CN112782148B - Method for rapidly identifying Arabica and Robertia coffee beans - Google Patents
Method for rapidly identifying Arabica and Robertia coffee beans Download PDFInfo
- Publication number
- CN112782148B CN112782148B CN202110192020.3A CN202110192020A CN112782148B CN 112782148 B CN112782148 B CN 112782148B CN 202110192020 A CN202110192020 A CN 202110192020A CN 112782148 B CN112782148 B CN 112782148B
- Authority
- CN
- China
- Prior art keywords
- data
- coffee bean
- lstm
- raman spectrum
- samples
- 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
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
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
- G01N21/65—Raman scattering
-
- 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
-
- 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
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Chemical & Material Sciences (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
Abstract
The invention discloses a method for rapidly identifying two coffee beans, namely, an arabica coffee bean and a robusta coffee bean, and belongs to the technical field of food detection.
Description
Technical Field
The invention relates to a coffee bean identification method, in particular to a quick identification method of Arabica and Robertsta coffee beans, and belongs to the technical field of food detection.
Background
Coffee is a tropical plant native to southwestern Eleobia, and its fruits are peeled, fermented, degummed, and dried to produce raw coffee. The two main coffee bean varieties are Arabica and Robertsta in the current coffee trading market, the appearances of the two coffee beans are similar, but the price difference is large, a plurality of illegal merchants are always secondary and good, and the difference is utilized to earn a riot. Therefore, the fast and accurate nondestructive coffee bean variety identification method has important significance for guaranteeing the rights of consumers.
Laser Raman spectroscopy is a molecular structure characterization technology based on a Raman scattering effect, and the spectral line position, the spectral band intensity and the like of the laser Raman spectroscopy can reflect information such as material components and the like, and the laser Raman spectroscopy is widely applied to various industries. Guo Pengcheng and the like are combined with the laser Raman technology to successfully distinguish the ganoderma lucidum spore oil; wen Danhua and the like establish a Shanxi mature vinegar age Raman spectrum rapid detection method; corvucci et al successfully used the laser Raman spectrum of honey in conjunction with principal component analysis models for honey production tracing. In addition, in the field of agricultural product detection, the laser Raman technology has been applied to qualitative and quantitative detection of capsicum sudan red, pesticide residues of vegetables and fruits, clenbuterol and the like in fresh meat.
Deep learning has strong learning ability by combining low-level features to form a more abstract high-level representation to discover attribute classes of data. Recurrent Neural Networks (RNNs) are important branches of deep learning, which can create a cycle in a network graph by adding extra weight to the network, and have demonstrated excellent feature extraction and learning capabilities in many fields such as fault diagnosis, speech recognition, emotion classification, and the like. Deep learning has also attracted attention in the field of spectral analysis in recent years, le and the like successfully screen out the rice containing the aflatoxin by combining the near-infrared technology with the deep learning; zhao Yong and the like perform Raman spectrum classification of three types of estrogen powder by using a one-dimensional convolutional neural network; a bidirectional long-term and short-term memory network is applied to the field of terahertz spectrum identification by Yuhaoyue and the like.
Currently, pawel et al successfully distinguished arabica from robusta by ion mobility of two coffee bean aromas, but the detection speed was slower. Marie et al used DNA detection techniques to distinguish between the two coffee beans, but the cost involved was significant.
The method combines the Raman spectrum with the recurrent neural network (LSTM) algorithm, is applied to the type identification of two coffee beans, and has the advantages of no damage to samples, rapidness, low cost and the like.
Disclosure of Invention
The technical problem to be solved by the invention is to provide the rapid identification method of the Arabica coffee beans and the Roberta coffee beans, which has the advantages of rapid identification, low cost, obviously higher noise resistance than the traditional classification algorithm and sample nondestructiveness.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a rapid identification method for two coffee beans, namely, arabic card coffee bean samples and Robert coffee bean samples, is characterized in that the phenomenon that the two coffee bean samples, namely, the Arabic card coffee bean samples and the Robert coffee bean samples, are obviously absorbed in a Raman spectrum wave band is utilized, original data of a Raman spectrum are obtained, a base line is deducted from the original data, smoothing filtering is carried out on the original data, gaussian white noise is added for expanding the data, the expanded data of each coffee bean is placed into an LSTM neural network for training and testing, an LSTM neural network model is built, optimal parameters are found through optimization, and finally classification results of the coffee bean samples are obtained.
