Method for evaluating internal and external physicochemical characteristics of powder banana based on svm multispectral model

Abstract

The present application relates to the technical field of quality detection of powder banana, and discloses a method for evaluating the internal and external physicochemical characteristics of powder banana based on a SVM multispectral model, which comprises the following steps: step S1: sample preparation; step S2: external physicochemical characteristic determination; step S3: internal physicochemical characteristic determination; step S4: image acquisition and processing, wherein a camera is connected to a computer and the parameters are adjusted, the powder banana sample is placed directly below the lens, and black velvet is laid at the bottom of the dark box to avoid reflection; the image processing comprises image conversion and storage, image visual optimization and image understanding; and step S5: partial least squares discrimination is established. The SVM multispectral model in the present application can effectively evaluate the external physicochemical characteristics of powder banana and has a certain potential feasibility for evaluating the internal physicochemical characteristics of powder banana.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of quality detection of powder bananas, and particularly relates to a method for evaluating internal and external physicochemical characteristics of powder bananas based on a multi-spectral model of SVM. BACKGROUND

[0002] As an important tropical fruit, the quality of powder bananas directly affects the consumer experience and industry benefits. Traditional powder banana quality detection methods mainly rely on manual experience judgment and destructive physicochemical analysis, which have the disadvantages of low efficiency, strong subjectivity, and inability to realize non-destructive detection, and are difficult to meet the needs of modern agricultural production and the market.

[0003] In recent years, optical non-destructive detection technology has been widely used in fruit quality detection due to its advantages of rapidity, non-destructivity, environmental protection, etc. Among them, visible / near-infrared spectroscopy technology (Vis / NIRS) can reflect the internal composition and structure information of fruits, providing a new technical means for non-destructive detection of fruit quality. At present, Vis / NIRS technology has achieved certain results in the quality detection of fruits such as apples, pears, and peaches, but the research on powder banana quality detection is relatively less, and there are the following problems:

[0004] Lack of systematic research on the optical properties of powder bananas: existing researches mainly focus on the prediction of single quality indicators, and lack of systematic measurement and analysis of optical property parameters (such as absorption coefficient μa and reduced scattering coefficient μs') of powder bananas, which is difficult to fully reflect the quality information of powder bananas.

[0005] The prediction accuracy and stability of the model need to be improved: existing prediction models are mainly based on single algorithm, and the prediction accuracy and stability are limited, which is difficult to meet the actual application needs.

[0006] Lack of comprehensive prediction of internal and external quality indicators of powder bananas: existing researches mainly focus on the prediction of single quality indicators, and lack of comprehensive prediction of internal and external quality indicators (such as single fruit weight, soluble solids, acidity, etc.) of powder bananas, which is difficult to fully evaluate the quality of powder bananas.

[0007] Therefore, the present application is proposed. SUMMARY

[0008] To solve the above technical problems, the basic idea of the technical solution of the present application is:

[0009] The method for evaluating internal and external physicochemical characteristics of powder bananas based on SVM includes the following steps:

[0010] Step S1: sample preparation, select powder bananas with maturity of 6-7; after placing the powder bananas in a constant temperature box at 30℃ for 3-4 days, select powder bananas with maturity of 8-9, and without pests, scars, deformities, and mechanical damage as samples;

[0011] Step S2: external physicochemical characteristics measurement, using an electronic balance to measure the fresh weight of the fruit, using a soft tape to measure the outer diameter, inner diameter and circumference of the fruit; using a 3nh precision color difference meter NR20XE to measure the color of the upper, middle and lower points of the selected sample, and taking the average value;

[0012] Step S3: measurement of internal physicochemical characteristics, cutting the banana into pieces, homogenizing and wrapping with sterile gauze, and squeezing the juice onto the Love PAL-BX / ACID6 banana special sugar acid integrated machine sensing mirror to measure the soluble solids content of the powder banana;

[0013] Step S4: image acquisition and processing, connecting the camera to the computer and adjusting the parameters, placing the powder banana sample directly below the lens, and placing black velvet on the bottom of the dark box to avoid reflection; image processing includes: image conversion and storage, image visual optimization and image understanding;

[0014] Step S5: by establishing a partial least squares discriminant method, after feature extraction of the training sample, a regression model between the independent variable and the dependent variable of the training sample is established, and then the most effective feature information for classification is found;

[0015] Step S6: representing the deviation between the observed value and the true value, the algorithm is as follows

[0016]

[0017] y i is the actual value, is the predicted value, and n is the sample number.

