CN115859029B - Spectrum quantitative analysis method based on two-dimensional reconstruction - Google Patents
Spectrum quantitative analysis method based on two-dimensional reconstruction Download PDFInfo
- Publication number
- CN115859029B CN115859029B CN202211511413.7A CN202211511413A CN115859029B CN 115859029 B CN115859029 B CN 115859029B CN 202211511413 A CN202211511413 A CN 202211511413A CN 115859029 B CN115859029 B CN 115859029B
- Authority
- CN
- China
- Prior art keywords
- spectrum
- quantitative analysis
- dimensional
- dimensional reconstruction
- order
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a spectrum quantitative analysis method based on two-dimensional reconstruction, which comprises the following steps: s1: acquiring a standard sample spectrum signal for modeling, and carrying out two-dimensional reconstruction on one-dimensional spectrum data based on signal transformation methods of different parameters; s2: constructing a spectrum high-order feature extraction network based on two-dimensional CNN; s3: establishing and training a component content regression model; s4: after two-dimensional reconstruction is carried out on the spectrum signal of the sample to be detected, a feature extraction network and a regression model are input, a high-order feature map is extracted, and the quantitative analysis result of each component is calculated. According to the spectrum quantitative analysis method based on two-dimensional reconstruction, the high-order characteristics of spectrum signals are extracted through two-dimensional reconstruction and a deep CNN network, and a component quantitative analysis model is established, so that the quantitative analysis of multi-component overlapped spectrums can be completed with high quality, the quantitative analysis precision is improved, the spectrum measurement of complex substance components is realized, and the spectrum quantitative analysis method can be used for the quantitative analysis of spectrum signals in the fields of chemical industry, foods, environment and the like.
Description
Technical Field
The invention relates to the technical field of spectrum quantitative analysis, in particular to a spectrum quantitative analysis method based on two-dimensional reconstruction.
Background
The spectrum is an important technology for quantitative analysis of the components of substances, and is widely applied to the fields of environment, food, chemical industry and the like. In the detection of complex samples, spectra of multiple components are mutually influenced and are highly overlapped, so that the characteristics of the spectra of each component are difficult to extract, and the quantitative analysis accuracy of the complex samples is reduced.
Spectral preprocessing is an important step in spectral quantitative analysis, which typically includes baseline correction, multiple scatter correction, filtering, wavelength selection, signal conversion, and the like. Parameters of spectrum pretreatment are critical to the performance of spectrum quantitative analysis, however, complex coupling relations exist between each spectrum pretreatment and modeling methods, pretreatment parameters are often obtained through repeated experimental trial and error, and complex pretreatment method selection and parameter optimization processes lead to complex spectrum quantitative analysis.
In recent years, a model based on deep learning is increasingly applied to spectral quantitative analysis. And an end-to-end quantitative analysis model is established through a model self-learning spectrum signal preprocessing mode, so that the complicated preprocessing method selection and parameter optimization process can be avoided in the spectrum quantitative analysis based on deep learning. However, the acquisition cost of the spectrum standard sample of the substance to be detected is still higher, and the spectrum quantitative analysis of complex, overlapped and small samples can greatly influence the model training difficulty and the quantitative analysis precision of the method, so that the precision of the measurement result is reduced.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the invention provides a spectrum quantitative analysis method based on two-dimensional reconstruction, which can effectively extract the characteristics of each component in a multi-component spectrum of a sample to be detected, establish a quantitative analysis model of each component, and particularly can effectively improve the precision of the quantitative analysis model in the spectrum analysis of complex, overlapped and small samples.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a method for quantitative analysis of spectra based on two-dimensional reconstruction, comprising the steps of:
step S1: acquiring a standard sample spectrum signal for modeling, and carrying out two-dimensional reconstruction on one-dimensional spectrum data based on signal transformation methods of different parameters;
step S2: constructing a spectrum high-order feature extraction network based on two-dimensional CNN;
step S3: establishing and training a component content regression model;
step S4: and (3) after two-dimensional reconstruction is carried out on the spectrum signal of the sample to be detected, a high-order characteristic diagram is extracted from the input model, and a quantitative analysis result of each component is obtained.
