Nondestructive detection method for tomato fruit physical and chemical indexes based on magnetic induction tomography

Abstract

The application discloses a nondestructive detection method for acid content of tomato fruits based on magnetic induction tomography, and comprises the following steps: S1, building a rapid nondestructive detection device based on the principle of magnetic induction tomography; S2, collecting conductivity data of corresponding tomato fruits by using a conductivity detector in the detection device; S3, determining the content of acid in the tomato fruits by using a refractometer method; S4, performing correlation analysis on the conductivity data obtained by the conductivity detector and the acid in the tomato fruits, and obtaining frequency band data with the highest correlation between the conductivity data of the tomato fruits and the acid content; S5, dividing the frequency band data obtained from S4 into a training set and a test set according to interval sampling and in a certain proportion; S6, constructing a regression prediction model for the acid in the tomato fruits by using a CNN (Convolutional Neural Network) convolutional neural network; and S7, inputting the conductivity data of a target obtained by the conductivity detector into the prediction model in S6, so as to obtain the acid content of the corresponding tomato.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of detecting the internal physical and chemical index content of fruits and vegetables, and is mainly applied to the field of non-destructive detection of the internal physical and chemical index of fruits and vegetables. Specifically, it relates to a method for non-destructive detection of the internal physical and chemical index of fruits and vegetables by using the principle of magnetic induction tomography to obtain high-throughput conductivity data. BACKGROUND

[0002] Fruits and vegetables are indispensable food in people's daily life, and the various nutritional components contained therein have always been the focus of attention of producers and consumers. In particular, with the continuous improvement of people's living quality in China, the public has also put forward higher requirements for the quality of fruits and vegetables. In order to meet the needs of consumers and the reasonable pricing of fruits and vegetables, the technology for non-destructive detection of the physical and chemical index of fruits and vegetables is widely used. This technology uses physical properties such as light, sound, electricity, and magnetism to obtain and analyze and evaluate the physical and chemical index of fruits and vegetables without damage, contact, or influence. The existing mainstream non-destructive detection technologies mainly include near-infrared spectroscopy, hyperspectral imaging, Raman spectroscopy, dielectric property analysis, acoustic property analysis, and other detection technologies. Among them, near-infrared spectroscopy, hyperspectral imaging, and Raman spectroscopy all belong to optical detection. Near-infrared spectroscopy uses the frequency doubling and frequency mixing absorption characteristics of hydrogen-containing groups such as C-H, N-H, and O-H to obtain spectral information, and through analysis of these absorption information, qualitative or quantitative detection of the sample can be achieved. Hyperspectral imaging combines spectral technology and imaging technology to obtain spectral information and image information of any point in three-dimensional space, thereby visualizing the distribution of each component. Raman spectroscopy provides vibration and rotation information of functional groups or chemical bonds to study the structure and properties of substances from the perspective of molecular vibration, without considering the influence of water molecule vibration, and complements infrared spectroscopy in molecular structure analysis. The dielectric property analysis detection technology selects dielectric property parameters sensitive to changes in the physical and chemical index of fruits and vegetables, constructs a correlation model between the physical and chemical index and the dielectric property parameters, and thereby realizes the detection of the quality of fruits and vegetables. In addition, different fruits and vegetables have different transmission, scattering, absorption, reflection, and attenuation characteristics under the action of different sound waves, and different frequencies. Therefore, by using these characteristics and the different absorption and scattering of sound waves by different fruits and vegetables, the degree of attenuation of sound waves is different, and this law is used to realize non-destructive detection of the internal physical and chemical index of fruits and vegetables.

[0003] As the invention patent with patent number CN115184295A proposes a non-destructive detection method for soluble solids content of pomelo based on near-infrared spectral detection information, by collecting sample spectral information and measuring sample sugar content data to establish a parameter model, the sugar content of Sanhong pomelo can be quickly and effectively detected without damaging the appearance of the fruit. However, this method is not suitable for fruits and vegetables with water content higher than 80% due to weak absorption in the near-infrared spectral region and low spectral signal-to-noise ratio. In addition, the invention patent with patent number CN116007753A proposes a real-time non-destructive detection method for fruit quality based on FPGA and hyperspectral, by acquiring spectral images of fruits and index parameters of fruit samples to construct a detection model, and outputting the sugar acidity and damage degree of each fruit through the detection model. However, the process of acquiring spectral data is easily affected by environmental and instrumental factors, has certain limitations, and the detection effect for fruits and vegetables with thick peels is not ideal.