[0018] As a preferred embodiment of the present application, in step S3, about 1g of banana sample is weighed by an electronic scale, placed in a 100mL beaker, then 50mL of distilled water is added by a measuring spoon, and the sample is stirred uniformly by a glass straw to fully dilute the sample, and the diluted acidity measurement sample is taken and dropped onto the instrument sensing mirror to obtain the powder banana acidity; each is repeated 3 times, and the average value is taken.

[0019] Compared with the prior art, the present application has the following advantages:

[0020] 1. Multi-index fusion and model optimization

[0021] By integrating the detection data of external physicochemical characteristics (single fruit weight, outer diameter, inner diameter, circumference, color) and internal physicochemical characteristics (soluble solids, acidity), combined with partial least squares regression, support vector machine (SVM), random forest and other models, comprehensive modeling is carried out. Compared with traditional single index detection or single model application, this method improves the model generalization ability through multi-dimensional data fusion, for example, the R 2 =0.81, RPD=2.30, which is significantly better than some existing chemical determination methods.

[0022] 2. High efficiency of non-destructive testing

[0023] The non-destructive evaluation of the quality of the banana is realized by using the multispectral technology combined with machine learning. Compared with the existing method relying on manual sampling chemical determination, the method does not need to destroy the fruit, the detection speed is faster, and the external morphology and internal component information can be obtained at the same time, and the detection cost is reduced.

[0024] 3. Model selection and improvement of grading accuracy

[0025] By comparing the modeling effects of different models (SVM, random forest, etc.), the SVM model with the highest recognition accuracy is preferentially selected. In the prediction of external features (inner diameter, circumference, and outer diameter), RPD>1.5, R 2 >0.8, which is better than the model relying only on spectral or image features in some existing researches.

[0026] 4. Potential evaluation ability of internal quality

[0027] Although the single fruit weight and pulp weight prediction effect is limited, the SVM model shows feasibility in internal indicators such as soluble solids and acidity.

[0028] The specific embodiments of the application will be described in further detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0029] In the drawings:

[0030] Figure 1 It is a schematic diagram of the determination position of the external physicochemical characteristics of the banana;

[0031] Figure 2 It is a schematic diagram of the image acquisition device;

[0032] Figure 3 Fig. (a) is a change diagram of the average multispectral and confidence interval of the banana; (b) is a normal distribution test diagram of the multispectral of the banana;

[0033] Figure 4 It is a histogram of the apparent data distribution of the lightness (L*) and red-green color component (a*) of the color of the banana;

[0034] Figure 5 It is a histogram of the apparent data distribution of the yellow-blue color component (b*) and chroma (c*) of the color of the banana;

[0035] Figure 6 It is a histogram of the apparent data distribution of the color hue angle (h*) and soluble solids of the banana;

[0036] Figure 7The apparent data distribution histogram of the outer diameter and inner diameter of the color of the powder banana;

[0037] Figure 8 The apparent data distribution histogram of the outer diameter and inner diameter of the color of the powder banana;

[0038] Figure 9 The apparent data distribution histogram of the girth and peel weight of the color of the powder banana. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application, and the following embodiments are used to illustrate the present application.

[0040] Embodiment 1:

[0041] As shown in the method for evaluating the internal and external physicochemical characteristics of the powder banana based on the SVM multispectral model, the method comprises the following steps: Figures 1 to 3

[0042] Step S1: sample preparation, selecting the powder banana with maturity of 6-7; after the powder banana is placed in a constant temperature box at 30 DEG C for 3-4 days, the powder banana with maturity of 8-9 and without diseases, pests, scars, deformities and mechanical damage is selected as the sample;

[0043] Step S2: determination of external physicochemical characteristics, using an electronic balance to determine the fresh weight of the fruit, using a soft tape measure to determine the outer diameter, inner diameter and girth of the fruit; using a 3nh precision colorimeter NR20XE to determine the color of the selected sample at three points of upper, middle and lower, and taking the average value;

[0044] Step S3: determination of internal physicochemical characteristics, after the banana is cut into pieces and homogenized and wrapped with sterile gauze, the juice is squeezed on the Love PAL-BX / ACID6 banana special sugar acid integrated machine sensing mirror to determine the soluble solid content (Brix value) of the powder banana, about 1g of banana sample is weighed by an electronic balance and placed in a 100mL beaker, then 50mL of distilled water is added by a measuring spoon, and the sample is stirred uniformly by a glass pipette to dilute the sample sufficiently, and the diluted acidity measurement sample is taken and dropped on the instrument sensing mirror to obtain the acidity (Acid value) of the powder banana; each is repeated for 3 times, and the average value is taken;

[0045] Step S4: image acquisition and processing, connecting the camera with the computer and adjusting the parameters, placing the powder banana sample directly below the lens, and laying black velvet on the bottom of the dark box to avoid reflection; image processing includes: image conversion and storage, image visual optimization and image understanding;

[0046] ​Step S5: After feature extraction of the training samples by establishing a partial least squares discriminant method, a regression model between the independent variables and dependent variables of the training samples is established, and then the most effective feature information for classification is found;

[0047] Step S6: The deviation between the observed value and the true value is represented, and the algorithm is as follows

[0048]

[0049] y i is the actual value, is the predicted value, and n is the sample number.

[0050] As Figures 1 to 3 shown, in the specific embodiment, in bananas, the acid value (acid value) reflects the concentration of sugar in banana juice. For example, a Brix value of 15% means that 100 grams of banana juice contains 15 grams of soluble solids (mainly sugar). The acid value refers to the content of organic acids (such as citric acid, malic acid, etc.) in fruits, usually expressed in percentage or acidity coefficient (such as pH). In bananas, the main organic acids are citric acid and malic acid, and the size of their acidity will affect the taste of bananas.

[0051] Further, the image acquisition device mainly consists of a portable multispectral imaging camera, a light source, a dark box and a computer. Among them, the portable multispectral imaging camera (Monarch Pro TM , Unispectral Company), spectral range: 705-920nm, spectral accuracy ±2.5nm; single frame / spectral image cube; 60 frames per second in single band mode, 25 frames per second in dual band mode. The camera lens is fixed in the center of the customized dark box, about 50cm away from the sample. The dark box is composed of black acrylic plates to block external light, and the internal size is 300x400x540mm. Halogen cold light source (XD-302, Shanghai Wenmei Optoelectronic Technology Co., Ltd.), with double-tube hard 1-meter optical fiber and adjustable condenser lens. The device is equipped with a computer with Monarch APP software for data acquisition. The physical map and schematic diagram of the device are shown in Figure 2 .

[0052] Image processing software: In terms of software, the image acquisition software Monarch is used to collect the image of the powder banana, and the Matlab software is used to process and model analysis of the image of the powder banana.

[0053] (1) Monarch image acquisition software. This image acquisition software can connect the camera at the same time, get real-time display image, and can adjust the camera parameters, and can be well compatible with Windows system. In the main interface of the software, it is divided into image display module, basic operation module and parameter adjustment module. Among them, the basic operation module includes initialization, screenshot, disconnection and parameter update; parameter setting: exposure time 8000, gain 5dB, frame rate 150fps, brightness 10.

[0054] (2) Matlab software. Matlab is developed by Mathworks Company in the United States, which has powerful computing function and various toolboxes for development, matrix calculation, etc. The image processing toolbox of Matlab has powerful image processing function, which can realize digital processing of image by writing code or using functions in the toolbox.

[0055] Image acquisition and processing: connect the camera with the computer and adjust the parameters, put the banana sample under the lens, and lay black velvet on the bottom of the dark box to avoid reflection. A total of 113 banana samples were collected, and the images were collected at 713, 736, 759, 782, 805, 828, 851, 874, 897 and 920 nm at a time. The resolution of each image was 1280×1024 pix, and the storage format was.png format. A sample was repeated 3 times at different positions.

[0056] Image processing can be roughly divided into three levels: (1) image transformation and storage. It refers to the digitization of image, spatial conversion of image and image encoding compression. (2) Image visual optimization. It is divided into image enhancement and image restoration, both of which are to improve the quality of image. (3) Image understanding. The key steps of image understanding are image segmentation and feature extraction. The meaning of image segmentation is to separate the region of interest from the background or other objects in the image, which is helpful for further analysis. Image feature extraction refers to extracting and identifying the useful information contained in the image, so as to judge the category and quantity of the target contained in the image and other effective information.

[0057] Data processing method: After feature extraction of the image, it is important to select the appropriate modeling method for data processing. The purpose of modeling is to enable the computer to automatically identify the research object and classify it, and to make the recognition rate as high as possible. In this study, two methods, support vector machine (SVM-C) and partial least squares discriminant analysis (PLS-DA), were selected. SVM-C: Support vector machine model is a binary linear classifier in machine learning method. Its basic principle is to maximize the gap between support vectors after classification, that is, to find an optimal boundary to divide the data as accurately as possible. Support vector machine can not only be used in linear classification problems, but also can be extended by kernel function to make it applicable to non-linear classification problems. Kernel function is to output data from a given space to a new high-dimensional space, so as to use hyperplane to classify data, which can solve the problem of high-dimensional feature classification and regression. PLS-DA: Partial least squares discriminant analysis is a technique that minimizes the sum of squared errors to find the optimal classification function. After feature extraction of the training sample, a regression model between the independent variable and the dependent variable of the training sample is established, and then the most effective feature information for classification is found.

[0058] After the model is established, some evaluation parameters are needed to measure the accuracy and precision of the model. In this study, root mean square error (RMSE) was used to determine the effect of the model after feature parameter screening. It can be used to represent the deviation between the observed value and the true value. Therefore, the smaller the RMSE, the better the stability of the model and the higher the precision.

[0059] External and internal physicochemical characteristics: The external physicochemical characteristics (single fruit weight, external diameter, internal diameter, circumference, color) and internal physicochemical characteristics (soluble solids, acidity) of 113 powder bananas were determined. The specific values are shown in Table 2.

[0060] Image data processing results: Therefore, the distribution of physicochemical values of the samples showed a strong concentration trend, close to normal distribution ( Figure 3 ), indicating that the obtained physicochemical parameter data can effectively represent the physicochemical state of the powder banana variety, and can be used for the construction of the next physicochemical regression model.

[0061] From Figure 4 It can be seen from (a) that the average reflectivity of powder banana changes in the range of 0.43 to 0.48, and shows an upward trend in the wavelength range of 713 nm to 851 nm, and a slight downward trend in the wavelength range of 851 nm to 920 nm. Figure 5(b) The normal distribution of reflectance in different wave bands can be seen from the figure, the reflectance of all wave bands is concentrated around 0.48, the change range is between 0.1 and 0.9, and the reflectance of all wave bands presents a good normal distribution, which indicates that there is reasonable individual difference between the multi-spectral data of powder banana, and the next step of regression model construction can be carried out.

[0062] Test example:

[0063] Table 1 Comparison of results of different regression models in evaluating the color of powder banana

[0064]

[0065]

[0066] Compared with the traditional linear model-partial least squares (PLS), the prediction result of the color of powder banana is poor, the residual prediction deviation (RPD) is less than 1.5, and the nonlinear machine learning method and the deep learning algorithm have better results in evaluating the color of powder banana (Table 2). Among all the prediction results of the color of powder banana, the regression model based on support vector machine (SVM) obtains the best prediction result, the prediction accuracy (Rv2) is greater than 0.82, and the RPD is greater than 2.36. Among them, the prediction of L*, a*, b* and h* obtains satisfactory results, the prediction accuracy is 0.91, 0.85, 0.87 and 0.92 respectively, and the RPD value reaches 3.40, 2.58, 2.82 and 3.46 respectively. The prediction result of c* is acceptable (Rv 2 =0.82, RPD=2.36).

[0067] Table 4 shows the comparison of results of different regression models in evaluating the sugar acidity and appearance of powder banana. The results show that in evaluating the soluble solids of powder banana, the SVM model shows certain feasibility (RPD>1.5). However, the results of all models in predicting the titratable acidity of powder banana are not ideal (RPD<1.5), which is due to the fact that the multi-spectral data of the surface of powder banana peel is obtained, the peel interferes with the response of the spectrum and the internal physicochemical properties, and on the other hand, there is no significant C-H absorption peak in the range of 713-920 nm, resulting in unsatisfactory prediction results. In addition, the SVM model also shows certain feasibility in predicting the size dimension (inner diameter, circumference) of powder banana (RPD>1.5), and obtains good results in predicting the outer diameter of powder banana (Rv 2= 0.81, RPD = 2.30). While in predicting the single fruit weight and pulp weight of the banana, the results were poor. This shows that the SVM-based multispectral model can effectively evaluate the external physicochemical characteristics (chroma, external diameter) of the banana and has certain potential feasibility in evaluating the internal physicochemical characteristics (soluble solids, single fruit weight) of the banana.

[0068] The feature band interpretation based on the SVM model is shown in Figure 5 The feature band weight distribution of the SVM model with good prediction results is shown in FIG. 6. It is shown that the wave bands that have a greater impact on the prediction results of the banana chroma data mainly concentrate on 713 nm, 736 nm, 782 nm, and 897 nm. The SHAP values of these wave bands show a large range of changes, indicating that they have a good effect on the differentiation and prediction of the samples. In addition, the wave band that contributes most to the prediction results of the banana external diameter is 713 nm, and the range of the SHAP values of 713 nm is greater than that of the other wave bands, which indicates that 713 nm can better distinguish and predict the external diameter of the banana.

[0069] Table 2 Comparison of the results of different regression models in evaluating the sugar acidity and appearance of the banana

[0070]

[0071]

[0072]

[0073] The advanced technical means of Sylvie Bureau and his team from the French Institute of Quality and Safety of Plant Origin were introduced, and the main tropical fruit banana was taken as the research object. The external physicochemical characteristics (single fruit weight, external diameter, internal diameter, girth, chroma), internal physicochemical characteristics (soluble solids, acidity) quality indicators of the banana were detected, and two grade discrimination models of partial least squares regression, support vector machine, random forest, decision tree, and one-dimensional convolutional neural network model were established. The modeling effects of different models were compared, and the model with the highest recognition accuracy was selected to be imported into the software. The SVM model also showed certain feasibility in predicting the size dimension (internal diameter, girth) of the banana (RPD > 1.5), and good results were obtained in predicting the external diameter of the banana (Rv2= 0.81, RPD = 2.30). While in predicting the single fruit weight and pulp weight of the banana, the results were poor. This shows that the SVM-based multispectral model can effectively evaluate the external physicochemical characteristics (chroma, external diameter) of the banana and has certain potential feasibility in evaluating the internal physicochemical characteristics (soluble solids, single fruit weight) of the banana.

Claims

1. A method for evaluating the internal and external physical and chemical characteristics of bananas based on a multispectral model based on SVM, characterized in that: The steps include: Step S1: Sample preparation: pink bananas with a maturity of 60% to 70% are selected; after the pink bananas are placed in a constant temperature box at 30°C for 3-4 days, pink bananas with a maturity of 80% to 90% and no diseases, insect pests, scars, deformities, or mechanical damage are selected as samples; Step S2: Determination of external physical and chemical characteristics: use an electronic balance to measure the fresh weight of the fruit, and use a soft tape measure to measure the outer diameter, inner diameter, and girth of the fruit; use a 3nh precision colorimeter NR20XE to measure the chromaticity of the top, middle, and bottom points of the selected sample and take the average value; Step S3: Determination of internal physicochemical characteristics: The bananas were cut into pieces and homogenized, wrapped with sterile gauze, and the juice was squeezed out and the soluble solids content of the banana powder was measured on the sensing mirror of the Aituo PAL-BX / ACID6 banana-specific sugar-acid integrated machine; Step S4: Image acquisition and processing: Connect the camera to the computer and adjust the parameters. Place the banana sample directly below the lens, and cover the bottom of the darkroom with black velvet to prevent reflections. Image processing includes image conversion and storage, image visual optimization, and image understanding. A total of 113 banana samples were imaged at 713, 736, 759, 782, 805, 828, 851, 874, 897, and 920 nm. Each image had a resolution of 1280 × 1024 pixels and was stored in .png format. Each sample was taken at different locations and repeated three times. Step S5: After extracting features from the banana image, an SVM regression model is established to evaluate the internal and external physical and chemical characteristics of the banana. Step S6: Characterize the deviation between the observed value and the true value. The algorithm is as follows: y i is the actual value, is the predicted value, and n is the number of samples.

2. The method for evaluating the internal and external physical and chemical characteristics of banana leaves based on a multispectral model based on SVM according to claim 1, characterized in that: In step S3, 1 g of banana sample was weighed using an electronic scale and placed in a 100 mL beaker. 50 mL of distilled water was then added using a measuring spoon and stirred evenly with a glass pipette to fully dilute the sample. An appropriate amount of the diluted acidity measurement sample was then dripped onto the sensing mirror of the instrument to obtain the banana acidity. Each step was repeated 3 times, and the average value was taken.

Images