The technical scheme is further improved as follows:
preferably, in the step S1, a signal conversion method with continuously variable parameters is used to process the same set of standard sample spectrum signals, then the processed spectrum signals are sequentially arranged along the parameter dimension, and then two-dimensional reconstruction is performed to obtain a two-dimensional spectrum matrix, where the signal conversion method includes but is not limited to wavelet transformation and fractional differentiation.
Preferably, in the step S2, the two-dimensional reconstructed spectrum matrix in the step S1 is used as an input layer of the higher-order feature extraction network, and the higher-order feature map is extracted through calculation of a multi-layer convolution layer and a pooling layer and is transmitted to an output layer of the higher-order feature extraction network.
Preferably, in the step S3, the method specifically includes the following steps:
s31: adopting the high-order feature map obtained in the step S2 as input;
s32: regression is carried out on the high-order feature map by adopting a regression network, and a regression result is obtained through calculation;
s33: taking the mean square error as a loss function;
s34: updating the network weights using an error back propagation algorithm;
s35: and S31-S34 are repeated until the preset training times are completed, and a quantitative analysis model is built.
Preferably, in the step S4, the method specifically includes the following steps:
s41: measuring the spectrum of a sample to be measured, and carrying out two-dimensional reconstruction to obtain a two-dimensional reconstructed spectrum matrix;
s42: inputting the high-order feature extraction network to obtain a high-order feature map of a sample to be detected;
s43: inputting the high-order characteristic diagram into a quantitative analysis model to obtain a quantitative analysis result.
Compared with the prior art, the spectrum quantitative analysis method based on two-dimensional reconstruction has the following advantages:
(1) According to the spectrum quantitative analysis method based on the two-dimensional reconstruction, through the two-dimensional reconstruction and the two-dimensional CNN feature extraction network, the detail features of the multi-element spectrum can be effectively extracted, the quantitative analysis precision of the overlapped spectrum is improved, the complicated parameter optimization process in the traditional chemometric preprocessing method is avoided, and the modeling efficiency of quantitative analysis is improved.
(2) The spectrum quantitative analysis method based on two-dimensional reconstruction integrates the advantages of chemometrics and deep learning, reduces the learning difficulty of a deep learning model by using the knowledge acquired by chemometrics, and particularly improves the model learning efficiency and modeling precision for small-sample, multi-element and overlapping spectrum analysis.
Drawings
FIG. 1 is a schematic flow chart of the method for quantitatively analyzing spectra of the present invention.
Fig. 2 is a graph of near infrared spectrum signals of a corn sample in an embodiment of the invention.
Fig. 3 is a near infrared spectrum signal diagram of a drug sample according to an embodiment of the present invention.
Fig. 4 is a flowchart of a method for spectral quantification analysis of near infrared spectrum signals of a corn sample based on two-dimensional reconstruction in an embodiment of the present invention.
Detailed Description
The following describes specific embodiments of the present invention in detail. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
FIG. 1 shows an embodiment of the two-dimensional reconstruction-based spectroscopic quantitative analysis method of the present invention, comprising the steps of:
s1: obtaining a set of standard spectrum signals X for modeling train And the corresponding component content y thereof train ，X train Consists of M samples, each sample containing spectral responses at N spectral wavelengths, X train The size is MxN, and two-dimensional reconstruction is carried out on one-dimensional spectrum data by using different k-parameter signal transformation methods to obtain a two-dimensional reconstruction spectrum matrix X train_2D The size is MXNXk.
The two-dimensional reconstructed spectrum matrix X train_2D The process of (2) is as follows:
s11: processing of a spectral signal X using a signal transformation method with continuous variation of parameters train ；
S12: sequentially arranging the processed spectrum signals along the parameter dimension to reconstruct into a two-dimensional spectrum matrix X train_2D In this embodiment, the optical spectrum signals after processing are arranged in sequence along the parameter dimension according to the order of the parameter sizes.
S2: constructing a two-dimensional CNN-based spectrum high-order feature extraction network N F 。
Wherein, the high-order feature extraction network N F Is input as the two-dimensional reconstructed spectral matrix X in step S1 train_2D And extracting a high-order feature map through calculation of a multi-layer convolution layer and a pooling layer, and transmitting the high-order feature map to an output layer of a high-order feature extraction network.
