wavelength, Represents the multimodal spectral data The average absorbance at each wavelength.

[0061] The standardized spectral sub-data , organized as a matrix .

[0062] Calculate the covariance matrix :

[0063] (5);

[0064] in, Represents the matrix transpose, for the covariance matrix Perform singular value decomposition to obtain the principal component matrix :

[0065] (6);

[0066] in, and is an orthogonal matrix, The elements on the diagonal are the covariance matrix The diagonal matrix of the singular values ​​of .

[0067] In one example, from the principal component matrix Select the first three column vectors as the spectral principal component vector matrix.

[0068] According to an embodiment of the present invention, for each spectral sub-data, a second spectral feature is extracted from the principal component matrix.

[0069] In one example, from the principal component matrix The second spectral features are extracted from the three modes, and the second spectral features corresponding to the three modes can be expressed as , , .

[0070] According to an embodiment of the present invention, a plurality of second spectral features are aligned and matched to obtain a plurality of first spectral features.

[0071] According to an embodiment of the present invention, the plurality of modalities includes a first modality and a plurality of second modalities.

[0072] In one example, the first modality may be a near infrared light modality, and the plurality of second modalities may be a fluorescent light modality and an ultraviolet-visible light modality.

[0073] According to an embodiment of the present invention, the second spectral features of the first modality are used as a reference matrix, and the second spectral features of the plurality of second modalities are aligned based on the reference matrix to obtain the third spectral features of the plurality of second modalities.

[0074] In one example, the second spectral features corresponding to the three modes are , , , as a three-dimensional space coordinate point, with the second spectral feature of the near-infrared light absorption spectrum As the reference matrix, the second spectral feature of the fluorescence spectrum and the second spectral feature of the UV-Vis absorption spectrum Align with the reference matrix respectively to obtain the third spectral feature of the fluorescence spectrum and the third spectral feature of the UV-Vis absorption spectrum .

[0075] (7);

[0076] (8).

[0077] According to an embodiment of the present invention, an iterative closest point algorithm is used to optimize and align the third spectral features of each of the plurality of second modes with the second spectral features of the first mode to obtain the first spectral features of each of the plurality of second modes.

[0078] The Iterative Closest Point (ICP) algorithm can be used to align two models. By iterative optimization, the distance between the two models is gradually reduced so that the two models match as closely as possible in space.

[0079] According to an embodiment of the present invention, an iterative closest point algorithm is used to calculate a matching error based on the third spectral features of each of the plurality of second modes and the second spectral features of the first mode.

[0080] In one example, an iterative closest point algorithm is used based on the third spectral feature of the fluorescence spectrum. , the third spectral feature of UV-visible absorption spectrum , the second spectral feature of near-infrared absorption spectrum , calculate the matching error :

[0081] (9);

[0082] in, The number of points representing the second spectral feature, The value range is .

[0083] According to an embodiment of the present invention, the third spectral feature is updated based on the matching error to obtain updated third spectral features of each of the plurality of second modalities.

[0084] In one example, based on the matching error Update the third spectral feature of fluorescence spectrum , the third spectral feature of UV-visible absorption spectrum , and obtain the updated third spectral feature of the fluorescence spectrum , the updated third spectral feature of UV-Vis absorption spectrum .

[0085] (10);

[0086] (11).

[0087] According to an embodiment of the present invention, when the matching error is less than or equal to the error threshold, the updated third spectral features of each of the plurality of second modalities are determined as the first spectral features of each of the plurality of second modalities.

[0088] According to an embodiment of the present invention, when the matching error is greater than an error threshold, optimized registration is performed based on the updated third spectral features of each of the plurality of second modalities and the second spectral features of the first modality.

[0089] According to an embodiment of the present invention, through principal component analysis, alignment based on a reference matrix can achieve a rough alignment of the second spectral features of each of the multiple second modes, and can extract key information about nitrogen. Using an iterative nearest point algorithm, a precise alignment of the second spectral features of each of the multiple second modes can be achieved, which can make up for the information differences between different modes, extract the nitrogen information between different modes into the same dimension for detection, eliminate interference, and improve the detection accuracy of nitrogen concentration.

[0090] According to an embodiment of the present invention, the plurality of first spectral features include a second spectral feature of a first modality and respective first spectral features of a plurality of second modalities.

[0091] Figure 2 A schematic diagram of a soil leachate detection method based on multimodal spectral feature fusion according to an embodiment of the present invention is shown.

