CN116026795A - Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum - Google Patents
Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum Download PDFInfo
- Publication number
- CN116026795A CN116026795A CN202211681383.4A CN202211681383A CN116026795A CN 116026795 A CN116026795 A CN 116026795A CN 202211681383 A CN202211681383 A CN 202211681383A CN 116026795 A CN116026795 A CN 116026795A
- Authority
- CN
- China
- Prior art keywords
- reflection
- spectrum
- regression
- quality
- 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.)
- Pending
Links
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a rice grain quality character nondestructive prediction method based on reflection and transmission spectrum. Firstly, collecting reflection and transmission spectrum information of rice grains by using a developed integrated hyperspectral reflection and transmission spectrum imaging system, and extracting spectrum data through image processing; secondly, determining quality character parameters of rice grains by using a traditional chemical method; then screening characteristic spectrums highly correlated with the quality traits of rice grains by a correlation analysis, a non-information variable screening method, a competitive self-adaptive re-weighted sampling method and a continuous projection algorithm, modeling by taking a characteristic spectrum set as an independent variable and taking a quality trait artificial value as a dependent variable, wherein the modeling regression method comprises stepwise linear regression, partial least squares regression, support vector machine regression, random forests and CNN-LSTM, and screening an optimal model; and finally, carrying out nondestructive prediction on the quality character of the rice grains by using an optimal model, and excavating the correlation between the reflection spectrum and the transmission spectrum of the rice grains and the quality character.
Description
Technical Field
The invention belongs to the technical fields of agricultural automation and agricultural information, and particularly relates to a rice grain quality character nondestructive prediction method based on reflection and transmission spectrums.
Background
For the measurement of rice grain quality traits, the traditional chemical method generally comprises the processes of shelling, grinding, instrument analysis of component content and the like, is time-consuming and labor-consuming, and the result is easily influenced by manual operation and the like, and belongs to the lossy measurement. With the development of optical imaging technology, computer information technology and the like, the spectrum technology becomes an important way for nondestructive prediction of the quality of agricultural products.
Shon et al (2004) established 3 regression models of polished rice protein content by near infrared transmission spectroscopy; xue Ligong et al (2004) found that rice grain processing quality was significantly inversely related to canopy spectral reflectance spectrum index at maturity, whereas grain protein content was significantly related to canopy spectral ratio index and normalized index at most of the growth period; tang Yanlin et al (2006) found that the rice ear and brown rice crude protein content has good correlation with the canopy spectral reflectivities ρ_lambda and the first derivative spectra d_lambda and the canopy spectra RVI in the grouting period and the lactation period, and that certain wave bands even reach significant and extremely significant correlation levels; zhang Hao et al (2012) research found that the canopy spectrum of the potted rice grain has obvious reflectivity difference at the wavelength of 700-1000 nm; zhang Xiao et al (2021) based on reflectance spectra of rice flour and dry ears found that sensitive wavelet features were effective in improving amylose content prediction accuracy compared to normalized spectral indices. The research analysis of the former shows that the spectral transmission and reflection characteristics can be used for nondestructive prediction of rice grain quality characteristics, but most researches can only predict a small quantity of quality characteristic indexes, the adopted sample size is small, and the applicability of the model is questionable.
Hyperspectral imaging techniques have been widely used in recent years in nondestructive testing because of their high resolution in obtaining spectral information. The data cube acquired by the hyperspectral imaging system not only has spectral data information of each point on the image, but also can be unfolded in the spectral dimension to acquire image information in any spectral band. In addition, the quality characteristics of rice grains are different in reflection results under different wave bands, so that the quality characteristics can be predicted by selecting characteristic wave bands.
The traditional chemical method is time-consuming and labor-consuming to obtain the rice grain quality character parameters, the sample is damaged and the accuracy is not high, the quality character indexes of the current nondestructive prediction through reflection and transmission spectrums are less, and the precision and the applicability of the model are required to be improved. With the development of computer and information science, machine learning and artificial intelligence technology are promoting the transformation of traditional data analysis methods. By combining the information technology and the hyperspectral imaging technology and carrying out mathematical modeling on hyperspectral image data, the quality character parameters can be predicted by using hyperspectrum.
Disclosure of Invention
First, the technical problem to be solved
The traditional chemical method is time-consuming and labor-consuming in obtaining the rice grain quality character parameters, is lossy to a sample and low in accuracy, and the quality character indexes predicted through reflection and transmission spectrums are less, and the model precision and applicability are required to be improved. In order to overcome the problems, the invention provides a rice grain quality character nondestructive prediction method based on reflection and transmission spectra.
(II) technical scheme
In order to solve the technical problems, the invention provides a rice grain quality character nondestructive prediction method based on reflection and transmission spectra. The method comprises the following steps:
step A, an integrated hyperspectral reflection and transmission spectrum imaging system is built, reflection and transmission spectrum images of a plurality of healthy and full rice seeds are collected, and reflection and transmission spectrum data are extracted;
step B, husking and grinding rice grains to obtain rice powder, and obtaining quality character parameters of the rice grains by a traditional chemical method;
c, processing the reflection and transmission spectrum data obtained in the step A to obtain a characteristic spectrum set;
step D, establishing a quality character-rice grain characteristic spectrum regression prediction model by taking the characteristic spectrum set as an independent variable and taking the quality character artificial value as a dependent variable;
and E, screening an optimal model, and carrying out nondestructive prediction on the quality character of the rice grains by using the optimal model.
More specifically, the reflection and transmission spectrum data are extracted in the step A, specifically by the following modes: spectral image information is stored in a system workstation in a binary data file format which is stored in a band line-by-line crossing manner; c++ is adopted to reform image data, and hyperspectral images under each wave band are extracted; dividing hyperspectral images in each wave band in a two-by-two circulation mode, carrying out image segmentation by combining an OTSU algorithm to obtain a binary image, screening out the binary image with the best segmentation effect, and masking the binary image with the hyperspectral images in each wave band to obtain a mask image only containing rice grains; reflection and transmission spectrum data are extracted based on each mask image.
More specifically, the rice grain quality trait parameters in step B include one or more of protein content, amylose content, moisture content, gelatinization temperature and consistency.
More specifically, in the step C, the reflection and transmission spectrum data are processed to obtain a characteristic spectrum set, specifically in the following manner: carrying out smoothing pretreatment on the reflection spectrum data and the transmission spectrum data based on a simple moving average method, an SG convolution smoothing method and a fast Fourier transform method, and then calculating derivative spectrum data of the smoothed spectrum value, wherein the derivative spectrum data comprises a first derivative and a second derivative; carrying out pearson correlation analysis on the smoothed spectrum data and the derivative spectrum data thereof and the quality character artificial value to determine the correlation between the spectrum data and the quality character, further screening to obtain a spectrum data set S1 highly correlated with the quality character, and screening characteristic wavelengths by adopting UVE, CARS and SPA respectively to obtain spectrum data sets S2, S3 and S4 corresponding to the characteristic wavelengths; the intersection of the spectral data sets S1, S2, S3 and S4 is found, and the result obtained is taken as a characteristic spectral set.
More specifically, in the step D, a stepwise linear regression method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, specifically,
step D1, in a characteristic spectrum set, selecting a variable and a candidate set at each time to establish a spectrum-quality character linear regression model together; the candidate set is initially empty;
step D2, finding the model with the minimum AIC score of the erythrocyte information criterion in all the models in the step D1, and selecting the corresponding variable to add into the candidate set;
step D3, repeating the steps D1 and D2 until the AIC score of the erythro pool information criterion is no longer reduced, and finally, the variables in the candidate set are called AIC initial selection variables;
step D4, using the Bayesian information criterion BIC as an evaluation index, and repeating the steps D1-D3 to obtain BIC primary selection variables;
step D5, obtaining an intersection of the AIC primary selection variable and the BIC primary selection variable to obtain a final variable;
step D6, establishing a linear regression model by using the final variable, and sorting the variable variance expansion factor VIF values in a descending order, and sequentially deleting one variable from high to low;
step D7, repeating the step D6 until the variance expansion factor VIF of all variables of the regression model is smaller than 10;
and D8, taking the variables screened in the step D7 as input, and establishing a linear regression model.
More specifically, in the step D, a traditional nonlinear regression method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, specifically, a range (0-N) of a principal component is empirically set for partial least squares regression, and then the optimal principal component size of the partial least squares regression is determined through a Bayes optimization algorithm; the Bayesian optimization algorithm utilizes the information obtained before to optimize the next iteration when each iteration is performed, and the specific mode is as follows: firstly, dividing a test set and a verification set, then initializing the number of principal components to be N/2 to fit a regression model, and calculating the loss on the verification set; then, the number of principal components is determined by continuous iteration of a binary search algorithm, and the loss on the verification set is calculated; after multiple iterations, the regression model with the highest score on the verification set is selected, and the corresponding K is the optimal principal component size.
More specifically, in the step D, a machine learning regression method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, and the specifically adopted method comprises support vector machine regression and random forest regression.
More specifically, in the step D, a deep learning method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, and the specific method is as follows: firstly, extracting the characteristics of a spectrum set by using a convolution layer and a maximum pooling layer of a CNN model, then inputting the characteristics into an LSTM model to obtain the relation between spectrum characteristic sequences, setting an excitation function of an output layer as linear to represent a regression task, optimizing by using a random gradient descent method, taking an average absolute error as an index of a loss function, training 10 groups of data each time, obtaining a result after 100 iterations, and visualizing each iteration result.
More specifically, in step E, the optimal model is selected by evaluating the regression model result according to parameters including the determination coefficient R2 and the root mean square error RMSE of the model.
(III) beneficial effects
The benefits of the present invention are seen in two ways compared to the prior art. On one hand, nondestructive detection of rice grain quality characteristics is realized through a hyperspectral reflection and transmission spectrum imaging system, and the problems of sample damage and time consumption in the traditional chemical method are solved. On the other hand, by preprocessing spectral data, extracting characteristic spectral bands, excavating an important relation between the quality traits of rice grains and the reflection transmission spectrum, establishing a regression prediction model of the quality traits of the rice grains by using a plurality of methods, selecting an optimal model for prediction, effectively improving the robustness, stability and applicability of the model, and solving the problem that the model precision and applicability of the current researcher for the prediction of the rice quality by using the reflection or transmission spectrum are lower. If the traditional chemical method is used for measuring a plurality of quality traits, a plurality of repeated samples are needed to be prepared, the loss is high, and based on the invention, the integrated hyperspectral reflection and transmission spectrum imaging system which is researched and developed only needs to be used for scanning the spectrum information of one part of rice grain, so that the plurality of quality traits can be predicted, the loss of samples is greatly reduced, the throughput of quality detection is improved, and the high-efficiency, accurate and nondestructive prediction of the quality traits of the rice grain is realized.
Drawings
Fig. 1 is a flow chart of the technical scheme of the invention.
Figure 2 is a diagram of a built integrated hyperspectral reflectance and transmittance spectral imaging system.
FIG. 3 is a diagram of a stepwise linear regression variable screening process.
Detailed Description
The invention is further illustrated in the following drawings and examples, which are given solely for the purpose of illustration and are not intended to limit the scope of the invention.
The invention provides a rice grain quality character nondestructive prediction method based on reflection and transmission spectrum in order to solve the technical problem. As shown in figure 1, the invention mainly comprises the following steps:
step A, an integrated hyperspectral reflection and transmission spectrum imaging system is built, reflection and transmission spectrum images of a plurality of healthy and full rice seeds are collected, and reflection and transmission spectrum data are extracted;
step B, husking and grinding rice grains to obtain rice powder, and obtaining quality character parameters of the rice grains by a traditional chemical method;
c, processing the reflection and transmission spectrum data obtained in the step A to obtain a characteristic spectrum set;
step D, establishing a quality character-rice grain characteristic spectrum regression prediction model by taking the characteristic spectrum set as an independent variable and taking the quality character artificial value as a dependent variable;
and E, screening an optimal model, and carrying out nondestructive prediction on the quality character of the rice grains by using the optimal model.
More specifically, the integrated hyperspectral reflection and transmission spectrum imaging system constructed in the step A is shown in the figure 2, and the rice seeds are conveyed to a transmission spectrum imaging center through a PVC transparent conveyor belt and two sensing elements; when the reflection spectrum information is collected, firstly, a transmission spectrum light source is turned off, the reflection spectrum light source is turned on, a system workstation sends an instruction to move a correction white board to a position above a conveyor belt (the length of the correction white board is larger than the width of the conveyor belt, namely, the correction white board is ensured to cover an imaging visual field), then, a hyperspectral camera is sent to move above the correction white board, white board and dark current are collected, then, the correction white board is moved back to an original point, and the reflection spectrum is started to be collected; after the collection of the reflection spectrum is finished, the reflection spectrum light source is turned off, the transmission spectrum light source is turned on, the collection of the transmission spectrum is started, and rice seeds are sent out through the conveyor belt after the collection is finished.
More specifically, the reflection and transmission spectrum data are extracted in the step A, specifically by the following modes: spectral image information is stored in a system workstation in a binary data file format which is stored in a band line-by-line crossing manner; c++ is adopted to reform image data, and hyperspectral images under each wave band are extracted; dividing hyperspectral images in each wave band in a two-by-two circulation mode, carrying out image segmentation by combining an OTSU algorithm to obtain a binary image, screening out the binary image with the best segmentation effect, and masking the binary image with the hyperspectral images in each wave band to obtain a mask image only containing rice grains; reflection and transmission spectrum data are extracted based on each mask image.
More specifically, the rice grain quality trait parameters in step B include one or more of protein content, amylose content, moisture content, gelatinization temperature and consistency.
More specifically, in the step C, the reflection and transmission spectrum data are processed to obtain a characteristic spectrum set, specifically in the following manner: carrying out smoothing pretreatment on the reflection spectrum data and the transmission spectrum data based on a simple moving average method, an SG convolution smoothing method and a fast Fourier transform method, and then calculating derivative spectrum data of the smoothed spectrum value, wherein the derivative spectrum data comprises a first derivative and a second derivative; carrying out pearson correlation analysis on the smoothed spectrum data and the derivative spectrum data thereof and the quality character artificial value to determine the correlation between the spectrum data and the quality character, further screening to obtain a spectrum data set S1 highly correlated with the quality character, and screening characteristic wavelengths by adopting UVE, CARS and SPA respectively to obtain spectrum data sets S2, S3 and S4 corresponding to the characteristic wavelengths; the intersection of the spectral data sets S1, S2, S3 and S4 is found, and the result obtained is taken as a characteristic spectral set.
More specifically, in the step D, a stepwise linear regression method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, specifically,
step D1, in a characteristic spectrum set, selecting a variable and a candidate set at each time to establish a spectrum-quality character linear regression model together; the candidate set is initially empty;
step D2, finding the model with the minimum AIC score of the erythrocyte information criterion in all the models in the step D1, and selecting the corresponding variable to add into the candidate set;
step D3, repeating the steps D1 and D2 until the AIC score of the erythro pool information criterion is no longer reduced, and finally, the variables in the candidate set are called AIC initial selection variables;
step D4, using the Bayesian information criterion BIC as an evaluation index, and repeating the steps D1-D3 to obtain BIC primary selection variables;
step D5, obtaining an intersection of the AIC primary selection variable and the BIC primary selection variable to obtain a final variable;
step D6, establishing a linear regression model by using the final variable, and sorting the variable variance expansion factor VIF values in a descending order, and sequentially deleting one variable from high to low;
step D7, repeating the step D6 until the variance expansion factor VIF of all variables of the regression model is smaller than 10; wherein, the variable screening process of the steps D1 to D7 is shown in the figure 3;
and D8, taking the variables screened in the step D7 as input, and establishing a linear regression model.
More specifically, in the step D, a traditional nonlinear regression method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, specifically, a range (0-N) of a principal component is empirically set for partial least squares regression, and then the optimal principal component size of the partial least squares regression is determined through a Bayes optimization algorithm; the Bayesian optimization algorithm utilizes the information obtained before to optimize the next iteration when each iteration is performed, and the specific mode is as follows: firstly, dividing a test set and a verification set, then initializing the number of principal components to be N/2 to fit a regression model, and calculating the loss on the verification set; then, the number of principal components is determined by continuous iteration of a binary search algorithm, and the loss on the verification set is calculated; after multiple iterations, the regression model with the highest score on the verification set is selected, and the corresponding K is the optimal principal component size.
More specifically, in the step D, a machine learning regression method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, and the specifically adopted method comprises support vector machine regression and random forest regression.
More specifically, in the step D, a deep learning method is adopted to establish a quality character-rice grain hyperspectral characteristic spectrum regression prediction model, and the specific method is as follows: firstly, extracting the characteristics of a spectrum set by using a convolution layer and a maximum pooling layer of a CNN model, then inputting the characteristics into an LSTM model to obtain the relation between spectrum characteristic sequences, setting an excitation function of an output layer as linear to represent a regression task, optimizing by using a random gradient descent method, taking an average absolute error as an index of a loss function, training 10 groups of data each time, obtaining a result after 100 iterations, and visualizing each iteration result.
More specifically, in step E, the optimal model is selected by evaluating the regression model result according to parameters including the determination coefficient R2 and the root mean square error RMSE of the model.
More specifically, based on the optimal model screened in the step E, multiple quality character contents can be obtained simultaneously only by using a developed integrated hyperspectral reflection and transmission spectrum imaging system to scan spectrum information of rice grains.
The method can be applied to nondestructive detection of quality characters of crop seeds such as corn, wheat, rape and the like and fruit and vegetable tubers such as apples, potatoes and the like
The specific examples described in this application are intended to be illustrative only of the spirit of the invention. Various modifications or additions may be made to the embodiments described herein by those skilled in the art, or the manner of practicing the invention may be substituted in the form thereof without departing from the spirit of the invention or exceeding the scope of the invention as defined in the appended claims.
Claims (9)
1. A rice grain quality character nondestructive prediction method based on reflection and transmission spectrum is characterized by comprising the following steps of,
step A, an integrated hyperspectral reflection and transmission spectrum imaging system is built, reflection and transmission spectrum images of a plurality of healthy and full rice seeds are collected, and reflection and transmission spectrum data are extracted;
step B, husking and grinding rice grains to obtain rice powder, and obtaining quality character parameters of the rice grains by a traditional chemical method;
c, processing the reflection and transmission spectrum data obtained in the step A to obtain a characteristic spectrum set;
step D, establishing a quality character-rice grain characteristic spectrum regression prediction model by taking the characteristic spectrum set as an independent variable and taking the quality character artificial value as a dependent variable;
and E, screening an optimal model, and carrying out nondestructive prediction on the quality character of the rice grains by using the optimal model.
2. The method for nondestructively predicting quality traits of rice grains based on reflection and transmission spectrums according to claim 1, wherein the reflection and transmission spectrum data is extracted in the step A by the following steps: spectral image information is stored in a system workstation in a binary data file format which is stored in a band line-by-line crossing manner; c++ is adopted to reform image data, and hyperspectral images under each wave band are extracted; dividing hyperspectral images in each wave band in a two-by-two circulation mode, carrying out image segmentation by combining an OTSU algorithm to obtain a binary image, screening out the binary image with the best segmentation effect, and masking the binary image with the hyperspectral images in each wave band to obtain a mask image only containing rice grains; reflection and transmission spectrum data are extracted based on each mask image.
3. The method for non-destructive prediction of quality traits of rice grains based on reflection and transmission spectra according to claim 1, wherein the quality trait parameters of rice grains in step B comprise one or more of protein content, amylose content, moisture content, gelatinization temperature and gel consistency.
4. The method for nondestructively predicting quality traits of rice grains based on reflection and transmission spectrums according to claim 1, wherein in the step C, reflection and transmission spectrum data are processed to obtain a characteristic spectrum set, specifically adopting the following modes: carrying out smoothing pretreatment on the reflection spectrum data and the transmission spectrum data based on a simple moving average method, an SG convolution smoothing method and a fast Fourier transform method, and then calculating derivative spectrum data of the smoothed spectrum value, wherein the derivative spectrum data comprises a first derivative and a second derivative; carrying out pearson correlation analysis on the smoothed spectrum data and the derivative spectrum data thereof and the quality character artificial value to determine the correlation of the spectrum data and the quality character, screening to obtain a spectrum data set S1 highly correlated with the quality character, and screening characteristic wavelengths by adopting a non-information variable screening method UVE, a competitive self-adaptive re-weighting sampling method CARS and a continuous projection algorithm SPA to obtain spectrum data sets S2, S3 and S4 corresponding to the characteristic wavelengths; the intersection of the spectral data sets S1, S2, S3 and S4 is found, and the result obtained is taken as a characteristic spectral set.
5. The method for nondestructive prediction of quality characteristics of rice grains based on reflection and transmission spectra according to claim 1, wherein the step D is characterized in that a stepwise linear regression method is adopted to build a regression prediction model of quality characteristics-rice grain hyperspectral characteristic spectra,
step D1, in a characteristic spectrum set, selecting a variable and a candidate set at each time to establish a spectrum-quality character linear regression model together; the candidate set is initially empty;
step D2, finding the model with the minimum AIC score of the erythrocyte information criterion in all the models in the step D1, and selecting the corresponding variable to add into the candidate set;
step D3, repeating the steps D1 and D2 until the AIC score of the erythro pool information criterion is no longer reduced, and finally, the variables in the candidate set are called AIC initial selection variables;
step D4, using the Bayesian information criterion BIC as an evaluation index, and repeating the steps D1-D3 to obtain BIC primary selection variables;
step D5, obtaining an intersection of the AIC primary selection variable and the BIC primary selection variable to obtain a final variable;
step D6, establishing a linear regression model by using the final variable, and sorting the variable variance expansion factor VIF values in a descending order, and sequentially deleting one variable from high to low;
step D7, repeating the step D6 until the variance expansion factor VIF of all variables of the regression model is smaller than 10;
and D8, taking the variables screened in the step D7 as input, and establishing a linear regression model.
6. The nondestructive prediction method for quality characteristics of rice grains based on reflection and transmission spectrums is characterized in that a traditional nonlinear regression method is adopted in the step D to establish a quality characteristic-rice grain hyperspectral characteristic spectrum regression prediction model, specifically, a range (0-N) of a principal component is empirically set for partial least squares regression, and then the optimal principal component size of the partial least squares regression is determined through a Bayesian optimization algorithm; the Bayesian optimization algorithm utilizes the information obtained before to optimize the next iteration when each iteration is performed, and the specific mode is as follows: firstly, dividing a test set and a verification set, then initializing the number of principal components to be N/2 to fit a regression model, and calculating the loss on the verification set; then, the number of principal components is determined by continuous iteration of a binary search algorithm, and the loss on the verification set is calculated; after multiple iterations, the regression model with the highest score on the verification set is selected, and the corresponding K is the optimal principal component size.
7. The method for nondestructive prediction of quality properties of rice grains based on reflection and transmission spectrums according to claim 1, wherein the method in the step D is characterized in that a machine learning regression method is adopted to establish a quality property-rice grain hyperspectral characteristic spectrum regression prediction model, and the specifically adopted method comprises support vector machine regression and random forest regression.
8. The method for nondestructive prediction of quality characteristics of rice grains based on reflection and transmission spectra according to claim 1, wherein the deep learning method is adopted in the step D to establish a regression prediction model of the quality characteristics-rice grain hyperspectral characteristics, and the specific method is as follows: firstly, extracting the characteristics of a spectrum set by using a convolution layer and a maximum pooling layer of a CNN model, then inputting the characteristics into an LSTM model to obtain the relation between spectrum characteristic sequences, setting an excitation function of an output layer as linear to represent a regression task, optimizing by using a random gradient descent method, taking an average absolute error as an index of a loss function, training 10 groups of data each time, obtaining a result after 100 iterations, and visualizing each iteration result.
9. The method for non-destructive prediction of quality traits of rice grains based on reflection and transmission spectrum according to claim 1, wherein in step E, the optimal model is selected specifically according to a decision coefficient R comprising the model 2 And parameters including Root Mean Square Error (RMSE) evaluate regression model results and screen an optimal model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211681383.4A CN116026795A (en) | 2022-11-25 | 2022-11-25 | Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211681383.4A CN116026795A (en) | 2022-11-25 | 2022-11-25 | Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116026795A true CN116026795A (en) | 2023-04-28 |
Family
ID=86073529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211681383.4A Pending CN116026795A (en) | 2022-11-25 | 2022-11-25 | Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116026795A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235533A (en) * | 2023-11-10 | 2023-12-15 | 腾讯科技（深圳）有限公司 | Object variable analysis method, device, computer equipment and storage medium |
-
2022
- 2022-11-25 CN CN202211681383.4A patent/CN116026795A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235533A (en) * | 2023-11-10 | 2023-12-15 | 腾讯科技（深圳）有限公司 | Object variable analysis method, device, computer equipment and storage medium |
| CN117235533B (en) * | 2023-11-10 | 2024-03-01 | 腾讯科技（深圳）有限公司 | Object variable analysis method, device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mishra et al. | Improving moisture and soluble solids content prediction in pear fruit using near-infrared spectroscopy with variable selection and model updating approach | |
| Yu et al. | Development of deep learning method for predicting firmness and soluble solid content of postharvest Korla fragrant pear using Vis/NIR hyperspectral reflectance imaging | |
| CN109493287B (en) | Deep learning-based quantitative spectral data analysis processing method | |
| Guo et al. | Nondestructive measurement of soluble solids content of kiwifruits using near-infrared hyperspectral imaging | |
| Zhang et al. | Fast prediction of sugar content in dangshan pear (Pyrus spp.) using hyperspectral imagery data | |
| Zhu et al. | Rapid detection and visualization of slight bruise on apples using hyperspectral imaging | |
| Khodabakhshian et al. | Application of Vis/SNIR hyperspectral imaging in ripeness classification of pear | |
| Kim et al. | Application of ensemble neural-network method to integrated sugar content prediction model for citrus fruit using Vis/NIR spectroscopy. | |
| Qi et al. | Moldy peanut kernel identification using wavelet spectral features extracted from hyperspectral images | |
| CN116026795A (en) | Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum | |
| Chen et al. | Probing 1D convolutional neural network adapted to near-infrared spectroscopy for efficient classification of mixed fish | |
| CN111693487A (en) | Fruit sugar degree detection method and system based on genetic algorithm and extreme learning machine | |
| Tian et al. | Detection of soluble solid content in apples based on hyperspectral technology combined with deep learning algorithm | |
| Walsh et al. | The evolution of chemometrics coupled with near infrared spectroscopy for fruit quality evaluation. II. The rise of convolutional neural networks | |
| CN115049902B (en) | Visual prediction method, system, equipment and storage medium for water content of citrus blade | |
| Chia et al. | Neural network and extreme gradient boosting in near infrared spectroscopy | |
| CN116519661A (en) | Rice identification detection method based on convolutional neural network | |
| CN114324215A (en) | Lemon leaf chlorophyll content and two-dimensional distribution detection method thereof | |
| Hsieh et al. | Separating Septicemic and Normal Chicken Livers by Visible/near–infrared Spectroscopy and Back–propagation Neural Networks | |
| Tegegn et al. | Convolutional neural networks for quantitative prediction of different organic materials using near-infrared spectrum | |
| Zhang et al. | Hyperspectral detection of walnut protein contents based on improved whale optimized algorithm. | |
| Yu | Development of bulk peanuts maturity predictive model using hyperspectral imaging system | |
| Ninh et al. | Determination of fruit freshness using near-infrared spectroscopy and machine learning techniques | |
| CN111693488A (en) | Fruit grade classification method and system based on DS evidence theory fusion | |
| Fu et al. | Fast measurement of fatty acid in soybean using reflective hyperspectral imaging |
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 |