The technical scheme of the invention is further improved as follows: the method comprises the following steps:
1) Pretreatment of a coffee bean sample: screening coffee bean samples, and airing;
2) Measuring a coffee bean sample by using a Raman spectrometer to obtain original Raman spectrum data, and preprocessing the original Raman spectrum data;
3) Performing data expansion on the preprocessed original Raman spectrum data to expand the spectrum data of each coffee bean sample to 2000 samples;
4) Establishing an LSTM neural network model for carrying out spectrum qualitative classification, determining optimal parameters of the model, and selecting an optimal network structure;
5) And obtaining classification results of two coffee bean samples according to the optimal network structure.
The technical scheme of the invention is further improved as follows: the model of the Raman spectrometer in the step 2) is i-
plus785s, an excitation light source of 785nm, and a spectral range of 175cm -1 ～3200cm -1 Resolution of 4.5cm -1 Maximum power of 300mW, scanning times of 3 times, and scanning resolution of 4cm -1 The integration time is 3000ms, and background noise is deducted in real time during scanning.
The technical scheme of the invention is further improved as follows: the preprocessing in the step 2) is to perform baseline correction and smooth filtering processing on the measured original Raman spectrum data.
The technical scheme of the invention is further improved as follows: in the step 3), gaussian white noise with different parameters is added to the measured original Raman spectrum data, various interference signals in the real spectrum acquisition process are simulated, and meanwhile, the spectrum data of each coffee bean is expanded to 2000 samples.
The technical scheme of the invention is further improved as follows: and (4) disturbing the label and data sequence of the extended Raman spectrum data of the two coffee bean samples.
The technical scheme of the invention is further improved as follows: in the step 4), the expanded spectral data is randomly divided into a training set of 60%, a testing set of 20% and a verification set of 20% to obtain characteristic spectral data, the characteristic spectral data is reasonably divided into LSTM neural units, then the dimensionality of each spectral data is measured, each spectral data is divided at equal intervals and input into an LSTM neural network for iteration, the optimal parameters of the model are determined, the model with the highest classification accuracy is obtained, and then the optimal network structure is obtained.
The technical scheme of the invention is further improved as follows: the determination process for determining the optimal parameters of the model comprises the following steps: and selecting reasonable iteration times, and selecting the optimal spectrum number in the LSTM unit and the number of LSTM neurons through accuracy.
The technical scheme of the invention is further improved as follows: and (5) running for 5 times according to the LSTM optimal network model, and taking the average value of the accuracy of the classification result.
Due to the adoption of the technical scheme, the invention has the technical progress that:
the method utilizes different spectrum data obtained by absorption of Raman spectra by Arabic and Robertia coffee beans, utilizes an LSTM neural network to classify and identify coffee samples, and finally classifies the two coffee bean samples, and the result also proves that the accuracy rate of the coffee bean classification result is reliable. The invention utilizes the LSTM neural network, can process a large number of coffee bean samples, can efficiently and accurately classify the samples, and finally completes the qualitative identification of the coffee bean types.
Drawings
FIG. 1 is a LSTM cell transfer diagram of the present invention;
FIG. 2 is a view of the internal structure of the LSTM of the present invention;
FIG. 3 is a Raman spectrum of an Abiraka and Robert's base line after calibration and smooth filtering in accordance with the present invention;
FIG. 4 is a comparison of different data structures and hidden node number results for the present invention;
FIG. 5 is a graph of the effect of iteration number on accuracy of the present invention;
FIG. 6 is a graph of LSTM accuracy versus loss values for the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following examples:
as shown in fig. 1 and fig. 2, a rapid identification method for two coffee beans, namely, arabica coffee beans and robusta coffee beans, utilizes the phenomenon that two coffee bean samples, namely, arabica coffee beans and robusta coffee beans, have obvious absorption in a raman spectrum band to obtain original data of a raman spectrum, subtracts a baseline from the original data and carries out smooth filtering treatment, then white gaussian noise is added for expanding data, then the expanded data of each coffee bean is put into an LSTM neural network for training and testing, an LSTM neural network model is built, optimal parameters are found by optimizing, and finally classification results of the coffee bean samples are obtained.