[0092] like Figure 2 As shown, the first mode may be a near-infrared light mode, and the second mode may include a fluorescence mode and an ultraviolet-visible light mode. The spectral sub-data of near-infrared light, the spectral sub-data of fluorescence, and the spectral sub-data of ultraviolet-visible light are decentralized and singular value decomposed respectively to obtain a principal component matrix. The second spectral characteristics of near-infrared light, the second spectral characteristics of fluorescence, and the second spectral characteristics of ultraviolet-visible light are determined from the principal component matrix. Taking the second spectral characteristics of near-infrared light as the reference matrix, the second spectral characteristics of fluorescence and the second spectral characteristics of ultraviolet-visible light are aligned and matched with the reference matrix to obtain the third spectral characteristics of fluorescence and the third spectral characteristics of ultraviolet-visible light. Determine whether the matching error between the third spectral characteristic of fluorescence and the third spectral characteristic of ultraviolet-visible light and the second spectral characteristic of near-infrared light is greater than an error threshold. If so, update the third spectral characteristic of fluorescence and the third spectral characteristic of ultraviolet-visible light. If so, determine the updated third spectral characteristic of fluorescence and the third spectral characteristic of ultraviolet-visible light as the first spectral characteristic of fluorescence and the first spectral characteristic of ultraviolet-visible light, and determine the second spectral characteristic of near-infrared light as the first spectral characteristic of near-infrared light.

[0093] In one example, the first spectral characteristics of the three modalities include the second spectral characteristics of the near infrared light absorption spectrum , the updated third spectral feature of the fluorescence spectrum and the updated third spectral feature of the UV-Vis absorption spectrum .

[0094] Perform feature fusion on the three first spectral features to obtain the fusion feature .

[0095] (12).

[0096] Based on the fusion features, the partial least squares regression method was used to establish the matrix of the prediction model. The soil leaching solution samples with known nitrogen concentrations were used as training sets to calculate the regression parameters. . The prediction model can be expressed as:

[0097] (13);

[0098] in, represents the regression error; represents the nitrogen concentration matrix used in the training set; Represents the fusion features used in the training set.

[0099] Regression parameters The calculation formula can be expressed as:

[0100] (14).

[0101] Based on fusion features Conduct concentration testing to obtain the nitrogen concentration matrix of the soil leaching solution :

[0102] (15).

[0103] The present invention also provides a soil leachate detection device based on multi-modal spectral feature fusion.

[0104] Figure 3 A schematic diagram of a soil leachate detection device based on multimodal spectral feature fusion according to an embodiment of the present invention is shown.

[0105] like Figure 3 As shown, the soil leachate detection device based on multimodal spectral feature fusion includes a fluorescent excitation light source 210, a broadband light source 220, a beam combiner 230, a detector 240 and a processor 250.

[0106] According to an embodiment of the present invention, the fluorescent excitation light source 210 is used to excite the first light beam. The broadband light source 220 is used to excite the second light beam. The beam combiner 230 is used to combine the first light beam and the second light beam to obtain a mixed light beam, and output the mixed light beam to the soil leaching solution 201. The detector 240 is used to perform photoelectric detection on the spectrum generated by the mixed light beam in the soil leaching solution 201 to obtain multimodal spectral data, wherein the multimodal spectral data includes multiple spectral sub-data belonging to multiple modes respectively. The processor 250 is used to extract spectral features from the multiple spectral sub-data to obtain multiple first spectral features, perform feature fusion on the multiple first spectral features to obtain fused features, and perform concentration detection based on the fused features to obtain the nitrogen concentration of the soil leaching solution 201.

[0107] In one example, the second light beam may be an ultraviolet-visible light beam, or may be a near-infrared light beam.

[0108] Figure 4 A schematic diagram of a soil leachate detection device based on multimodal spectral feature fusion according to another embodiment of the present invention is shown.

[0109] According to an embodiment of the present invention, the soil leaching solution 201 is used to output a mixed spectrum based on the mixed light beam.

[0110] like Figure 4As shown, the soil leachate detection device based on multimodal spectral feature fusion also includes a light modulation module 260.

[0111] According to an embodiment of the present invention, the light modulation module 260 is used to receive the mixed spectrum, perform spatial light modulation on the mixed spectrum to obtain multiple target spectra, and project the multiple target spectra to the detector 240. The detector 240 is also used to perform photoelectric detection on the multiple target spectra to obtain multimodal spectral data.

[0112] Figure 5 A schematic diagram of a soil leachate detection device based on multimodal spectral feature fusion according to another embodiment of the present invention is shown.

[0113] like Figure 5 As shown, the light modulation module 260 includes a grating 261 , a first lens 262 , a spatial light modulator 263 and a second lens 264 .

[0114] According to an embodiment of the present invention, the grating 261 is used to perform spectroscopic processing on the mixed spectrum to obtain multiple initial spectra belonging to multiple modes. The first lens 262 is used to converge the multiple initial spectra to the spatial light modulator 263. The spatial light modulator 263 is used to perform spectral spatial distribution modulation on the multiple initial spectra to obtain multiple target spectra. The second lens 264 is used to project the multiple target spectra to multiple arrays of the detector 240 respectively.

[0115] Figure 6 A schematic diagram of a soil leachate detection device based on multimodal spectral feature fusion according to yet another embodiment of the present invention is shown.

[0116] like Figure 6 As shown, the soil leachate detection device based on multimodal spectral feature fusion can also include a control device 270. The control device 270 can be connected to the processor 250, the fluorescent excitation light source 210 and the broadband light source 220. The control device 270 can be used to control the fluorescent excitation light source 210 and the broadband light source 220 to work in time sharing.