In this embodiment, the higher-order feature extraction network N in step S2 F Using a 2-layer two-dimensional CNN network, namely Conv1 and Conv2; the convolution kernel sizes of Conv1 and Conv2 are 3×3, and the convolution kernel carries out convolution operation by taking one wavelength as a step length.
S3: and (5) establishing and training a component content regression model.
The method specifically comprises the following steps:
s31: extracting network N using higher order features F Extracting the two-dimensional reconstruction spectrum matrix X in the step S1 train_2D Higher order features of (2) to obtain higher order feature map
S32: using regression network N R For high-order feature mapRegression is carried out, and a regression result y is obtained through calculation train_p ；
S33: taking the mean square error as a loss function, adding an L2 regular term into the loss function, wherein the loss function L is as follows:
wherein w represents N R And N F Is a weight matrix of the whole system;
s34: updating N using error back propagation algorithm R And N F The weight of (a);
s35: repeating S31-S34 until the preset training times are completed, and establishing a final quantitative analysis model.
S4: and (3) after two-dimensional reconstruction of the spectrum signal of the sample to be detected, inputting the spectrum signal into a model to extract a high-order characteristic diagram, and obtaining a quantitative analysis result of each component.
The method specifically comprises the following steps:
s41: measuring spectrum X of sample to be measured test And performing two-dimensional reconstruction to obtain a two-dimensional reconstructed spectrum matrix X test_2D ；
S42: x is to be test_2D Input high-order feature extraction network N F Obtaining a high-order characteristic diagram of a sample to be detected
S43: will be a high-order feature mapInputting the quantitative analysis model to obtain a quantitative analysis result y test_p 。
According to the spectrum quantitative analysis method, through the two-dimensional reconstruction and the two-dimensional CNN feature extraction network, the detail features of the multi-element spectrum can be effectively extracted, and the quantitative analysis precision of the overlapped spectrum is improved.
In order to explain the quantitative analysis method of the present invention in detail, the analysis of the moisture content of corn by near infrared spectrum will be described in further detail. It is understood that within the scope of the present invention, the above-described technical features of the present invention and technical features specifically described below (e.g., in the examples) may be combined with each other and associated with each other, thereby constituting a preferred technical solution.
Example 1
As shown in fig. 2 and 4, the near infrared spectrum of the corn sample is two-dimensionally reconstructed by using the near infrared spectrum of the corn as an analysis object, and then passed through a spectrum Gao Jiete based on two-dimensional CNNSign extraction network N F Extracting high-order features of the spectrum, and inputting the high-order feature map into a component content regression model N R Obtaining a predicted value of the sample. And calculating a loss function according to the deviation of the predicted value and the true value, training the model by an error back propagation method, and finally obtaining the near infrared spectrum quantitative analysis model of the corn moisture content.
The specific process is as follows:
step S1: near infrared spectra X and corresponding moisture content y of a set of corn samples are acquired as a sample set.
The near infrared spectrum signal of each sample is one-dimensional data, the spectrum matrix X comprises 80 samples and spectrum response (1100-2490 nm) at 700 wavelengths, 64 samples are selected to form a training set X by an SPXY method train And y train The remaining 16 samples form test set X test And y test ，X train The size is 64 multiplied by 700, X test The size is 16×700.
The fractional differential is adopted as a signal transformation method, and the definition of the adopted Caputo fractional differential is shown as a formula (3):
wherein alpha is differential order and n is the smallest integer greater than or equal to alpha) Γ () is a gamma function.
The differential orders are respectively 0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and 1.0, and 11 transformed signals are obtained by combining a filtering algorithm, wherein the 0.0 order is the original spectrum data, and the 1.0 order is the first derivative of the spectrum signal. And normalizing the spectrum signals with the same differential order in the data set, and eliminating the dimension difference caused by differential operation. Then, the differential signals are arranged from small to large in differential order, a two-dimensional reconstructed spectrum matrix is obtained through reconstruction, and a two-dimensional reconstructed spectrum matrix X of a training set is obtained train_2D Size and dimensions of64X 700X 11.