[0004] Considering the close relationship between electric field and magnetic field, and the fact that magnetic field is the result of charge motion in matter and has no effect on fruits and vegetables, compared to spectral methods, magnetic induction has almost zero requirement for environmental light, and is not easily affected by instrument performance and processing speed. Compared to acoustic property methods, magnetic induction is not affected by bubbles inside fruits and vegetables and the hardness of fruits and vegetables. Therefore, by using the principle of magnetic induction and Maxwell's equations, relevant electrical property data can be obtained by applying an external magnetic field to quickly analyze the water content, sugar content, acidity, density, and other properties of fruits and vegetables. The invention proposes a non-destructive detection device and method for tomato fruit acidity based on the principle of magnetic induction tomography, by constructing an electrical conductivity meter based on magnetic induction tomography, obtaining relevant data of the electrical conductivity of tomato fruits (hereinafter referred to as electrical conductivity data), and constructing a physicochemical index prediction model by obtaining relevant physicochemical indexes of tomato fruits, finally using the prediction model to obtain the corresponding tomato fruit acidity content. SUMMARY

[0005] The purpose of the present invention is to propose a non-destructive detection method for tomato fruit acidity based on magnetic induction tomography. The present invention is a method of modeling from electrical conductivity data to physicochemical index analysis, thereby realizing rapid and non-destructive detection of fruit acidity in tomatoes, with simple operation and accurate detection effect.

[0006] The technical concept of the present invention is to place the corresponding fruit and vegetable target in the detection device, obtain its electrical conductivity data, analyze the correlation between the corresponding physicochemical index and it, determine whether there is a corresponding relationship between them, then perform error analysis on the obtained electrical conductivity data to exclude incorrect data, and then construct a prediction model between the corresponding physicochemical index of acidity and the electrical conductivity data using appropriate deep learning algorithms, and finally use the model to predict the corresponding physicochemical index content by inputting the electrical conductivity data obtained by the detection device.

[0007] The technical scheme adopted by the present application to achieve the above-mentioned application purposes is as follows:

[0008] The nondestructive detection method for the acid content of tomato fruits based on magnetic induction tomography comprises the following steps:

[0009] S1: A rapid nondestructive detection device based on the principle of magnetic induction tomography is built.

[0010] S2: The conductivity data of the corresponding tomato fruits are collected by using the conductivity detector in the detection device.

[0011] S3: The content of the acid in the tomato fruits is determined by using the refractometer method.

[0012] S4: The conductivity data obtained by the conductivity detector and the acid content of the tomato fruits are analyzed for correlation, and the frequency band data with the highest correlation between the conductivity data of the tomato fruits and the acid content are obtained.

[0013] S5: The frequency band data obtained from S4 are sampled according to intervals and divided into training sets and test sets according to a certain proportion.

[0014] S6: A regression prediction model for the acid content of tomato fruits is constructed by using a CNN convolutional neural network.

[0015] S7: Based on the prediction model in S6, the conductivity data of the corresponding target obtained by the conductivity detector are input, so as to obtain the acid content of the corresponding tomato.

[0016] Preferably, the rapid nondestructive detection device in step S1 comprises a fixer for fixing the fruits to be measured, which is placed on a conveyor belt, and the conveyor belt is annular; the conveyor belt passes through the center hole of the conductivity detector; the motor drives the conveyor belt, and the motor is controlled by a single-chip microcomputer.