[0117] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination advantageously. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A soil leachate detection method based on multimodal spectral feature fusion, characterized in that: include: In response to a concentration detection request, irradiating the soil leaching solution with a mixed light beam to obtain multimodal spectral data, wherein the multimodal spectral data includes a plurality of spectral sub-data respectively belonging to a plurality of modalities; Extracting features based on the multiple spectral sub-data respectively to obtain multiple second spectral features; Aligning and matching the multiple second spectral features to obtain multiple first spectral features; Performing feature fusion on the plurality of first spectral features to obtain a fused feature; and The concentration is detected based on the fusion feature to obtain the nitrogen concentration of the soil leaching solution.

2. The method according to claim 1, characterized in that The plurality of modalities include a first modality and a plurality of second modalities; The step of aligning and matching the plurality of second spectral features to obtain the plurality of first spectral features includes: Using the second spectral feature of the first mode as a reference matrix, aligning the second spectral features of the plurality of second modes based on the reference matrix to obtain third spectral features of the plurality of second modes; and An iterative closest point algorithm is used to optimize and align the third spectral features of each of the multiple second modes with the second spectral features of the first mode to obtain the first spectral features of each of the multiple second modes, wherein the multiple first spectral features include the second spectral features of the first mode and the first spectral features of each of the multiple second modes.

3. The method according to claim 2, characterized in that The method of using an iterative closest point algorithm to optimize and align the third spectral features of each of the plurality of second modes with the second spectral features of the first mode to obtain the first spectral features of each of the plurality of second modes includes: Using the iterative closest point algorithm, based on the third spectral features of each of the plurality of second modes and the second spectral features of the first mode, a matching error is calculated; updating the third spectral feature based on the matching error to obtain updated third spectral features of each of the plurality of second modes; and When the matching error is less than or equal to an error threshold, the updated third spectral features of each of the plurality of second modalities are determined as the first spectral features of each of the plurality of second modalities.

4. The method according to claim 3, characterized in that The method further comprises: When the matching error is greater than the error threshold, an optimized alignment is performed based on the updated third spectral features of each of the plurality of second modalities and the second spectral features of the first modality.

5. The method according to claim 1, characterized in that The extracting features based on the plurality of spectral sub-data to obtain a plurality of second spectral features comprises: For each spectral sub-data, performing decentralization processing on the spectral sub-data to obtain pre-processed spectral sub-data; Performing principal component analysis on the preprocessed spectral sub-data to obtain a principal component matrix; and The second spectral feature is extracted from the principal component matrix.

6. The method according to claim 1, characterized in that The method of irradiating the soil leaching solution with a mixed light beam to obtain multimodal spectral data includes: irradiating the soil leaching solution with the mixed light beam to obtain a mixed spectrum, wherein the mixed spectrum includes a plurality of initial spectra respectively belonging to the plurality of modes; Performing spatial light modulation on the multiple initial spectra to obtain multiple target spectra; and The detector is used to perform photoelectric detection on the multiple target spectra to obtain the multimodal spectral data.

7. A soil leaching solution detection device based on multimodal spectral feature fusion, characterized in that: include: A fluorescence excitation light source, used to excite the first light beam; a broadband light source for exciting the second light beam; A beam combiner, used for combining the first light beam and the second light beam to obtain a mixed light beam, and outputting the mixed light beam to the soil leaching solution; a detector, configured to perform photoelectric detection on the spectrum of the soil leaching solution generated by the mixed light beam to obtain multimodal spectral data, wherein the multimodal spectral data includes a plurality of spectral sub-data respectively belonging to a plurality of modes; and The processor is used to perform feature extraction based on the multiple spectral sub-data to obtain multiple second spectral features; align and match the multiple second spectral features to obtain multiple first spectral features; perform feature fusion on the multiple first spectral features to obtain fusion features, and perform concentration detection based on the fusion features to obtain the nitrogen concentration of the soil leaching solution.

8. The device according to claim 7, characterized in that The soil leaching solution is used to output a mixed spectrum based on the mixed light beam; The device also includes: an optical modulation module, configured to receive the mixed spectrum, perform spatial optical modulation on the mixed spectrum to obtain a plurality of target spectra, and project the plurality of target spectra to a detector; Wherein, the detector is also used to perform photoelectric detection on the multiple target spectra to obtain the multimodal spectral data.

9. The device according to claim 8, characterized in that The optical modulation module comprises: A grating, used for performing spectroscopic processing on the mixed spectrum to obtain a plurality of initial spectra respectively belonging to the plurality of modes; A first lens, configured to converge the plurality of initial light spectra to a spatial light modulator; The spatial light modulator is used to perform spectral spatial distribution modulation on the multiple initial spectra respectively to obtain the multiple target spectra; and The second lens is used to project the multiple target spectra to the multiple arrays of the detector respectively.

Images