Step S2: constructing a two-dimensional CNN-based spectrum high-order feature extraction network N F 。
The two-dimensional convolution layer is composed of two-dimensional convolution layers, namely Conv1, conv2, and Conv1 and Conv2, wherein the convolution kernels are 3 multiplied by 3, the convolution kernels take one wavelength as a step length to carry out convolution operation, conv1 comprises 8 convolution kernels, and Conv2 comprises 16 convolution kernels. The convolution layers include convolution operations, batchNorm normalization operations, and an activation layer. The activation layer in the feature extraction network adopts a LeakyRelu nonlinear activation function.
Step S3: training and establishing a component content regression model.
Construction of component content regression model N Using fully connected and regressive layers R . The activation layer in the regression model uses a LeakyRelu nonlinear activation function.
Training a component content regression model according to the model structure designed in the step S2, wherein the model uses a mean square error as a loss function in the training process, and an L2 regular term is added in the loss function, and the loss function is shown in a formula (4):
the error back propagation method and the Adam optimizer are adopted to train the network weights. After the preset training times are completed for I times, a final quantitative analysis model is established.
Step four: using test set X test And y test Test model performance, X test Two-dimensional reconstruction is carried out according to the step S1 to obtain a reconstruction matrix X test_2D The size is 16×700×11. X is to be test_2D Inputting the higher-order feature extraction network in the step S2 and the regression network in the step S3 to obtain measured values obtained by solving the modelBy determining the coefficient R 2 And evaluating the performance of the model by predicting root mean square error, as shown in the formula (5) and the formula (6):
where N is the number of test set samples.
In this embodiment, n=16, y i Is the true value of the moisture content of the ith sample in the test set,the prediction value obtained by the spectrum quantitative analysis method based on the two-dimensional reconstruction.
Based on the original spectrum, the fractional differential spectrum and the one-dimensional CNN and the two-dimensional reconstructed spectrum and the two-dimensional CNN, the corn moisture content measurement results are subjected to comparative analysis, 100 times of repeated modeling is performed by using each method, and the results are counted. The mean and standard deviation of the results are shown in table 1.
TABLE 1 measurement results of corn moisture content by combining original spectra, fractional differential spectra, one-dimensional CNN, and two-dimensional reconstructed spectra, two-dimensional CNN
| Method | RMSEP | R 2 |
| Original Spectrum-CNN | 0.0623±0.0046 | 0.8834±0.0175 |
| Fractional order derivative-CNN | ||
| 0.1 order | 0.0636±0.0061 | 0.8780±0.0245 |
| 0.2 order | 0.0692±0.0096 | 0.8541±0.0415 |
| 0.3 order | 0.0599±0.0104 | 0.8896±0.0401 |
| 0.4 th order | 0.0510±0.0054 | 0.9214±0.0179 |
| 0.5 th order | 0.0473±0.0051 | 0.9325±0.0157 |
| 0.6 th order | 0.0517±0.0064 | 0.9188±0.0203 |
| 0.7 th order | 0.0517±0.0062 | 0.9189±0.0193 |
| 0.8 th order | 0.0511±0.0070 | 0.9206±0.0212 |
| 0.9 th order | 0.0483±0.0061 | 0.9291±0.0182 |
| 1.0 order | 0.0522±0.0049 | 0.9179±0.0157 |
| Two-dimensional reconstruction spectrum-2 DCNN | 0.0394±0.0018 | 0.9536±0.0043 |
As can be seen from Table 1, the method provided by the invention has a RMSEP of 0.0394, R 2 0.9536, the model precision is higher, and a smaller standard deviation also indicates that the model has better stability. In conclusion, the two-dimensional reconstruction-based spectrum quantitative analysis method provided by the invention can effectively establish a quantitative analysis model with high precision and good stability in small sample and complex spectrum analysis.
Example 2
As shown in fig. 3, the near infrared spectrum of the drug is taken as an analysis object, the near infrared spectrum of the drug sample is subjected to two-dimensional reconstruction, and then the network N is extracted by the spectrum higher-order feature based on the two-dimensional CNN F Extracting high-order features of the spectrum, and inputting the high-order feature map into a component content regression model N R Obtaining a predicted value of the sample. And calculating a loss function according to the deviation of the predicted value and the true value, training the model by an error back propagation method, and finally obtaining a near infrared spectrum quantitative analysis model of the content of the effective components of the medicine.
