CN114778485B - Variety identification method and system based on near infrared spectrum and attention mechanism network - Google Patents
Variety identification method and system based on near infrared spectrum and attention mechanism network Download PDFInfo
- Publication number
- CN114778485B CN114778485B CN202210678616.9A CN202210678616A CN114778485B CN 114778485 B CN114778485 B CN 114778485B CN 202210678616 A CN202210678616 A CN 202210678616A CN 114778485 B CN114778485 B CN 114778485B
- Authority
- CN
- China
- Prior art keywords
- near infrared
- attention mechanism
- infrared spectrum
- mechanism network
- network model
- 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
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 173
- 230000007246 mechanism Effects 0.000 title claims abstract description 164
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000001228 spectrum Methods 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 38
- 238000007781 pre-processing Methods 0.000 claims abstract description 20
- 230000003595 spectral effect Effects 0.000 claims description 109
- 230000006870 function Effects 0.000 claims description 59
- 238000000605 extraction Methods 0.000 claims description 54
- 238000011176 pooling Methods 0.000 claims description 24
- 239000000126 substance Substances 0.000 claims description 15
- 238000004497 NIR spectroscopy Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 239000010410 layer Substances 0.000 description 28
- 241000209094 Oryza Species 0.000 description 5
- 235000007164 Oryza sativa Nutrition 0.000 description 5
- 235000009566 rice Nutrition 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 241000209140 Triticum Species 0.000 description 2
- 235000021307 Triticum Nutrition 0.000 description 2
- 240000008042 Zea mays Species 0.000 description 2
- 235000005824 Zea mays ssp. parviglumis Nutrition 0.000 description 2
- 235000002017 Zea mays subsp mays Nutrition 0.000 description 2
- 235000005822 corn Nutrition 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000003912 environmental pollution Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 108010044467 Isoenzymes Proteins 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- Strategic Management (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Tourism & Hospitality (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
Abstract
The invention provides a variety identification method and system based on a near infrared spectrum and an attention mechanism network, wherein the method comprises the following steps: preprocessing the obtained initial near infrared spectrum of the crop to be identified to obtain the near infrared spectrum to be identified; and inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction and identification to obtain an authenticity prediction result of variety identification, wherein the attention mechanism network model is obtained by training an initial attention mechanism network model through a spectrum data set sample. By inputting the near infrared spectrum to be identified into the trained attention mechanism network model, the authenticity of a plurality of crop varieties can be quickly and accurately judged, the accuracy of identifying and identifying the crop seed varieties is improved, and simple, convenient and efficient crop variety classification is realized.
Description
Technical Field
The invention relates to the technical field of information processing, in particular to a variety identification method and system based on near infrared spectrum and attention mechanism network.
Background
Wheat, rice and corn are the major food crops in our country. The method has the advantages that the crops of different varieties are correctly classified, the method has important significance for researching the yield of seeds and the variety breeding work of the crops, and the traditional crop variety authenticity identification technology such as DNA molecular identification, isoenzyme identification, field identification and the like has the defects of complex operation, slow detection, sample damage and environmental pollution, so that the method is necessary to explore a simple and efficient crop variety classification method.
Disclosure of Invention
The invention provides a variety identification method and system based on a near infrared spectrum and an attention mechanism network, which are used for solving the problems of complex operation, slow detection, sample damage and environmental pollution in the prior art and realizing simple, convenient and efficient crop variety classification.
The invention provides a variety identification method based on a near infrared spectrum and an attention mechanism network, which comprises the following steps:
preprocessing the obtained initial near infrared spectrum of the crop to be identified to obtain the near infrared spectrum to be identified;
inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction identification to obtain an authenticity prediction result of variety identification;
wherein the attention mechanism network model is obtained by training an initial attention mechanism network model through a training set.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, the obtained near infrared spectrum of the crop to be identified is preprocessed to obtain the near infrared spectrum to be identified, and the method comprises the following steps:
determining a near infrared spectrum mean value and a near infrared spectrum standard deviation according to the initial near infrared spectrum;
and standardizing the initial near infrared spectrum according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, the near infrared spectrum mean value and the near infrared spectrum standard deviation are determined through the following formulas:
wherein the content of the first and second substances,
is the second in the initial near infrared spectrumiThe first of each varietyjThe data of the bar spectrum is shown,Nthe number of the varieties is shown as follows,
is the mean value of the near infrared spectrum,
is the standard deviation of the near infrared spectrum.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, provided by the invention, the attention mechanism network model is obtained through the following steps:
preprocessing the acquired spectral data set sample to obtain a spectral data set sample to be identified;
inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value;
calculating a loss function according to the sample predicted value and the sample true value of the spectral data set sample;
and updating parameters of the initial attention mechanism network model according to the loss function and a random gradient descent strategy to obtain the attention mechanism network model.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, the initial attention mechanism network model comprises a convolution layer, an attention module and a full connection layer, wherein the attention module comprises a channel attention module and a space attention module;
inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value, wherein the method comprises the following steps:
inputting the spectral data set sample to be identified into the convolution layer to obtain a spectral convolution characteristic;
performing first feature extraction on the spectrum convolution features based on the channel attention module, and summing the first feature extraction results to obtain spectrum channel features;