The method comprises the following steps:
1) Pretreatment of a coffee bean sample: screening coffee bean samples, and airing;
2) Measuring a coffee bean sample by using a Raman spectrometer to obtain original Raman spectrum data, and preprocessing the original Raman spectrum data;
3) Performing data expansion on the preprocessed original Raman spectrum data to expand the spectrum data of each coffee bean sample to 2000 samples;
4) Establishing an LSTM neural network model for carrying out spectrum qualitative classification, determining optimal parameters of the model, and selecting an optimal network structure;
5) And obtaining classification results of the two coffee bean samples according to the optimal network structure.
Further, the model of the Raman spectrometer in the step 2) is i-
plus785s, an excitation light source of 785nm, and a spectral range of 175cm -1 ～3200cm -1 Resolution of 4.5cm -1 Maximum power of 300mW, scanning times of 3 times, and scanning resolution of 4cm -1 The integration time is 3000ms, and background noise is deducted in real time during scanning.
Further, the preprocessing in the step 2) is to perform baseline correction and smooth filtering processing on the measured raw raman spectrum data.
Further, in the step 3), gaussian white noise with different parameters is added to the measured original raman spectrum data, various interference signals in the real spectrum acquisition process are simulated, and meanwhile, the spectrum data of each coffee bean is expanded to 2000 samples. And (4) disturbing the label and data sequence of the extended Raman spectrum data of the two coffee bean samples.
Further, in step 4), the expanded spectral data is randomly divided into a training set of 60%, a testing set of 20% and a verification set of 20% to obtain characteristic spectral data, the characteristic spectral data is reasonably divided into LSTM neural units, then the dimensionality of each piece of spectral data is measured, each piece of spectral data is divided at equal intervals and input into an LSTM neural network for iteration, the optimal parameters of the model are determined, the model with the highest classification accuracy is obtained, and then the optimal network structure is obtained.
Further, internal input gate i of LSTM t Forgetting door f t Output gate o t Internal hidden state c tx New cell state c t New state h t Is calculated by the formula
i t ＝σ(W i h t-1 +U i x t )；
f t ＝σ(W f h t-1 +U f x t )；
o t ＝σ(W o h t-1 +U o x t )；
c tx ＝tanh(W ct h t-1 +U ct x t )；
c t ＝(c t-1 ×f t )+(c tx ×i t )；
h t ＝tanh(c t )×o t ；
Further, the process of determining the optimal parameters of the model comprises: and selecting reasonable iteration times, and selecting the optimal spectrum number in the LSTM unit and the number of LSTM neurons through accuracy.
Further, the operation is carried out for 5 times according to the LSTM optimal network model, and the average value of the accuracy of the classification result is taken.
Example 1:
based on the identification of the coffee bean types by combining the raman spectrum with the LSTM neural network, the samples adopted in this example were two coffee beans of arabica and robusta.
1) Screening and airing two purchased coffee beans;
2) Measuring 30 Raman spectra of two coffee beans by using a Raman spectrometer;
3) Obtaining raman spectrum average graphs of two coffee beans through baseline correction and smoothing filtering, as shown in fig. 3;
4) Selecting the optimal spectrum number in the LSTM unit and the number of LSTM neurons through accuracy;
5) And obtaining classification results of the two coffee bean samples according to the optimal network structure.
The LSTM neural network is an improved Recurrent Neural Network (RNN), can effectively solve the problem of long-time previous important information loss in an RNN structure, and has unique advantages when being applied to Raman spectrum.
Measuring 30 Raman spectra of each type of coffee beans, wherein the measured Raman spectra are 1 × 1809 data, expanding the Raman spectrum data in a noise adding mode, expanding each type of coffee bean data to 2000, setting the input data size of each unit of the LSTM network to be 201, 67, 27 and 9 respectively because 1809 can be divided by 9, 27, 67 and 201, and carrying out experiments aiming at four data structures; meanwhile, the hidden layer node numbers (neuron numbers) discuss 5 cases of 96, 112, 128, 144 and 160, respectively, and the classification accuracy results of coffee beans are shown in table 1 and are plotted as fig. 4 in different data structures and different hidden layer node numbers.