The specific process is as follows:
step S1: and acquiring a near infrared spectrum X of a group of medicine samples and the corresponding content y of the effective components as a sample set.
The near infrared spectrum signal of each sample is one-dimensional data, and the spectrum matrix X comprises 309 samples and spectrum response (7400-10507 cm -1 ) Selecting 240 samples to form training set X by SPXY method train And y train The remaining 69 samples form test set X test And y test ，X train Size 240X 404, X test The size is 69×404.
The fractional differential is adopted as a signal transformation method, and the definition of the adopted Caputo fractional differential is shown as a formula (3):
wherein alpha is differential order and n is the smallest integer greater than or equal to alpha) Γ () is a gamma function.
The differential orders are respectively 0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and 1.0, and 11 transformed signals are obtained by combining a filtering algorithm, wherein the 0.0 order is the original spectrum data, and the 1.0 order is the first derivative of the spectrum signal. And normalizing the spectrum signals with the same differential order in the data set, and eliminating the dimension difference caused by differential operation. Then, the differential signals are arranged from small to large in differential order, a two-dimensional reconstructed spectrum matrix is obtained through reconstruction, and a two-dimensional reconstructed spectrum matrix X of a training set is obtained train_2D The size is 240×404×11.
Step S2: constructing a two-dimensional CNN-based spectrum high-order feature extraction network N F 。
The two-dimensional convolution layer is composed of two-dimensional convolution layers, namely Conv1, conv2, and Conv1 and Conv2, wherein the convolution kernels are 3 multiplied by 3, the convolution kernels take one wavelength as a step length to carry out convolution operation, conv1 comprises 8 convolution kernels, and Conv2 comprises 16 convolution kernels. The convolution layers include convolution operations, batchNorm normalization operations, and an activation layer. The activation layer in the feature extraction network adopts a LeakyRelu nonlinear activation function.
Step S3: training and establishing a component content regression model.
Construction of component content regression model N Using fully connected and regressive layers R . The activation layer in the regression model uses a LeakyRelu nonlinear activation function.
Training a component content regression model according to the model structure designed in the step S2, wherein the model uses a mean square error as a loss function in the training process, and an L2 regular term is added in the loss function, and the loss function is shown in a formula (7):
the error back propagation method and the Adam optimizer are adopted to train the network weights. After the preset training times are completed for I times, a final quantitative analysis model is established.
Step four: using test set X test And y test Test model performance, X test Two-dimensional reconstruction is carried out according to the step S1 to obtain a reconstruction matrix X test_2D The size is 69×404×11. X is to be test_2D Inputting the higher-order feature extraction network in the step S2 and the regression network in the step S3 to obtain measured values obtained by solving the modelBy determining the coefficient R 2 And evaluating the performance of the model by predicting root mean square error, as shown in the formulas (8) and (9):
where N is the number of test set samples.
In this embodiment, n=69,y i is the true value of the active ingredient content of the ith sample in the test set,the prediction value obtained by the spectrum quantitative analysis method based on the two-dimensional reconstruction.
Based on the original spectrum, the fractional differential spectrum and the one-dimensional CNN and the two-dimensional reconstruction spectrum and the two-dimensional CNN, the measurement results of the content of the effective components of the medicine are subjected to comparative analysis, the modeling is repeated 100 times by using each method, and the results are counted. The mean and standard deviation of the results are shown in table 2.