performing second feature extraction on the spectral channel features based on the spatial attention module, and performing splicing and convolution dimensionality reduction on second feature extraction results to obtain spectral spatial features;
and inputting the spectral space characteristics into the full-connection layer to obtain the predicted value of the sample.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, the loss function is calculated according to the sample predicted value and the sample true value of the spectrum data set sample, and the method comprises the following steps:
the loss function is determined by the following equation:
wherein, the first and the second end of the pipe are connected with each other,lossthe function of the loss is represented by,
is as followsiThe actual value of each sample was determined,
is as followsiAnd (4) predicting the value of each sample.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, provided by the invention, the parameters of the initial attention mechanism network model are updated according to the loss function and the random gradient descent strategy to obtain the attention mechanism network model, and the method comprises the following steps:
determining the initial attention mechanism network model as an attention mechanism network model under the condition that the loss function meets a preset condition;
and under the condition that the loss function does not meet the preset condition, updating the model parameters of the initial attention mechanism network model according to a random gradient descent strategy, and returning to re-execute the step of inputting the spectral data set sample to be identified into the initial attention mechanism network model.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, provided by the invention, the updating of the model parameters of the initial attention mechanism network model according to the stochastic gradient descent strategy comprises the following steps:
updating model parameters of the initial attention mechanism network model by the following formula:
wherein the content of the first and second substances,tin order to be able to perform the number of iterations,Wfor the model parameters of the initial attention mechanism network model,
the number of iterations istThe parameters that are updated at the time of the day,
the number of iterations istThe learning rate at the time of the day,
in order to be a function of the cost,
representing a randomly selected one of the gradient directions,Xrepresenting a sample of the spectral dataset to be identified.
The invention also provides a variety identification system based on the near infrared spectrum and the attention mechanism network, which comprises the following components:
the acquisition and processing unit is used for preprocessing the acquired initial near infrared spectrum of the crop to be identified to obtain the near infrared spectrum to be identified;
the prediction unit is used for inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction and identification to obtain an authenticity prediction result of variety identification;
and the training unit is used for training the initial attention mechanism network model through a training set to obtain the attention mechanism network model.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the obtaining and processing unit is specifically used for determining the near infrared spectrum mean value and the near infrared spectrum standard deviation according to the initial near infrared spectrum; and standardizing the initial near infrared spectrum according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the training unit is specifically used for determining the near infrared spectrum mean value and the near infrared spectrum standard deviation through the following formula:
wherein the content of the first and second substances,
is the second in the initial near infrared spectrumiThe first of each varietyjThe data of the bar spectrum is shown,Nthe number of the varieties is shown as follows,
is the mean value of the near infrared spectrum,
is the standard deviation of the near infrared spectrum.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the training unit is specifically used for preprocessing the acquired spectral data set sample to obtain a spectral data set sample to be identified; inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value; calculating a loss function according to the sample predicted value and the sample true value of the spectral data set sample; and updating parameters of the initial attention mechanism network model according to the loss function and a random gradient descent strategy to obtain the attention mechanism network model.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the initial attention mechanism network model comprises a convolution layer, an attention module and a full connection layer, wherein the attention module comprises a channel attention module and a space attention module;
the training unit is specifically used for inputting the spectral data set sample to be identified into the convolution layer to obtain a spectral convolution characteristic; performing first feature extraction on the spectrum convolution features based on the channel attention module, and summing the first feature extraction results to obtain spectrum channel features; performing second feature extraction on the spectral channel features based on the spatial attention module, and performing splicing and convolution dimensionality reduction on second feature extraction results to obtain spectral spatial features; and inputting the spectral space characteristics into the full-connection layer to obtain the predicted value of the sample.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the training unit is specifically used for determining the loss function through the following formula:
wherein the content of the first and second substances,lossthe function of the loss is represented by,
is as followsiThe actual value of each sample was determined,
is as followsiAnd (4) predicting the value of each sample.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the training unit is specifically used for determining the initial attention mechanism network model as the attention mechanism network model under the condition that the loss function meets the preset condition; and under the condition that the loss function does not meet the preset condition, updating the model parameters of the initial attention mechanism network model according to a random gradient descent strategy, and returning to re-execute the step of inputting the spectral data set sample to be identified into the initial attention mechanism network model.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the training unit is specifically used for updating the model parameters of the initial attention mechanism network model through the following formula:
wherein the content of the first and second substances,tin order to be able to perform the number of iterations,Wto model parameters of the initial attention mechanism network model,
the number of iterations istThe parameters that are updated at the time of the update,
is the number of iterations oftThe learning rate at the time of the day,
in order to be a function of the cost,
representing a randomly selected one of the gradient directions,Xrepresenting a sample of the spectral dataset to be identified.
According to the variety identification method and system based on the near infrared spectrum and the attention mechanism network, the near infrared spectrum to be identified is obtained by preprocessing the obtained initial near infrared spectrum of the crop to be identified; and inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction identification to obtain an authenticity prediction result of variety identification. By inputting the near infrared spectrum to be identified into the attention mechanism network model, the authenticity of a plurality of crop varieties can be rapidly and accurately judged, the accuracy of identifying and identifying the crop seed varieties is improved, and simple, convenient and efficient crop variety classification is realized.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a variety identification method based on near infrared spectroscopy and an attention mechanism network provided by the invention.
FIG. 2 is a schematic flow chart of attention mechanism network training provided by the present invention.
Fig. 3 is a schematic diagram of an initial attention mechanism network provided by the present invention.
Fig. 4 is a schematic structural diagram of an attention module provided in the present invention.
FIG. 5 is a schematic diagram of an architecture of a near infrared spectroscopy and attention mechanism network-based variety identification system provided by the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a variety identification method based on a near infrared spectrum and an attention mechanism network, which comprises the following steps of:
and S11, preprocessing the obtained initial near infrared spectrum of the crop to be identified to obtain the near infrared spectrum to be identified.
In particular, the crop to be identified may include, but is not limited to, seeds of single or multi-grain rice, wheat, corn, or progeny thereof.
In one example, taking rice as an example, 21 mature and full rice single-grain samples with known variety types are collected, and each variety adopts 100 seeds as a sample and carries out spectrum collection in a near-infrared high-throughput spectrum collection system. The collection range of the spectrometer is 1100-2500nm, and the collection gate width is 1 ms. Spectra are collected once for each seed, and 2100 pieces of rice near infrared spectrum data are obtained as initial near infrared spectra. Randomly taking 80% of each variety as a training set and 20% as a testing set. The training set is used for constructing the model, and the testing set is used for verifying the prediction effect of the model.
And preprocessing the acquired initial near infrared spectrum before verifying the model or performing authenticity prediction identification by using the model, wherein the preprocessing comprises but is not limited to screening or normalization and other operations, and the preprocessed initial near infrared spectrum is used for obtaining the near infrared spectrum to be identified, which is convenient for the attention mechanism network model to perform authenticity prediction identification.
And S12, inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction and identification to obtain an authenticity prediction result of variety identification.
Wherein the attention mechanism network model is obtained by training an initial attention mechanism network model through a training set.
In the embodiment of the invention, the near infrared spectrum to be identified is obtained by preprocessing the obtained initial near infrared spectrum of the crop to be identified; and inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction identification to obtain an authenticity prediction result of variety identification. By inputting the near infrared spectrum to be identified into the attention mechanism network model, the authenticity of a plurality of crop varieties can be rapidly and accurately judged, the accuracy of identifying and identifying the crop seed varieties is improved, and simple, convenient and efficient crop variety classification is realized.
According to the method for identifying the variety based on the near infrared spectrum and the attention mechanism network, the step S11 comprises the following steps:
and S111, determining a near infrared spectrum mean value and a near infrared spectrum standard deviation according to the initial near infrared spectrum.
S112, standardizing the initial near infrared spectrum according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified.
Specifically, the near infrared spectrum mean and the near infrared spectrum standard deviation may be determined from the initial near infrared spectrum, and the initial near infrared spectrum may be normalized based on the near infrared spectrum mean and the near infrared spectrum standard deviation.
Further, the near infrared spectrum mean value and the near infrared spectrum standard deviation are determined by the following formula 1 and formula 2:
wherein the content of the first and second substances,
is the second in the initial near infrared spectrumiThe first of each varietyjThe data of the bar spectrum is shown,Nthe number of the varieties is shown as follows,
is the mean value of the near infrared spectrum,
is the standard deviation of the near infrared spectrum.
In one example, Z-score normalization may be used to calculate the mean and standard deviation of the near infrared spectrum from multiple spectral data for different species in the initial near infrared spectrum. And calculating the difference value between the spectral data and the near infrared spectrum mean value of the spectral data of the initial near infrared spectrum, and taking the quotient of the difference value and the near infrared spectrum standard deviation as the near infrared spectrum to be identified.
In the embodiment of the invention, the near infrared spectrum mean value and the near infrared spectrum standard deviation are calculated through the initial near infrared spectrum, and the initial near infrared spectrum is standardized according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified. The standardized near infrared spectrum to be identified has a simple structure, so that the near infrared spectrum to be identified can be conveniently identified and identified subsequently through the attention mechanism network model, the difficulty of identifying and identifying the attention mechanism network model is reduced, and the time cost of identifying and identifying is reduced.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, as shown in fig. 2, the attention mechanism network model is obtained through the following steps:
and S21, preprocessing the acquired spectral data set sample to obtain a spectral data set sample to be identified.
Specifically, the acquired spectral dataset sample may be preprocessed to obtain a spectral dataset sample to be identified.
In an example, step S21 may be performed by preprocessing the acquired spectral data set sample as described in steps S11 and S111-S112, so as to obtain a to-be-identified spectral data set sample, which is not described herein again.
And S22, inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value.
Specifically, the initial attention mechanism network may be constructed according to actual needs, and in a preferred example, as shown in fig. 3, the initial attention mechanism network may include an input layer, a convolutional layer, an attention module, 3 fully-connected layers, and an output layer, where the convolutional layer has a size of 1 × 3 × 8, a BN (Batch Normalization) is performed after the convolutional layer, and the attention module passes through a Relu (rectified linear unit) activation function, where the attention module may include a spatial attention module and a channel attention module, and inputs the fully-connected layers. The total connecting layer has 3 layers, and the parameters are respectively 100, 50 and 25. And inputting the spectral data set sample to be identified into an initial attention mechanism network, sequentially passing through the structure, and finally outputting a classification result to obtain a sample prediction value.
And S23, calculating a loss function according to the predicted value of the sample and the true value of the sample of the spectral data set.
Specifically, the variety of the spectral data set sample subjected to variety prediction identification through the initial attention mechanism network model is represented by the sample prediction value, the real variety of the spectral data set sample is represented by the sample real value, and the loss function can be calculated according to the obtained sample prediction value and the sample real value pre-labeled by the spectral data set sample.
Further, for step S23, the loss function may be determined by equation 3:
wherein the content of the first and second substances,lossthe function of the loss is expressed as,
is a firstiThe actual value of each sample was determined,
is as followsiAnd (4) predicting the value of each sample.
And S24, updating the parameters of the initial attention mechanism network model according to the loss function and the random gradient descent strategy to obtain the attention mechanism network model.
Further, step S24 includes S241-S242.
And S241, under the condition that the loss function meets a preset condition, determining the initial attention mechanism network model as an attention mechanism network model.
Specifically, the preset condition may be set according to actual needs, and in one example, the training may be ended when the loss function reaches the preset threshold, and the initial attention mechanism network model is determined as the attention mechanism network model. In another example, the initial attention mechanism network model may be determined as the attention mechanism network model after obtaining the loss function for the specified training round.
And S242, under the condition that the loss function does not meet the preset condition, updating the model parameters of the initial attention mechanism network model according to a random gradient descent strategy, and returning to re-execute the step of inputting the spectral data set sample to be identified into the initial attention mechanism network model.
Further, for step S242, the model parameters of the initial attention mechanism network model may be updated by equation 4 and equation 5.
Wherein the content of the first and second substances,tin order to be the number of iterations,Wto model parameters of the initial attention mechanism network model,
the number of iterations istThe parameters that are updated at the time of the update,
the number of iterations istThe learning rate at the time of the day,
in the form of a cost function, the cost function,
representing a randomly selected one of the gradient directions,Xrepresenting a sample of the spectral dataset to be identified.
In the embodiment of the invention, the spectral data set sample to be identified after the spectral data set sample is preprocessed is input into the initial attention mechanism network model, so that the initial attention mechanism network model is convenient to identify, predict and recognize the input data to obtain the sample predicted value. Calculating a loss function according to the predicted value of the sample and the true value of the sample of the spectral data set to be identified, representing the identification, prediction and identification capability of the model through the loss function, and updating the parameters of the initial attention mechanism network model according to the loss function and a random gradient descent strategy, so that the finally obtained attention mechanism network model has good identification, prediction and identification capability.
According to the variety identification method based on the near infrared spectrum and the attention mechanism network, the initial attention mechanism network model comprises a convolution layer, an attention module and a full connection layer, as shown in fig. 4, wherein the attention module comprises a channel attention module and a space attention module;
step S22 includes:
s221, inputting the spectral data set sample to be identified into the convolution layer to obtain the spectral convolution characteristic.
S222, performing first feature extraction on the spectrum convolution features based on the channel attention module, and determining spectrum channel features based on the first feature extraction result and the spectrum convolution features.
Specifically, in one example, the spectral convolution characteristic C of the convolution layer output is taken as an input to the channel attention module. The spectrum convolution characteristic C of the channel attention module is subjected to global maximum pooling and global average pooling respectively, then two results of the global maximum pooling and the global average pooling are sent to a weight-sharing multilayer perceptron (MLP) respectively, the two output characteristics are subjected to summation operation, first characteristic extraction is completed through a sigmoid (an activation function) activation function, a first characteristic extraction result M _ C is output, and the first characteristic extraction result M _ C is multiplied by the spectrum convolution characteristic C to output the spectrum channel characteristic M _ C.
And S223, performing second feature extraction on the spectral channel features based on the spatial attention module, and determining spectral spatial features based on second feature extraction results and the spectral channel features.
Specifically, in the previous example, the spectral channel feature M _ C output by the channel attention module is used as the input of the spatial attention module, the spectral channel feature M _ C is subjected to global maximum pooling and global average pooling respectively, the features subjected to global maximum pooling and the features subjected to global average pooling are subjected to channel splicing, dimensionality reduction is performed through convolution, second feature extraction is completed through a sigmoid activation function, a second feature extraction result M _ S is output, and the second feature extraction result M _ S is multiplied by the spectral channel feature M _ C to output a spectral spatial feature M _ S.
S224, inputting the spectral space characteristics into the full-connection layer to obtain the predicted value of the sample.
In the embodiment of the invention, the spectral convolution characteristics are obtained by extracting the characteristics of the spectral data set sample to be identified through the convolution layer. And performing first feature extraction on the spectrum convolution features based on the channel attention module, and determining spectrum channel features based on the first feature extraction result and the spectrum convolution features. And performing second feature extraction on the spectral channel features based on the spatial attention module, and determining spectral spatial features based on second feature extraction results and the spectral channel features. And inputting the spectral space characteristics into the full-connection layer to obtain a predicted value of the sample. The attention mechanism network can be identified, predicted and identified on the basis of the characteristics of more channels, spatial information and spectra through the spectrum convolution characteristics, the spectrum channel characteristics and the spectrum spatial characteristics, and more accurate sample prediction values are obtained.
In one embodiment of the invention, as a comparison, a single-layer 1D-CNN network model and a CBAM _ CNN network model added with an attention module are used as comparative construction models to predict a test set sample.
And comparing the predicted value and the true value of the varieties of the two model prediction test sets, and evaluating the recognition performance of the models.
The recognition performance of the model is evaluated by Accuracy (ACC), Precision (PRE), Recall (REC) and F1 scores, and the calculation formula is as follows:
the Accuracy (ACC) calculation formula is:
the accuracy ratio (PRE) is calculated as:
the recall Ratio (REC) is calculated by the formula:
wherein the content of the first and second substances,
the number of target variety grains correctly judged by the model;
the number of seeds of the target variety which are wrongly judged as seeds of the non-target variety by the model;
misjudging the number of seeds of the non-target variety as seeds of the target variety by the model;
the number of seeds of the non-target variety correctly judged by the model;
the score is a harmonic mean between accuracy and recall.
The accuracy rate after 500epoch of the 1D _ CNN network training is 90.24%, the accuracy rate of the network CBAM _ CNN after the attention module is added is 94.05%, the accuracy rate is improved by 3.81% compared with the accuracy rate of the 1D _ CNN network, and the F1 parameter is also improved by 3.7%, so that the model effect is good, and the effectiveness of the embodiment is proved. The specific model evaluation is shown in table 1.
TABLE 1 evaluation table of prediction result model
The variety identification system based on the near infrared spectrum and the attention mechanism network provided by the invention is described below, and the variety identification system based on the near infrared spectrum and the attention mechanism network described below and the variety identification method based on the near infrared spectrum and the attention mechanism network described above can be correspondingly referred to.
The invention also provides a variety identification system based on near infrared spectrum and attention mechanism network, as shown in fig. 5, comprising:
the acquisition and processing unit 51 is used for preprocessing the acquired initial near infrared spectrum of the crop to be identified to obtain the near infrared spectrum to be identified;
the prediction unit 52 is used for inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction and identification to obtain an authenticity prediction result of variety identification;
and the training unit 53 is configured to train the initial attention mechanism network model through a training set to obtain the attention mechanism network model.
In the embodiment of the invention, the near infrared spectrum to be identified is obtained by preprocessing the obtained initial near infrared spectrum of the crop to be identified; and inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction identification to obtain an authenticity prediction result of variety identification. By inputting the near infrared spectrum to be identified into the attention mechanism network model, the authenticity of a plurality of crop varieties can be rapidly and accurately judged, the accuracy of identifying and identifying the crop seed varieties is improved, and simple, convenient and efficient crop variety classification is realized.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network provided by the invention, the obtaining and processing unit 51 is specifically used for determining the near infrared spectrum mean value and the near infrared spectrum standard deviation according to the initial near infrared spectrum; and standardizing the initial near infrared spectrum according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network provided by the invention, the training unit 53 is specifically used for determining the near infrared spectrum mean value and the near infrared spectrum standard deviation through a formula 1 and a formula 2.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network, the training unit 53 is specifically used for preprocessing the acquired spectral data set sample to obtain a spectral data set sample to be identified; inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value; calculating a loss function according to the sample predicted value and the sample true value of the spectral data set sample; and updating parameters of the initial attention mechanism network model according to the loss function and a random gradient descent strategy to obtain the attention mechanism network model.
According to the variety identification system based on the near infrared spectrum and the attention system network, the initial attention system network model comprises a convolution layer, an attention module and a full connection layer, wherein the attention module comprises a channel attention module and a space attention module;
the training unit 53 is specifically configured to input the spectral data set sample to be identified into the convolution layer to obtain a spectral convolution characteristic; performing first feature extraction on the spectrum convolution features based on the channel attention module, and determining spectrum channel features based on a first feature extraction result and the spectrum convolution features; performing second feature extraction on the spectral channel features based on the spatial attention module, and determining spectral spatial features based on second feature extraction results and the spectral channel features; and inputting the spectral space characteristics into the full-connection layer to obtain the predicted value of the sample.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network provided by the invention, the training unit 53 is specifically used for determining the loss function through the formula 3.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network provided by the invention, the training unit 53 is specifically configured to determine the initial attention mechanism network model as the attention mechanism network model under the condition that the loss function meets the preset condition; and under the condition that the loss function does not meet the preset condition, updating the model parameters of the initial attention mechanism network model according to a random gradient descent strategy, and returning to re-execute the step of inputting the spectral data set sample to be identified into the initial attention mechanism network model.
According to the variety identification system based on the near infrared spectrum and the attention mechanism network provided by the invention, the training unit 53 is specifically used for updating the model parameters of the initial attention mechanism network model through a formula 4 and a formula 5.
The above-described system embodiments are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (12)
1. A variety identification method based on a near infrared spectrum and an attention mechanism network is characterized by comprising the following steps:
preprocessing the obtained initial near infrared spectrum of the crop to be identified to obtain a near infrared spectrum to be identified;
inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction identification to obtain an authenticity prediction result of variety identification;
the attention mechanism network model is obtained by training an initial attention mechanism network model through a spectrum data set sample;
the attention mechanism network model is obtained through the following steps: preprocessing the acquired spectral data set sample to obtain a spectral data set sample to be identified; inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value; calculating a loss function according to the predicted value of the sample and the actual value of the sample of the spectral data set; updating parameters of the initial attention mechanism network model according to the loss function and a random gradient descent strategy to obtain the attention mechanism network model;
the initial attention mechanism network model comprises a convolutional layer, an attention module and a full connection layer, wherein the attention module comprises a channel attention module and a space attention module;
inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample prediction value, wherein the method comprises the following steps: inputting the spectral data set sample to be identified into the convolution layer to obtain a spectral convolution characteristic; performing first feature extraction on the spectrum convolution features based on the channel attention module, and determining spectrum channel features based on a first feature extraction result and the spectrum convolution features; performing second feature extraction on the spectral channel features based on the spatial attention module, and determining spectral spatial features based on second feature extraction results and the spectral channel features; inputting the spectral space characteristics into the full-connection layer to obtain the predicted value of the sample;
the performing, by the channel attention module, a first feature extraction on the spectrum convolution feature, and determining a spectrum channel feature based on a first feature extraction result and the spectrum convolution feature includes: performing global maximum pooling and global average pooling on the spectrum convolution characteristics respectively, inputting results of the global maximum pooling and the global average pooling into a multilayer perceptron with shared weight respectively, summing outputs of the multilayer perceptrons, completing first characteristic extraction on the summed result through an activation function, and outputting a first characteristic extraction result; multiplying the first feature extraction result and the spectrum convolution feature to output a spectrum channel feature;
the second feature extraction is performed on the spectral channel features by the spatial attention module, and the spectral spatial features are determined based on the second feature extraction result and the spectral channel features, including: respectively carrying out global maximum pooling and global average pooling on the spectral channel characteristics, carrying out channel splicing on the characteristics subjected to global maximum pooling and the results subjected to global average pooling, carrying out dimension reduction through convolution, carrying out second characteristic extraction on the dimension reduction results through an activation function, and outputting second characteristic extraction results; and multiplying the second feature extraction result by the spectral channel feature to output the spectral space feature.
2. The method for variety identification based on nir spectroscopy and attention mechanism network of claim 1, wherein the pre-processing of the obtained nir spectrum of the crop to be identified to obtain the nir spectrum to be identified comprises:
determining a near infrared spectrum mean value and a near infrared spectrum standard deviation according to the initial near infrared spectrum;
and standardizing the initial near infrared spectrum according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified.
3. The method of claim 2, wherein determining the near infrared spectrum mean and the near infrared spectrum standard deviation from the initial near infrared spectrum comprises:
determining a near infrared spectrum mean and the near infrared spectrum standard deviation by the following formula:
wherein the content of the first and second substances,
is the second in the initial near infrared spectrum
The first of each variety
The data of the bar spectrum is shown,
the number of the varieties is shown as follows,
is the average value of the near infrared spectrum,
is the standard deviation of the near infrared spectrum.
4. The method for breed identification based on near infrared spectroscopy and attention mechanism networks according to claim 1, wherein the calculating a loss function from the predicted value of the sample and the true value of the sample of the spectral dataset comprises:
the loss function is determined by the following equation:
wherein the content of the first and second substances,
the function of the loss is represented by,
is a first
The true value of each of the samples was determined,
is as follows
And (4) predicting the value of each sample.
5. The method for identifying a variety based on a near infrared spectrum and an attention mechanism network as claimed in claim 1, wherein the updating parameters of the initial attention mechanism network model according to the loss function and the stochastic gradient descent strategy to obtain the attention mechanism network model comprises:
determining the initial attention mechanism network model as an attention mechanism network model under the condition that the loss function meets a preset condition;
and under the condition that the loss function does not meet the preset condition, updating the model parameters of the initial attention mechanism network model according to a random gradient descent strategy, and returning to re-execute the step of inputting the spectral data set sample to be identified into the initial attention mechanism network model.
6. The method for identifying a variety based on NIR spectroscopy and an attention mechanism network as claimed in claim 5, wherein the updating the model parameters of the initial attention mechanism network model according to a stochastic gradient descent strategy comprises:
updating model parameters of the initial attention mechanism network model by the following formula:
wherein the content of the first and second substances,
in order to be able to perform the number of iterations,
for the model parameters of the initial attention mechanism network model,
the number of iterations is
The parameters that are updated at the time of the update,
the learning rate when the iteration number is t,
in order to be a function of the cost,
representing a randomly selected one of the gradient directions,
representing a sample of the spectral dataset to be identified.
7. A kind of variety appraisal system based on near infrared spectroscopy and attention mechanism network, characterized by that, including:
the acquisition and processing unit is used for preprocessing the acquired initial near infrared spectrum of the crop to be identified to obtain the near infrared spectrum to be identified;
the prediction unit is used for inputting the near infrared spectrum to be identified into an attention mechanism network model for authenticity prediction and identification to obtain an authenticity prediction result of variety identification;
the training unit is used for training an initial attention mechanism network model through a training set to obtain the attention mechanism network model;
the training unit is specifically used for preprocessing the acquired spectral data set sample to obtain a spectral data set sample to be identified; inputting the spectral data set sample to be identified into an initial attention mechanism network model to obtain a sample predicted value; calculating a loss function according to the sample predicted value and the sample true value of the spectral data set sample; updating parameters of the initial attention mechanism network model according to the loss function and a random gradient descent strategy to obtain the attention mechanism network model;
the initial attention mechanism network model comprises a convolutional layer, an attention module and a fully connected layer, wherein the attention module comprises a channel attention module and a spatial attention module;
the training unit is specifically used for inputting the spectral data set sample to be identified into the convolution layer to obtain a spectral convolution characteristic; performing first feature extraction on the spectrum convolution features based on the channel attention module, and determining spectrum channel features based on a first feature extraction result and the spectrum convolution features; performing second feature extraction on the spectral channel features based on the spatial attention module, and determining spectral spatial features based on second feature extraction results and the spectral channel features; inputting the spectral space characteristics into the full-connection layer to obtain the predicted value of the sample;
the performing, by the channel attention module, a first feature extraction on the spectrum convolution feature, and determining a spectrum channel feature based on a first feature extraction result and the spectrum convolution feature includes: performing global maximum pooling and global average pooling on the spectrum convolution characteristics respectively, inputting results of the global maximum pooling and the global average pooling into a multilayer perceptron with shared weight respectively, summing outputs of the multilayer perceptrons, completing first characteristic extraction on the summed result through an activation function, and outputting a first characteristic extraction result; multiplying the first feature extraction result by the spectrum convolution feature to output a spectrum channel feature;
the second feature extraction is performed on the spectral channel features by the spatial attention module, and the spectral spatial features are determined based on the second feature extraction result and the spectral channel features, including: respectively carrying out global maximum pooling and global average pooling on the spectral channel characteristics, carrying out channel splicing on the characteristics subjected to global maximum pooling and the results subjected to global average pooling, carrying out dimension reduction through convolution, carrying out second characteristic extraction on the dimension reduction results through an activation function, and outputting second characteristic extraction results; and multiplying the second feature extraction result by the spectral channel feature to output the spectral space feature.
8. The near infrared spectroscopy and attention mechanism network-based variety identification system of claim 7,
the acquisition and processing unit is specifically used for determining a near infrared spectrum mean value and a near infrared spectrum standard deviation according to the initial near infrared spectrum; and standardizing the initial near infrared spectrum according to the near infrared spectrum mean value and the near infrared spectrum standard deviation to obtain the near infrared spectrum to be identified.
9. The near infrared spectroscopy and attention mechanism network-based variety identification system of claim 8,
the training unit is specifically configured to determine the near infrared spectrum mean value and the near infrared spectrum standard deviation by the following formula:
wherein the content of the first and second substances,
is the second in the initial near infrared spectrumiThe first of each variety
The data of the bar spectrum is shown,Nthe number of the varieties is shown as follows,
is the mean value of the near infrared spectrum,
is the standard deviation of the near infrared spectrum.
10. The near infrared spectroscopy and attention mechanism network based variety identification system of claim 7,
the training unit is specifically configured to determine the loss function by the following formula:
wherein the content of the first and second substances,
the function of the loss is expressed as,
is as followsiThe actual value of each sample was determined,
is as followsiAnd (4) predicting the value of each sample.
11. The near infrared spectroscopy and attention mechanism network based variety identification system of claim 7,
the training unit is specifically configured to determine the initial attention mechanism network model as an attention mechanism network model when the loss function satisfies a preset condition; and under the condition that the loss function does not meet the preset condition, updating the model parameters of the initial attention mechanism network model according to a random gradient descent strategy, and returning to re-execute the step of inputting the spectral data set sample to be identified into the initial attention mechanism network model.
12. The near infrared spectroscopy and attention mechanism network-based variety identification system of claim 11,
the training unit is specifically configured to update the model parameters of the initial attention mechanism network model by using the following formula:
wherein the content of the first and second substances,
in order to be able to perform the number of iterations,
to model parameters of the initial attention mechanism network model,
the number of iterations is
The parameters that are updated at the time of the update,
is the learning rate when the number of iterations is t,
in order to be a function of the cost,
representing a randomly selected one of the gradient directions,
representing a sample of the spectral dataset to be identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210678616.9A CN114778485B (en) | 2022-06-16 | 2022-06-16 | Variety identification method and system based on near infrared spectrum and attention mechanism network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210678616.9A CN114778485B (en) | 2022-06-16 | 2022-06-16 | Variety identification method and system based on near infrared spectrum and attention mechanism network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114778485A CN114778485A (en) | 2022-07-22 |
| CN114778485B true CN114778485B (en) | 2022-09-06 |
Family
ID=82422016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210678616.9A Active CN114778485B (en) | 2022-06-16 | 2022-06-16 | Variety identification method and system based on near infrared spectrum and attention mechanism network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114778485B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115201144B (en) * | 2022-09-14 | 2022-12-09 | 武汉工程大学 | Quantitative detection method, system and medium for amino acid and protein of rapeseed |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113011499A (en) * | 2021-03-22 | 2021-06-22 | 安徽大学 | Hyperspectral remote sensing image classification method based on double-attention machine system |
| WO2022073452A1 (en) * | 2020-10-07 | 2022-04-14 | 武汉大学 | Hyperspectral remote sensing image classification method based on self-attention context network |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110751A (en) * | 2019-03-31 | 2019-08-09 | 华南理工大学 | A kind of Chinese herbal medicine recognition methods of the pyramid network based on attention mechanism |
| US20220246244A1 (en) * | 2019-06-07 | 2022-08-04 | Leica Microsystems Cms Gmbh | A system and method for training machine-learning algorithms for processing biology-related data, a microscope and a trained machine learning algorithm |
| CN111985543B (en) * | 2020-08-06 | 2024-05-10 | 西北大学 | Construction method, classification method and system of hyperspectral image classification model |
| CN114062305B (en) * | 2021-10-15 | 2024-01-26 | 中国科学院合肥物质科学研究院 | Single grain variety identification method and system based on near infrared spectrum and 1D-In-Resnet network |
| CN114112984B (en) * | 2021-10-25 | 2022-09-20 | 上海布眼人工智能科技有限公司 | Fabric fiber component qualitative method based on self-attention |
-
2022
- 2022-06-16 CN CN202210678616.9A patent/CN114778485B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022073452A1 (en) * | 2020-10-07 | 2022-04-14 | 武汉大学 | Hyperspectral remote sensing image classification method based on self-attention context network |
| CN113011499A (en) * | 2021-03-22 | 2021-06-22 | 安徽大学 | Hyperspectral remote sensing image classification method based on double-attention machine system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114778485A (en) | 2022-07-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109493287B (en) | Deep learning-based quantitative spectral data analysis processing method | |
| Bendel et al. | Evaluating the suitability of hyper-and multispectral imaging to detect foliar symptoms of the grapevine trunk disease Esca in vineyards | |
| CN112924412B (en) | Single-grain rice variety authenticity distinguishing method and device based on near infrared spectrum | |
| CN110907393B (en) | Method and device for detecting saline-alkali stress degree of plants | |
| CN114778485B (en) | Variety identification method and system based on near infrared spectrum and attention mechanism network | |
| CN107704883A (en) | A kind of sorting technique and system of the grade of magnesite ore | |
| US8145585B2 (en) | Automated methods and systems for the detection and identification of money service business transactions | |
| Ünal et al. | Fruit recognition and classification with deep learning support on embedded system (fruitnet) | |
| CN116204831A (en) | Road-to-ground analysis method based on neural network | |
| CN116168766A (en) | Variety identification method, system and terminal based on ensemble learning | |
| Dantas et al. | Reduction of sampling intensity in forest inventories to estimate the total height of eucalyptus trees | |
| CN114062305B (en) | Single grain variety identification method and system based on near infrared spectrum and 1D-In-Resnet network | |
| CN113159225A (en) | Multivariable industrial process fault classification method | |
| CN112365093A (en) | GRU deep learning-based multi-feature factor red tide prediction model | |
| CN113538021A (en) | Machine learning algorithm for store continuity prediction of shopping mall | |
| CN113889274B (en) | Method and device for constructing risk prediction model of autism spectrum disorder | |
| CN113793217A (en) | Stock exchange inversion point and abnormal point detection method based on convolutional neural network | |
| CN114357855A (en) | Structural damage identification method and device based on parallel convolution neural network | |
| CN111160419A (en) | Electronic transformer data classification prediction method and device based on deep learning | |
| CN112364773A (en) | Hyperspectral target detection method based on L1 regular constraint depth multi-instance learning | |
| CN117929356B (en) | LIBS quantitative analysis method based on Gaussian process regression | |
| CN117077544B (en) | Oil-gas separator outlet pressure prediction method and device and electronic equipment | |
| CN117235624B (en) | Emission data falsification detection method, device and system and storage medium | |
| Custodio et al. | Performance analysis of deep learning architectures in classifying calamansi (citrofortunella microcarpa) | |
| CN115830444A (en) | Rice smut extraction method and system based on hyperspectral data band selection |
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 |