[0017] Preferably, step S2 specifically comprises: 200 conductivity data of the corresponding tomato fruits are collected within a specified time by using an 8-channel conductivity detector, and the data are composed of 21 frequency band data from 15 MHz to 35 MHz, and each frequency band data is composed of 256 numerical values, which are converted from the phase data θ collected by the conductivity detector into corresponding conductivity data, and the definition is as follows:

[0018] V=A+Bi (1)

[0019] θ=arctan(B / A) (2)

[0020] Δσ=-[(S k ) T S k +λR T R]-1 (S k ) T [θ1-θ2] (3)

[0021] where A is the real part related to the dielectric constant of the conductor caused by displacement current, B is the imaginary part related to the conductivity and the excitation frequency caused by internal eddy current. Δσ is the conductivity difference, θ1 is the voltage phase data when the measurement area exists the measured object, and θ2 is the voltage phase data of the measurement area in the air background. Where R is the regularization matrix and λ is the regularization parameter.(S k ) T S k is the Hession matrix under the kth iteration.

[0022] Preferably, the step S3 specifically comprises:

[0023] S3.1: Take the tomato fruit and vegetable sample to be measured, cut it at the equatorial position, and take three times of tomato tissue liquid as the titration sample. Each sample is 1.00g of sample, 50.00ml of distilled water is added for dilution, and it is fully stirred and mixed.

[0024] S3.2: Use the PAL-EazyACID3 digital handheld tomato acidity meter to measure the titration sample of the tomato fruit and vegetable to be measured. The same titration sample is measured three times, and the average value is taken as the measurement result of the titration sample. The average value of the three titration samples of the same tomato is taken as the measurement result of the tomato fruit acid content.

[0025] Preferably, the step S4 specifically comprises:

[0026] S4.1: As can be seen from S1, the conductivity data obtained by the conductivity detector has a dimension of (200, 21, 256) matrix A1, as follows:

[0027] A1 = [a1 a2 a3 … a 200 ] (4)

[0028]

[0029] where a i is the conductivity data of the i-th tomato fruit and vegetable, with a dimension of (21, 256), and Δσ p,j is the conductivity value of the jth column under the corresponding frequency band p.

[0030] S4.2: Since each tomato fruit and vegetable has corresponding conductivity data of 21 frequency bands, in order to obtain the frequency band data that can best represent the highest correlation between the conductivity data and the tomato fruit acid, it is necessary to analyze the Pearson correlation degree of each frequency band respectively. The correlation degree calculation formula is as follows:

[0031]

[0032] μ X = E(X), μ Y = E(Y) (7)

[0033]

[0034] wherein μ X , μ Y and σX, σYare the expectation and standard deviation. X is a column of conductivity data in 256 columns of data under the corresponding frequency band, and Y is 200 tomato fruit acid data.

[0035] The tomato fruit acid content is defined as Y = [y1 y2 … y 200 ] T , and the data of the tomato fruit under the frequency band i is defined as W = [w i,1 w i,2 … w i,200 ] T , wherein w i,1 represents the conductivity data [Δσ i,1 Δσ i,2 … Δσ i,256 ] of the No. 1 tomato fruit under the frequency band i. The jth column of conductivity data in 256 columns of data under the corresponding frequency band of 200 tomato fruits is defined as X = [x 1,j x 2,j … x 200,j ] T .

[0036] S4.3: The data of the frequency band with the highest total value of correlation calculated and counted according to S3.2 is taken as the target conductivity data.

[0037] Preferably, the step S5 specifically comprises:

[0038] The conductivity data obtained by S5 is defined as:

[0039]

[0040] The above conductivity data D 200,256 is input as a data set, and the tomato fruit acid content data Y in S4 is output as a data set. The training set and the test set are divided according to 7:3, 10 consecutive samples are regarded as an independent subset, and the 1st, 5th and 10th samples in each subset are selected as the test set μ X = E(X), μ Y = E(Y), and the rest are taken as the training set.

[0041] Preferably, the step S6 specifically comprises:

[0042] S6.1: The convolutional neural network regression prediction model comprises 8 convolutional layers and 3 fully connected layers.

[0043] S6.2: Each convolutional layer uses a 3x1 convolutional kernel, and the number of kernels is 64, 64, 128, 128, 256, 512, and 512, respectively, and all strides are 1. After each convolutional layer, a ReLU activation function is followed, and each pooling layer uses a MaxPool1d function. The convolutional layers and pooling layers included in the model can extract features from the input 1x256 one-dimensional data. Finally, the output layer is a fully connected layer with a single neuron, which maps the extracted features to a single scalar output.

[0044] Preferably, the step S7 specifically comprises: using the training set in S5 to train the prediction model in S5, using the test set in S5 to test the performance of the model in S6, and using the determination coefficient R 2 and RMSE as indicators to determine the effectiveness of the model.

[0045] A tomato fruit and vegetable fruit acid content nondestructive detection device and method based on the principle of magnetic induction tomography are implemented, which comprises: a tomato conductivity data and fruit acid content acquisition module, a tomato conductivity and fruit acid content correlation analysis module, and a tomato fruit acid prediction model construction module.

[0046] The conductivity data acquisition module acquires fruit and vegetable conductivity data and fruit acid content. An 8-channel conductivity detector is used to collect 200 sets of conductivity data of corresponding tomato fruits and vegetables within a specified time. The data is composed of 21 frequency bands, from 15MHz to 35MHz, and each frequency band data is composed of 256 values. The conductivity data set is constructed from these data. In addition, a PAL-EazyACID3 digital handheld tomato acidimeter is used to measure the fruit acid content of the test tomato.

[0047] The tomato conductivity and fruit acid content correlation analysis module calculates the correlation between tomato conductivity and fruit acid to obtain the highest correlation frequency band data. Specifically, since each tomato fruit and vegetable has corresponding conductivity data of 21 frequency bands, in order to obtain the frequency band data that can best represent the strongest correlation between conductivity data and tomato fruit acid, Pearson correlation degree analysis is performed on each frequency band, and the frequency band data with the highest total correlation value calculated and counted is used as the target conductivity data.

[0048] The fruit acid prediction model construction module is specifically: the conductivity data obtained under the target frequency band is divided into a training set and a test set according to interval sampling and a 7:3 ratio. A convolutional neural network is used to construct a tomato fruit acid regression prediction model, and the test set is input into the conductivity data corresponding to the target obtained by the conductivity detector, so that the required fruit acid content of the corresponding numbered tomato is obtained. And on this basis, the R 2 And RMSE.

[0049] The tomato conductivity data and fruit acid content acquisition module, the tomato conductivity and fruit acid content correlation analysis module, and the tomato fruit acid prediction model construction module are sequentially connected.

[0050] The beneficial effects of the present application are:

[0051] (1). The non-destructive testing device based on the principle of magnetic induction tomography is reasonably utilized, compared with the traditional electrical characteristic instrument device which is too large in size and difficult to miniaturize, the device provided by the present application is convenient to carry and use, so that the detection process is more convenient and efficient. Compared with the optical characteristic detection method, the device of the present application is not easily affected by environmental factors such as light pollution and site restrictions, and can be detected under various environmental conditions, improving its practicality and applicability.

[0052] (2). The present application can find the corresponding law between the two by calculating and analyzing the correlation between conductivity and physicochemical indicators, so as to construct a prediction model between the related physicochemical indicators and conductivity.

[0053] (3). The present application adopts a convolutional neural network to construct a corresponding physicochemical indicator prediction model based on target conductivity data and tomato fruit acid physicochemical indicators, which can extract deep feature information of the conductivity data, avoid the influence of human factors such as feature selection in traditional chemometrics modeling methods. And can be transplanted for detecting multiple fruit and vegetable physicochemical indicators, such as soluble solids, water content, maturity and hardness, reducing the intervention of manual operation, improving the detection efficiency and reliability, and also reducing the detection cost and time.

[0054] (4). The present application is superior to the existing detection method in detection efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 The device schematic diagram of the present application;

[0056] Figure 2 The flowchart of the method of the present application;

[0057] Figure 3 The system framework schematic diagram of the conductivity detector of the present application;

[0058] Figure 4 Figure 1 is a scatter plot between predicted and actual values of tomato fruit acid;

[0059] Figure 5 Figure 2 is a schematic diagram of the system structure of the present application;

[0060] Figure 6 Figure 3 is a schematic diagram of the prediction model network structure of the present application. DETAILED DESCRIPTION

[0061] The specific embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0062] REFERENCE Figures 1 to 6 A non-destructive testing method for tomato fruit and vegetable physical and chemical indicators based on magnetic induction tomography, comprising the following steps:

[0063] S1: Construct a rapid non-destructive testing device based on the principle of magnetic induction tomography, specifically comprising:

[0064] A rapid non-destructive testing device is built, including: fruit and vegetable to be tested 1, detection fixture 2, motor 3, single-chip microcomputer 4, conveyor belt 5, conductivity detector 6; the fruit and vegetable to be tested is placed on the detection fixture, and then the fixture is placed on the conveyor belt, from left to right are the conductivity detector, the conveyor belt, the tomato fruit and vegetable to be tested, the detection fixture, the motor, and the single-chip microcomputer; there are eight detection coils in the conductivity detector, evenly distributed (period 45°). The conveyor belt and the motor form a conveying device, and the conveying rate is controlled by the single-chip microcomputer. The detection fixture is composed of resin material, which is a long rectangular box with a downwardly recessed top, with a size of length x width x height = 5 cm x 5 cm x 2.5 cm. The conveyor belt is made of polyimide (PI) material, with a length of 50 cm and a width of 8 cm. Due to its low dielectric constant, low dielectric loss, high temperature stability, and excellent mechanical properties, it is widely used in electronic, optical, and aerospace fields. When conveying fruit and vegetables, the use of polyimide material conveyor belt can reduce the interaction between the fruit and vegetable and the coil magnetic field. The single-chip microcomputer is STM32F103C8T6, a 32-bit microcontroller based on ARM Cortex-M core STM32 series, used to control the operation of the conveyor belt. The shell of the conductivity detector is a circular ring-shaped shell made of polytetrafluoroethylene material, with an inner ring diameter of 10 cm and an outer ring diameter of 20 cm. The specific process is as follows: place the tomato on the right side of the conveyor belt as shown in the figure, start the motor and then start the conveyor belt through the single-chip microcomputer, when the tomato reaches the left detection device, the conveyor belt pauses, and the conductivity detector is turned on to collect the conductivity of the tomato fruit and vegetable. Figure 1

[0065] S2: Collect the conductivity data of the corresponding tomato fruit and vegetable using the conductivity detector, specifically comprising: ​

[0066] The conductivity data corresponding to the tomato fruits and vegetables is collected by using an 8-channel conductivity detector, and the system framework diagram of the conductivity detector is as shown in Figure 3 The target tomato fruits and vegetables are placed in the central measurement field, 200 measurement data corresponding to the tomato fruits and vegetables are collected within a specified time, the collected data is transmitted to the mobile phone end through Bluetooth, and then transmitted to the computer end for processing, and the corresponding conductivity data is extracted through matlab. The data is composed of 21 frequency bands from 15MHz to 35MHz, and the data of each frequency band is composed of 256 values, which is the phase data θ collected by the conductivity detector converted into corresponding conductivity data, and its definition is as follows:

[0067] V=A+Bi (1)

[0068] θ=arctan(B / A) (2)

[0069] Δσ=-[(S k ) T S k +λR T R] -1 (S k ) T [θ1-θ2] (3)

[0070] Wherein, the real part A is caused by displacement current, which is related to the dielectric constant of the conductor, and the imaginary part B is caused by internal eddy current, which is related to conductivity and excitation frequency. Δσ is the conductivity difference, θ1 is the voltage phase data when the measurement area exists the measured object, and θ2 is the voltage phase data of the measurement area in the air background. Wherein R is the regularization matrix and λ is the regularization parameter.(S k ) T S k is the Hession matrix under the kth iteration.

[0071] S3: The content of tomato fruit acid is determined by refractometer method, specifically including:

[0072] The tomato fruit and vegetable sample to be measured is taken, and the equatorial position is cut, and three times of tomato tissue fluid are taken as the titration sample, each time 1.00g of sample is added to 50.00ml of distilled water for dilution, and fully stirred and mixed. The PAL-EazyACID3 digital handheld tomato acidity meter is used to measure the titration sample of the tomato fruits and vegetables to be measured, the same titration sample is measured three times, and the average value is taken as the measurement result of the titration sample, and the average value of the three titration samples of the same tomato is taken as the measurement result of the content of tomato fruit acid.

[0073] S4: Correlation analysis is performed on the conductivity data obtained by the conductivity detector and the fruit acid of the tomato fruits and vegetables to obtain the frequency band data with the highest correlation between the conductivity data of the tomato fruits and vegetables and the fruit acid content, which specifically includes:

[0074] As can be seen from S2, the shape of the conductivity data obtained by the conductivity detector is a matrix A1 of (200, 21, 256), as shown below:

[0075] A1 = [a1 a2 a3 … a 200 ] (4)

[0076]

[0077] Wherein, a i is the conductivity data of the i-th tomato fruit and vegetable, and its shape is (21, 256), Δσ p,j is the conductivity value of the j-th column under the corresponding frequency band p. Since each tomato fruit and vegetable has corresponding conductivity data of 21 frequency bands, in order to obtain the frequency band data that can optimally represent the highest correlation between the conductivity data and the tomato fruit acid, Pearson correlation degree analysis needs to be performed on each frequency band respectively. The correlation degree calculation formula is as follows:

[0078]

[0079] μ X = E(X), μ Y = E(Y) (7)

[0080]

[0081] Wherein, μ X , μ Y and σX, σY are the expectation and standard deviation. X is a column of conductivity data in the 256 columns of data under the corresponding frequency band, and Y is 200 tomato fruit acid data.

[0082] The tomato fruit acid content is defined as Y = [y1 y2 … y 200 ] T , and the data of the tomato fruit and vegetable under the frequency band i is defined as W = [w i,1 w i,2 … w i,200 ] T , wherein w i,1 represents the conductivity data [Δσ i,1 Δσ i,2 … Δσ i,256 ] of the 1st tomato fruit and vegetable under the frequency band i. The j-th column of conductivity data of the 256 columns of data of the 200 tomato fruits and vegetables under the corresponding frequency band is defined as X = [x 1,j x2,j … x 200,j ] T The frequency band data with the highest total correlation value calculated and counted according to S3.2 is taken as the target conductivity data.

[0083] S5: The frequency band data obtained from S4 is sampled according to the interval and divided into training set and test set according to a certain proportion. Specifically, it includes:

[0084] The conductivity data obtained from S4 is defined as:

[0085]

[0086] The above conductivity data D 200,256 is input as the data set, and the tomato fruit and vegetable acid content data Y in S3 is taken as the data set output. The training set and test set are divided according to 7:3, and 10 consecutive samples are regarded as an independent subset. The 1st, 5th and 10th samples in each subset are selected as the test set, and the rest are taken as the training set.

[0087] S6: A tomato fruit and vegetable acid regression prediction model is constructed using a CNN convolutional neural network, specifically including:

[0088] The convolutional neural network regression prediction model includes 8 convolutional layers and 3 fully connected layers, as shown in Figure 6 . Each convolutional layer uses a 3x1 convolutional kernel, with quantities of 64, 64, 128, 128, 256, 512, and 512, respectively, and all strides are 1. Each convolutional layer is followed by a ReLU activation function, and each pooling layer uses a MaxPool1d function. The convolutional layers and pooling layers included in the model can extract features from the input 1x256 one-dimensional data. Finally, the output layer is a fully connected layer with a single neuron, which maps the extracted features to a single scalar output.

[0089] S7: Based on the prediction model in S6, the conductivity data corresponding to the target obtained by the conductivity detector is input, so as to obtain the acid content of the corresponding tomato. Specifically, it includes:

[0090] The training set in S5 is used to train the prediction model in S6, and the test set in S5 is used to test the performance of the model in S6. The Adam optimizer is used to train the model, the learning rate is set to 0.001, the batch size is set to 16, the maximum number of training generations is set to 500, and the Adam optimizer is used for training. The determination coefficient R 2 and RMSE of the predicted value and the actual value of the test set are 0.849 and 0.028, respectively, as shown in Figure 4 .

[0091] The application discloses a nondestructive detection device and method for acid content of tomatoes based on the principle of magnetic induction tomography.

[0092] The conductivity data acquisition module acquires the conductivity data and the acid content of the fruits and vegetables, and 200 pieces of conductivity data of corresponding tomatoes and fruits and vegetables are acquired within a specified time by using an 8-channel conductivity detector, the data is composed of 21 frequency band data, from 15 MHz to 35 MHz, and each frequency band data is composed of 256 numerical values, and then the conductivity data set is constructed by using the data. In addition, the PAL-EazyACID3 digital handheld tomato acidimeter is used to measure the acid content of the tomatoes.

[0093] The tomato conductivity and acid content correlation analysis module acquires the highest correlation frequency band data by calculating the correlation between the conductivity and the acid of the tomatoes, and the specific method is as follows: since each tomato and fruit and vegetable has corresponding 21 frequency band conductivity data, in order to acquire the frequency band data which can optimally represent the strongest correlation between the conductivity data and the acid of the tomatoes, Pearson correlation degree analysis is required to be performed on each frequency band, and the frequency band data with the highest total correlation value calculated and counted is used as the target conductivity data.

[0094] The acid prediction model construction module specifically comprises the following steps: the acquired conductivity data under the target frequency band is divided into a training set and a test set according to interval sampling and in a 7:3 ratio. A convolutional neural network is used to construct a tomato fruit and vegetable acid regression prediction model, and the test set is input into the corresponding target conductivity data acquired by the conductivity detector, so that the required acid content of the corresponding numbered tomato is obtained. And on this basis, the R 2 and RMSE between the actual value and the measured value are calculated.

[0095] The tomato conductivity data and acid content acquisition module, the tomato conductivity and acid content correlation analysis module and the tomato acid prediction model construction module are sequentially connected. Figure 5

[0096] The content described in the embodiments of the present application is only a list of implementation forms of the inventive concept, and the protection scope of the present application should not be regarded as being limited to the specific forms described in the embodiments, and the protection scope of the present application also extends to equivalent technical means that can be thought of by those skilled in the art according to the inventive concept.​

Claims

1. A non-destructive method for detecting fruit acid content in tomatoes and vegetables based on magnetic induction tomography, comprising the following steps: S1: Build a rapid non-destructive testing device based on the principle of magnetic induction tomography; S2: Collect conductivity data of the corresponding tomato fruits and vegetables using the conductivity meter in the detection device; S3: The content of fruit acids in tomatoes was determined by refractometer method; S4: Perform correlation analysis between conductivity data obtained through a conductivity meter and fruit acid content in tomatoes and vegetables to obtain the frequency band data with the highest correlation between conductivity data and fruit acid content in tomatoes and vegetables; specifically including: S4.1: The matrix A1 with dimensions (200, 21, 256) of the conductivity data obtained by the conductivity meter is shown below: A1=[a1 a2 a3 … a 200 (4) Among them, a i The conductivity data for tomato fruit and vegetable with ID i is in the dimension (21, 256), Δσ p,j This represents the conductivity value in the j-th column of the corresponding frequency band p; S4.2: Since each tomato fruit has corresponding conductivity data for 21 frequency bands, in order to obtain the frequency band data that best represents the highest correlation between conductivity data and tomato fruit acids, it is necessary to perform Pearson correlation analysis on each frequency band. The correlation calculation formula is as follows: m X =E(X),μ Y =E(Y) (7) Where, μ X μ Y σX and σY are the expected value and standard deviation, respectively; X is the conductivity data in one column of 256 data in the corresponding frequency band, and Y is 200 data points of tomato fruit acid. The fruit acid content of tomatoes is defined as Y = [y1 y2 … y 200 ] T Define the data of tomatoes and vegetables in frequency band i as W = [w i,1 w i,2 … w i,200 ] T , where w i,1 This represents the conductivity data of tomato fruit and vegetable No. 1 in frequency band i [△σ]. i,1 △σ i,2 … △σ i,256 ]; Define the conductivity data of the j-th column of 256 columns of data for 200 tomato fruits and vegetables in the corresponding frequency band as X=[x 1,j x 2,j … x 200,j ] T ; S4.3: The frequency band data with the highest total correlation value calculated and statistically analyzed according to S4.2 shall be used as the target conductivity data; S5: The frequency band data obtained from S4 will be sampled from intervals and divided into training and test sets according to a certain ratio; S6: Construct a regression prediction model for fruit acids in tomatoes and vegetables using a CNN convolutional neural network; S7: Based on the prediction model in S6, input the conductivity data of the corresponding target obtained by the conductivity meter to obtain the corresponding fruit acid content of the tomato.

2. The non-destructive detection method for fruit acid content in tomatoes and vegetables based on magnetic induction tomography as described in claim 1, characterized in that, The rapid non-destructive testing device described in step S1 includes: a fixture for fixing the fruits and vegetables to be tested is placed on a conveyor belt, the conveyor belt being circular; the conveyor belt passes through the central hole of the conductivity meter; a motor drives the conveyor belt, the motor being controlled by a microcontroller.

3. The non-destructive detection method for fruit acid content in tomatoes and vegetables based on magnetic induction tomography as described in claim 1, characterized in that, Step S2 specifically includes: collecting conductivity data for 200 corresponding tomato fruits and vegetables within a specified time using an 8-channel conductivity meter. This data consists of data from 21 frequency bands, ranging from 15MHz to 35MHz. Each frequency band consists of 256 values, which are converted from the phase data θ collected by the conductivity meter into the corresponding conductivity data. Its definition is as follows: V = A + Bi (1) θ=arctan(B / A) (2) △σ=-[(S k ) T S k +λR T R] -1 (S k ) T [θ1-θ2] (3) Wherein, the real part A is caused by the displacement current and is related to the dielectric constant of the conductor; the imaginary part B is caused by the internal eddy current and is related to the conductivity and excitation frequency; Δσ is the conductivity difference; θ1 is the voltage phase data when the measured object is present in the measurement area; θ2 is the voltage phase data of the measurement area against an air background; where R is the regularization matrix and λ is the regularization parameter; (S k ) T S k This is the Hession matrix under the k-th iteration.

4. The non-destructive detection method for fruit acid content in tomatoes and vegetables based on magnetic induction tomography as described in claim 1, characterized in that, Step S3 specifically includes: S3.1: Take out the tomato fruit and vegetable sample to be tested, cut it open at the equator, take three portions of tomato tissue fluid as the titration sample, each sample is 1.00g of sample, add 50.00ml of distilled water to dilute, and stir thoroughly. S3.2: The tomato fruit acid content was determined by using a PAL-Eazy ACID3 digital handheld tomato acidity meter. The same titration sample was measured three times, and the average value was taken as the result of the titration sample. The average value of the three titration samples of the same tomato was taken as the result of the tomato fruit acid content.

5. The non-destructive detection method for fruit acid content in tomatoes and vegetables based on magnetic induction tomography as described in claim 1, characterized in that, Step S5 specifically includes: The conductivity data obtained from S5 is defined as follows: The above conductivity data D 200,256 The dataset input is the tomato and vegetable fruit acid content data Y from S4, and the dataset output is the tomato and vegetable fruit acid content data Y. The training set and test set are divided in a 7:3 ratio, and 10 consecutive samples are regarded as an independent subset. The 1st, 5th, and 10th samples in each subset are selected as the test set μ. X =E(X),μ Y =E(Y), and the rest are used as the training set.

6. The non-destructive detection method for fruit acid content in tomatoes and vegetables based on magnetic induction tomography as described in claim 1, characterized in that, Step S6 specifically includes: S6.1: This convolutional neural network regression prediction model includes 8 convolutional layers and 3 fully connected layers; S6.2: Each convolutional layer uses a 3x1 convolutional kernel, with the following numbers: 64, 64, 128, 128, 256, 512, and 512, respectively, and all strides are 1. Each convolutional layer is followed by a ReLU activation function, and each pooling layer uses the MaxPool1d function. The convolutional and pooling layers in this model can extract features from the 1x256 one-dimensional input data. Finally, the output layer is a fully connected layer with a single neuron, which maps the extracted features to a single scalar output.

7. The non-destructive detection method for fruit acid content in tomatoes and vegetables based on magnetic induction tomography as described in claim 1, characterized in that, Step S7 specifically includes: using the training set from S5 to train the prediction model in S5, and using the test set from S5 to test the performance of the model in S6; and based on the determination coefficient R between the predicted and actual values ​​of the test set... 2 RMSE is used as an indicator to determine the effectiveness of the model.

Images