TABLE 2 measurement results of the content of active ingredients in the drug by combining the original spectrum, each fractional differential spectrum, one-dimensional CNN and two-dimensional reconstruction spectrum, and two-dimensional CNN
| Method | RMSEP | R 2 |
| Original Spectrum-CNN | 0.2968±0.0111 | 0.9574±0.0032 |
| Fractional order derivative-CNN | ||
| 0.1 order | 0.2892±0.0083 | 0.9596±0.0024 |
| 0.2 order | 0.2894±0.0064 | 0.9595±0.0018 |
| 0.3 order | 0.2887±0.0062 | 0.9597±0.0017 |
| 0.4 th order | 0.2849±0.0058 | 0.9608±0.0016 |
| 0.5 th order | 0.2802±0.0065 | 0.9621±0.0018 |
| 0.6 th order | 0.2749±0.0092 | 0.9635±0.0025 |
| 0.7 th order | 0.2795±0.0123 | 0.9622±0.0034 |
| 0.8 th order | 0.2760±0.0111 | 0.9632±0.0030 |
| 0.9 th order | 0.2794±0.0139 | 0.9622±0.0038 |
| 1.0 order | 0.2752±0.0127 | 0.9634±0.0034 |
| Two-dimensional reconstruction spectrum-2 DCNN | 0.2719±0.0059 | 0.9643±0.0016 |
As can be seen from Table 2, the method provided by the invention has a RMSEP of 0.2719 and R 2 0.9634, the model precision is higher, and a smaller standard deviation also indicates that the model has better stability. In conclusion, the two-dimensional reconstruction-based spectrum quantitative analysis method provided by the invention can effectively establish a quantitative analysis model with high precision and good stability in small sample and complex spectrum analysis.
The above embodiments are merely preferred embodiments of the present invention, and are not intended to limit the present invention in any way. While the invention has been described with reference to preferred embodiments, it is not intended to be limiting. Therefore, any simple modification, equivalent variation and modification of the above embodiments according to the technical substance of the present invention shall fall within the scope of the technical solution of the present invention.
Claims (3)
1. A method for quantitative analysis of spectra based on two-dimensional reconstruction, the method comprising the steps of:
step S1: acquiring a standard sample spectrum signal for modeling, and carrying out two-dimensional reconstruction on one-dimensional spectrum data based on signal transformation methods of different parameters;
step S2: constructing a spectrum high-order feature extraction network based on two-dimensional CNN;
step S3: establishing and training a component content regression model;
step S4: after two-dimensional reconstruction is carried out on a spectrum signal of a sample to be detected, a high-order feature map is extracted from an input model, and a quantitative analysis result of each component is calculated;
the step S4 specifically includes the following steps:
s41: measuring the spectrum of a sample to be measured, and carrying out two-dimensional reconstruction to obtain a two-dimensional reconstructed spectrum matrix;
s42: inputting the two-dimensional reconstructed spectrum matrix into a high-order feature extraction network to obtain a high-order feature map of a sample to be detected;
s43: inputting the high-order characteristic diagram into a quantitative analysis model to obtain a quantitative analysis result;
in the step S1, signal processing is carried out on the same group of standard sample spectrum signals by adopting a signal conversion method with continuously-changing parameters, the processed spectrum signals are sequentially arranged along parameter dimensions, and two-dimensional reconstruction is carried out to obtain a two-dimensional spectrum matrix;
in the step S1, the signal transformation method uses fractional differentiation.
2. The method for quantitative analysis of spectrum based on two-dimensional reconstruction according to claim 1, wherein in the step S2, the two-dimensional reconstruction spectrum matrix in the step S1 is used as an input layer of a higher-order feature extraction network, and the higher-order feature map is extracted through calculation of a multi-layer convolution layer and a pooling layer and transmitted to an output layer of the higher-order feature extraction network.