TABLE 1
As can be seen from fig. 4, when 201 data are input into each LSTM unit, the number of LSTM of each sample spectrum is 9, and the accuracy is the highest, which reaches 92.51%, because the raman spectrum of coffee beans is relatively completely preserved, and more LSTM unit spectrum information can be extracted, thereby improving the accuracy.
The iteration times of deep learning also have important influence on the accuracy, the accuracy changes with the increase of the iteration times, model iteration time parameters are experimentally researched on the basis of determining the data structure and the number of nodes of the hidden layer, and the classification accuracy and program running time results corresponding to different iteration times are shown in fig. 5. It can be seen from fig. 5 that the accuracy rate has a significant upward trend when the number of iterations is 40 to 60, and the accuracy rate is substantially constant when the number of iterations is 60 to 90. The time is prolonged due to the increase of the iteration times, the time factor is considered comprehensively, the iteration times are selected to be 60 times, and the running time is 26.60 seconds.
In summary, the unit type of LSTM is 201 data, the number of nodes in the hidden layer is 128, and when the iteration time is 60 times, the coffee bean classification accuracy is the highest, reaching 92.51%, and the running time is the minimum, and is 26.60 seconds. The results of the accuracy and loss curves of the training results are shown in fig. 6.
In order to verify the classification performance of the proposed LSTM model, the coffee bean Raman spectrum data is processed by adopting classification methods of K Nearest Neighbor (KNN), gradient boosting (Gradient boosting), random forest (Random forest) and Decision tree (Decision tree), and the experimental environment is the same as the building environment of the LSTM. Both the LSTM classification algorithm and the traditional classification algorithm are subjected to baseline correction and smooth filtering in advance, and then classification research is carried out. The invention compares different classification methods based on four indexes of accuracy, precision, recall rate and F1 value, and compares the time used by different methods, and the comparison results of different classification algorithms are shown in Table 2.
TABLE 2
As can be seen from Table 2, although the time used for the LSTM classification method is not the shortest, the accuracy and F1 value of the LSTM algorithm are the highest, which can reach 92.51%, and the F1 values of the two coffee beans are 92.83% and 93.67%, respectively. The experimental results prove that the LSTM method is effective and can be combined with Raman spectrum to be used for identifying the types of two coffee beans.
The Raman spectrum of the coffee beans is inevitably interfered by a plurality of uncontrollables such as environment light and sample positions in the field collection process, so that the strong anti-interference capability of the designed model is required to be higher. Based on the determined network structure and model parameters, gaussian white noise with different intensities is added to the original spectrum, the noise intensities are respectively 65dBW, 70dBW, 75dBW, 80dBW, 85dBW and 90dBW, the anti-noise performance of the proposed LSTM method is evaluated by comparing different classification algorithms, and the result table 3 shows.
TABLE 3
As is clear from table 3, when the additive noise is relatively weak, the accuracy of the various classification methods is not very different; with the increase of the noise intensity, the classification effect of the traditional classification method is greatly reduced, and the LSTM method still maintains higher classification accuracy; when the noise reaches 90dBW, the accuracy of LSTM can reach 84.12%, which is the highest among the five classification algorithms. The results fully show that the LSTM neural network is very suitable for the Raman spectrum accurate classification under the condition of noise interference.
The invention provides a Raman spectrum qualitative classification model based on deep learning, which is used for realizing qualitative identification of two types of coffee beans. Gaussian white noise is added into the spectrum, so that Raman data of original coffee beans are expanded, an LSTM network classification model is built, optimal parameters are determined, LSTM model training and testing are completed, and compared with the traditional method, the classification accuracy and the anti-interference capability of the LSTM method are obviously superior to those of other methods. The experimental result shows that the LSTM neural network is an excellent Raman spectrum qualitative analysis method, has the obvious advantage of strong robustness, and has important application value in the field of Raman spectrum analysis of field acquisition.
While embodiments of the invention have been disclosed above, it is not intended to be limited to the details shown, and described, but is to be accorded the widest scope consistent with the principles and novel features herein disclosed, and to such extent that those skilled in the art may readily devise many other modifications of the present invention without departing from the spirit and scope of the appended claims.
Claims (2)
1. The rapid identification method of the Arabica coffee beans and the Robertia coffee beans is characterized in that: the method comprises the steps of obtaining original data of a Raman spectrum by utilizing the phenomenon that two coffee bean samples of an Arabic card and a Roberta have obvious absorption in a Raman spectrum wave band, deducting a base line from the original data and carrying out smooth filtering treatment, adding the Gaussian white noise for expanding the data, putting the expanded data of each coffee bean into an LSTM neural network for training and testing, building an LSTM neural network model, finding out optimal parameters through optimization, and finally obtaining a classification result of the coffee bean samples;
the method comprises the following steps:
1) Pretreatment of coffee bean samples: screening coffee bean samples, and airing;
2) Measuring coffee bean sample with Raman spectrometer to obtain original Raman spectrum data, and comparingPreprocessing original Raman spectrum data; the model of the Raman spectrometer in the step 2) is i-RAMAN plus785s, the excitation light source is 785nm, and the spectral range is 175cm -1 ～3200cm -1 Resolution of 4.5cm -1 Maximum power of 300mW, scanning times of 3 times, and scanning resolution of 4cm -1 The integration time is 3000ms, and background noise is deducted in real time during scanning;
3) Performing data expansion on the preprocessed original Raman spectrum data to expand the spectrum data of each coffee bean sample to 2000 samples;
4) Establishing an LSTM neural network model for carrying out spectrum qualitative classification, determining optimal parameters of the model, and selecting an optimal network structure, wherein the unit types of the LSTM are 201 data, the number of nodes of a hidden layer is 128, the iteration times are carried out for 60 times, the classification accuracy of coffee beans is highest and reaches 92.51%, and the running time at the moment is minimum and is 26.60 seconds;
5) Obtaining classification results of two coffee bean samples according to the optimal network structure;
the preprocessing in the step 2) is to perform baseline correction and smooth filtering processing on the measured original Raman spectrum data;
in the step 3), gaussian white noise with different parameters is added to the measured original Raman spectrum data, various interference signals in the real spectrum acquisition process are simulated, and meanwhile, the spectrum data of each coffee bean is expanded to 2000 samples;
in the step 4), the expanded spectral data is randomly divided into a training set of 60%, a testing set of 20% and a verification set of 20% to obtain characteristic spectral data, the characteristic spectral data is reasonably divided into LSTM neural units, then the dimensionality of each spectral data is measured, each spectral data is divided at equal intervals and input into an LSTM neural network for iteration, the optimal parameters of the model are determined, the model with the highest classification accuracy is obtained, and then the optimal network structure is obtained;
the determination process of the optimal parameters of the model comprises the following steps: and selecting reasonable iteration times, and selecting the optimal spectrum number in the LSTM unit and the number of LSTM neurons through accuracy.
2. The method for the rapid identification of two coffee beans, arabica and robusta, according to claim 1, characterized in that: and in the step 3), the labels and the data sequence of the extended Raman spectrum data of the two coffee bean samples are disturbed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110192020.3A CN112782148B (en) | 2021-02-19 | 2021-02-19 | Method for rapidly identifying Arabica and Robertia coffee beans |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110192020.3A CN112782148B (en) | 2021-02-19 | 2021-02-19 | Method for rapidly identifying Arabica and Robertia coffee beans |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112782148A CN112782148A (en) | 2021-05-11 |
| CN112782148B true CN112782148B (en) | 2023-01-03 |
Family
ID=75761625
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110192020.3A Active CN112782148B (en) | 2021-02-19 | 2021-02-19 | Method for rapidly identifying Arabica and Robertia coffee beans |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112782148B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113420795B (en) * | 2021-06-07 | 2024-04-26 | 核工业北京地质研究院 | Mineral spectrum classification method based on cavity convolutional neural network |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106706546A (en) * | 2016-12-28 | 2017-05-24 | 中山市腾创贸易有限公司 | Analysis method for artificial intelligence learning materials on basis of infrared and Raman spectrum data |
| CN109142317B (en) * | 2018-08-29 | 2020-05-08 | 厦门大学 | Raman spectrum substance identification method based on random forest model |
| CN109543763B (en) * | 2018-11-28 | 2022-10-21 | 重庆大学 | Raman spectrum analysis method based on convolutional neural network |
| CN110309707A (en) * | 2019-05-08 | 2019-10-08 | 昆明理工大学 | A kind of recognition methods of the coffee drupe maturity based on deep learning |
-
2021
- 2021-02-19 CN CN202110192020.3A patent/CN112782148B/en active Active
Non-Patent Citations (3)
| Title |
|---|
| Discrimination between Arabica and Robusta green coffee using visible micro Raman spectroscopy and chemometric analysis;Rasha M. El-Abassy et al.;《Food Chemistry》;20101126;第126卷;摘要，第2节 * |
| Noise Reduction Technique for Raman Spectrum using Deep Learning Network;Liangrui Pan et al.;《2020 13th International Symposium on Computational Intelligence and Design (ISCID)》;20210120;第160页左栏第2段 * |
| Rapid screening of hepatitis B using Raman spectroscopy and long short-term memory neural network;Xin Wang et al.;《Lasers in Medical Science》;20200413;第35卷;摘要，材料和方法部分，结果部分 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112782148A (en) | 2021-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ma et al. | Rapid and non-destructive seed viability prediction using near-infrared hyperspectral imaging coupled with a deep learning approach | |
| Ji et al. | Non-destructive classification of defective potatoes based on hyperspectral imaging and support vector machine | |
| Feng et al. | Detection of subtle bruises on winter jujube using hyperspectral imaging with pixel-wise deep learning method | |
| Zheng et al. | AFFU-Net: Attention feature fusion U-Net with hybrid loss for winter jujube crack detection | |
| CN108844917A (en) | A kind of Near Infrared Spectroscopy Data Analysis based on significance tests and Partial Least Squares | |
| Zhou et al. | Identification of the variety of maize seeds based on hyperspectral images coupled with convolutional neural networks and subregional voting | |
| Gill et al. | A review of automatic fruit classification using soft computing techniques | |
| CN112782148B (en) | Method for rapidly identifying Arabica and Robertia coffee beans | |
| Wang et al. | Shelf-life prediction of ‘Gros Michel’bananas with different browning levels using hyperspectral reflectance imaging | |
| CN110208211B (en) | Near infrared spectrum noise reduction method for pesticide residue detection | |
| CN108489927A (en) | Fish place of production source tracing method, electronic equipment, storage medium and device | |
| Sheng et al. | Data fusion strategy for rapid prediction of moisture content during drying of black tea based on micro-NIR spectroscopy and machine vision | |
| Zhang et al. | Analysis of E-tongue data for tea classification based on semi-supervised learning of generative adversarial network | |
| Li et al. | Detection storage time of mild bruise’s yellow peaches using the combined hyperspectral imaging and machine learning method | |
| CN114112983A (en) | Python data fusion-based Tibetan medicine all-leaf artemisia rupestris L producing area distinguishing method | |
| Xie et al. | Modeling for mung bean variety classification using visible and near-infrared hyperspectral imaging | |
| Zhou et al. | Green plums surface defect detection based on deep learning methods | |
| CN109685099B (en) | Apple variety distinguishing method based on spectrum band optimization fuzzy clustering | |
| CN112485238A (en) | Method for identifying turmeric essential oil producing area based on Raman spectrum technology | |
| CN112069911A (en) | Fruit and vegetable quality detection method based on multispectral image information and TLMD-WOA-SIFT | |
| CN107886115A (en) | A kind of tealeaves mid-infrared light profile classification method of adaptively possible C mean clusters | |
| CN110567888A (en) | silkworm cocoon dry shell amount nondestructive testing method based on visible/near infrared spectrum and deep learning | |
| CN106442390A (en) | Transgenic soybean classification and identification method based on PCA-SVM algorithm | |
| CN113049526B (en) | Corn seed moisture content determination method based on terahertz attenuated total reflection | |
| Hu et al. | Research on rice seed fullness detection method based on terahertz imaging technology and feature extraction method |
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 |