3. The method for quantitative analysis of spectrum based on two-dimensional reconstruction according to claim 1, wherein in step S3, the method specifically comprises the following steps:
s31: adopting the high-order feature map obtained in the step S2 as input;
s32: regression is carried out on the high-order feature map by adopting a regression network, and a regression result is obtained through calculation;
s33: taking the mean square error as a loss function;
s34: updating the network weights using an error back propagation algorithm;
s35: and S31-S34 are repeated until the preset training times are completed, and a quantitative analysis model is built.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211511413.7A CN115859029B (en) | 2022-11-29 | 2022-11-29 | Spectrum quantitative analysis method based on two-dimensional reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211511413.7A CN115859029B (en) | 2022-11-29 | 2022-11-29 | Spectrum quantitative analysis method based on two-dimensional reconstruction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115859029A CN115859029A (en) | 2023-03-28 |
| CN115859029B true CN115859029B (en) | 2023-09-15 |
Family
ID=85667777
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211511413.7A Active CN115859029B (en) | 2022-11-29 | 2022-11-29 | Spectrum quantitative analysis method based on two-dimensional reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115859029B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106124449A (en) * | 2016-06-07 | 2016-11-16 | 中国科学院合肥物质科学研究院 | A kind of soil near-infrared spectrum analysis Forecasting Methodology based on degree of depth learning art |
| CN109493287A (en) * | 2018-10-10 | 2019-03-19 | 浙江大学 | A kind of quantitative spectra data analysis processing method based on deep learning |
| CN110033032A (en) * | 2019-03-29 | 2019-07-19 | 中国科学院西安光学精密机械研究所 | A kind of histotomy classification method based on micro- high light spectrum image-forming technology |
-
2022
- 2022-11-29 CN CN202211511413.7A patent/CN115859029B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106124449A (en) * | 2016-06-07 | 2016-11-16 | 中国科学院合肥物质科学研究院 | A kind of soil near-infrared spectrum analysis Forecasting Methodology based on degree of depth learning art |
| CN109493287A (en) * | 2018-10-10 | 2019-03-19 | 浙江大学 | A kind of quantitative spectra data analysis processing method based on deep learning |
| CN110033032A (en) * | 2019-03-29 | 2019-07-19 | 中国科学院西安光学精密机械研究所 | A kind of histotomy classification method based on micro- high light spectrum image-forming technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115859029A (en) | 2023-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109493287B (en) | Deep learning-based quantitative spectral data analysis processing method | |
| Türker-Kaya et al. | A review of mid-infrared and near-infrared imaging: principles, concepts and applications in plant tissue analysis | |
| Zhang et al. | Noise reduction in the spectral domain of hyperspectral images using denoising autoencoder methods | |
| Cogdill et al. | Single-kernel maize analysis by near-infrared hyperspectral imaging | |
| CN110736707B (en) | Spectrum detection optimization method for transferring spectrum model from master instrument to slave instrument | |
| CN110702656A (en) | Vegetable oil pesticide residue detection method based on three-dimensional fluorescence spectrum technology | |
| Liu et al. | Analysis of chlorophyll concentration in potato crop by coupling continuous wavelet transform and spectral variable optimization | |
| Pham et al. | Soil profile analysis using interactive visualizations, machine learning, and deep learning | |
| Fang et al. | Optical absorption and scattering properties at 900–1650 nm and their relationships with soluble solid content and soluble sugars in apple flesh during storage | |
| Shao et al. | Near-infrared spectroscopy for classification of oranges and prediction of the sugar content | |
| Li et al. | Near-infrared spectroscopy coupled chemometric algorithms for rapid origin identification and lipid content detection of Pinus koraiensis seeds | |
| Li et al. | Hyperspectral imaging with machine learning approaches for assessing soluble solids content of tribute citru | |
| Li et al. | Quantitative analysis of aflatoxin B1 of peanut by optimized support vector machine models based on near-infrared spectral features | |
| Abbaspour-Gilandeh et al. | Nondestructive estimation of the chlorophyll b of apple fruit by color and spectral features using different methods of hybrid artificial neural network | |
| Hong-hong et al. | Application of wavelength selection strategy based on SWCSS-UVE algorithm in improving model transfer for near infrared analysis of holocellulose and lignin | |
| CN112926625B (en) | Deviation influence factor analysis method for satellite radiation data | |
| CN115859029B (en) | Spectrum quantitative analysis method based on two-dimensional reconstruction | |
| CN107247033B (en) | Identify the method for Huanghua Pear maturity based on rapid decay formula life cycle algorithm and PLSDA | |
| Wang et al. | Mark-Spectra: A convolutional neural network for quantitative spectral analysis overcoming spatial relationships | |
| Raki et al. | Combining AI tools with non-destructive technologies for crop-based food safety: A comprehensive review | |
| Liu et al. | Estimation of chlorophyll content in maize canopy using wavelet denoising and SVR method | |
| CN116883720A (en) | Fruit and vegetable pesticide residue detection method and system based on spatial spectrum attention network | |
| Liu et al. | Non-Destructive Quality-Detection Techniques for Cereal Grains: A Systematic Review | |
| Zhu et al. | A UV–Vis spectroscopic detection method for cobalt ions in zinc sulfate solution based on discrete wavelet transform and extreme gradient boosting | |
| Tegegn et al. | Convolutional neural networks for quantitative prediction of different organic materials using near-infrared spectrum |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |