CN-121994748-A - Near infrared spectrum detection device and method for fresh glucose

CN121994748ACN 121994748 ACN121994748 ACN 121994748ACN-121994748-A

Abstract

The invention relates to the technical field of intelligent detection, and provides a device and a method for detecting a near infrared spectrum of fresh glucose, wherein the device for detecting the near infrared spectrum of the fresh glucose comprises a spectrum acquisition module, a controller module, a storage module and a cloud platform module; the spectrum acquisition module comprises an LED array, a signal conditioning unit, a micro spectrometer and an electric signal processing unit, the controller module comprises an MCU, the MCU is connected with the electric signal processing unit, the storage module and the cloud platform module, the storage module is used for storing a plurality of sample spectrum data and current sugar degree detection general models, and the cloud platform module is used for constructing the latest sugar degree detection general models. By adopting the structure, the on-site nondestructive sugar degree detection can be efficiently carried out on the fresh grape, and the accuracy of the detection result is effectively improved.

Inventors

XIAO HUAICHUN
LUO JIANG
CHEN SICONG
HUANG HEXIANG
TIAN YONGLE
YAO JINKUI
Gong meng

Assignees

东华理工大学南昌校区

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The near infrared spectrum detection device for the fresh glucose level is characterized by comprising a spectrum acquisition module, a controller module, a storage module and a cloud platform module, wherein the spectrum acquisition module is connected with the controller module, and the controller module is connected with the storage module and the cloud platform module; The spectrum acquisition module comprises an LED array, a signal conditioning unit, a micro spectrometer and an electric signal processing unit, wherein the LED array is used for emitting light with the wavelength of 650 nm-1050 nm so as to irradiate a white board and fresh grapes, the micro spectrometer is used for receiving reflected light of the white board and the fresh grapes and converting the reflected light into a first reference electric signal and a first detection electric signal, the signal conditioning unit is connected with the micro spectrometer so as to adjust the temperature drift of the micro spectrometer and convert the first reference electric signal and the first detection electric signal into a second reference electric signal and a second detection electric signal, the electric signal processing unit is connected with the signal conditioning unit so as to convert the second reference electric signal into a reference digital signal, convert the second detection electric signal into a detection digital signal and output sample spectrum data based on the reference digital signal and the detection digital signal; the controller module comprises an MCU, and the MCU is connected with the electric signal processing unit, the storage module and the cloud platform module; The storage module is used for storing a plurality of sample spectrum data and a current sugar degree detection universal model; The cloud platform module is used for acquiring a plurality of variety spectrum data, generating a generalization spectrum data set based on the variety spectrum data, and generating a general training set based on the variety spectrum data and the generalization spectrum data set so as to construct a latest sugar degree detection general model according to the general training set.
2. The near infrared spectrum detection device for the fresh glucose level according to claim 1, wherein the spectrum acquisition module further comprises a sample tray, the sample tray is used for supporting the fresh grape, the LED array comprises a plurality of square LED light sources which are arranged around the sample tray, and the LED array emits light in a PWM dimming mode.
3. The near infrared spectrum detection device for the fresh glucose level according to claim 1, wherein the signal conditioning unit comprises a voltage follower, a conditioning power supply, a conditioning capacitor and a conditioning resistor, the voltage follower is connected with the conditioning power supply and the conditioning capacitor, and the conditioning resistor is connected between the voltage follower and the electric signal processing unit.
4. The apparatus according to claim 1, wherein the step of outputting sample spectral data based on the reference digital signal and the detection digital signal comprises: Obtaining total reflection light intensity based on the reference digital signal analysis, and obtaining sample light intensity based on the detection digital signal analysis; Obtaining dark current light intensity; And obtaining sample absorbance based on the sample light intensity, the total reflection light intensity and the dark current light intensity, and obtaining sample spectrum data based on the sample absorbance analysis.
5. The near infrared spectrum detection device for the fresh glucose level according to claim 1, wherein the MCU is further connected with a communication module, the communication module comprises a serial communication unit and a Bluetooth communication unit, the serial communication unit is used for connecting the MCU and an upper computer, the Bluetooth communication unit is used for wireless transmission, and the communication module is connected with the cloud platform module.
6. The apparatus of claim 1, wherein the step of generating a generalized spectral dataset based on a plurality of the variety spectral data comprises: setting a generalization translation distance, and translating the variety spectrum data along a wavelength axis according to the generalization translation distance to obtain translation spectrum data; Setting a generalization signal-to-noise ratio, acquiring generalization Gaussian noise, and adding the generalization Gaussian noise into the variety spectrum data to acquire noise spectrum data; and forming a plurality of translation spectrum data and a plurality of noise spectrum data into a generalization spectrum data set.
7. The apparatus according to claim 1, wherein the step of generating a general training set based on the plurality of variety spectrum data and the generalized spectrum data set to construct a latest sugar degree detection general model from the general training set comprises: Obtaining a master spectrum matrix based on a plurality of variety spectrum data, and obtaining a slave spectrum matrix based on the generalization spectrum data set; performing global linear transformation based on the inverse of the slave spectrum matrix and the master spectrum matrix to obtain a global correction matrix; Forming an initial training set by a plurality of variety spectrum data and the generalization spectrum data set, and performing matrix multiplication on the initial training set through the global correction matrix to generate a general training set; Based on the general training set, a latest sugar degree detection general model is trained by adopting a PLSR algorithm and a 1D-CNN algorithm.
8. The near infrared spectrum detection device of a fresh glucose level according to claim 1, wherein the controller module is connected to a power module, and the power module is used for supplying power.
9. The near infrared spectrum detection device for the fresh glucose level according to claim 1, wherein the MCU is connected with a display module, the display module comprises a serial screen and a touch screen unit, the serial screen is used for displaying data and sample spectrum, and the touch screen unit is used for sending touch screen operation instructions to the MCU.
10. A method for detecting a fresh glucose level by using the near infrared spectrum detection device for the fresh glucose level according to any one of claims 1 to 9, wherein the method comprises the following steps: The method comprises the steps that a white board is irradiated by light with the wavelength of 650-1050 nm through an LED array, the micro spectrometer receives reflected light of the white board and converts the reflected light into a first reference electric signal, the first reference electric signal is converted into a second reference electric signal through a signal conditioning unit, and the second reference electric signal is transmitted to an electric signal processing unit so as to obtain a reference digital signal; The method comprises the steps that light with the wavelength of 650-1050nm is emitted by the LED array to irradiate fresh grapes, the micro spectrometer receives reflected light of the fresh grapes and converts the reflected light into a first detection electric signal, the first detection electric signal is converted into a second detection electric signal by the signal conditioning unit, the second detection electric signal is transmitted to the electric signal processing unit, so that a detection digital signal is obtained, and sample spectrum data is output based on the reference digital signal and the detection digital signal; receiving the sample spectrum data through an MCU in a controller module, and transmitting the sample spectrum data to a storage module and a cloud platform module; Judging whether a current sugar degree detection general model is stored in the storage module, if so, inputting the sample spectrum data into the current sugar degree detection general model to output a sugar degree detection result of the fresh glucose; If the current sugar degree detection general model is not stored in the storage module, acquiring a plurality of variety spectrum data through the cloud platform module, generating a generalized spectrum data set based on the variety spectrum data, generating a general training set based on the variety spectrum data and the generalized spectrum data set, constructing a latest sugar degree detection general model according to the general training set, determining the latest sugar degree detection general model as the current sugar degree detection general model, transmitting the current sugar degree detection general model to the storage module, inputting the sample spectrum data into the current sugar degree detection general model, and outputting the sugar degree detection result of the fresh glucose.

Description

Near infrared spectrum detection device and method for fresh glucose Technical Field The invention relates to the technical field of intelligent detection, in particular to a device and a method for detecting a near infrared spectrum of fresh glucose. Background Fresh grapes are rich in various nutrients such as saccharides, organic acids, vitamins and the like, and are favored by consumers by virtue of bright color, unique flavor and rich nutritional value. Along with the increasing promotion of living standard, the requirements on the quality of the fresh grapes are also the water-rise ship height. Sugar degree is one of the key indexes for measuring the quality of the food, and is directly related to the taste and flavor of the food. The traditional detection method for the fresh glucose degree mainly depends on manual identification, has the problems of strong subjectivity and low discrimination rate, and most of the current methods are used for destroying samples to different degrees through laboratory chemical means and a glycometer, and the method is high in cost and difficult to meet the requirement of large-scale on-site detection. In the prior art, the on-site nondestructive sugar degree detection of fruits generally adopts a near infrared spectrum technology, diffuse reflection light of the fruits is collected through a spectrometer, and the sugar degree is predicted by combining a deep learning model, however, the existing detection device adopting the near infrared spectrum technology has the problems of larger error, poor universality, lower precision, lack of targeted setting on thin-skin thick-skin fruits and insufficient intelligent degree, and the existing deep learning model based on the near infrared spectrum has systematic difference in spectrum when facing fresh grapes in different varieties, production places and growing environments, so that the model has weak generalization capability and large detection error, and the problem of limited spectral data volume further restricts the improvement of model performance. Disclosure of Invention Aiming at the defects of the prior art, the invention aims to provide the near infrared spectrum detection device for the fresh glucose level, which can realize the calibration of the sample spectrum of the fresh grape through a reference value by a hardware module and a software module, and can be used for carrying out targeted light source setting and interconnection with a cloud platform so as to realize the update iteration of a general model for the glucose level detection. The invention aims to solve the technical problems of large detection error, poor universality, insufficient precision, low intelligent degree, weak model generalization capability and limited spectrum data volume of the fresh glucose degree of the fruit sugar degree detection device in the prior art. In order to achieve the above object, the present invention is achieved by the following technical scheme: The near infrared spectrum detection device for the fresh glucose comprises a spectrum acquisition module, a controller module, a storage module and a cloud platform module, wherein the spectrum acquisition module is connected with the controller module, and the controller module is connected with the storage module and the cloud platform module; The spectrum acquisition module comprises an LED array, a signal conditioning unit, a micro spectrometer and an electric signal processing unit, wherein the LED array is used for emitting light with the wavelength of 650 nm-1050 nm so as to irradiate a white board and fresh grapes, the micro spectrometer is used for receiving reflected light of the white board and the fresh grapes and converting the reflected light into a first reference electric signal and a first detection electric signal, the signal conditioning unit is connected with the micro spectrometer so as to adjust the temperature drift of the micro spectrometer and convert the first reference electric signal and the first detection electric signal into a second reference electric signal and a second detection electric signal, the electric signal processing unit is connected with the signal conditioning unit so as to convert the second reference electric signal into a reference digital signal, convert the second detection electric signal into a detection digital signal and output sample spectrum data based on the reference digital signal and the detection digital signal; the controller module comprises an MCU, and the MCU is connected with the electric signal processing unit, the storage module and the cloud platform module; The storage module is used for storing a plurality of sample spectrum data and a current sugar degree detection universal model; The cloud platform module is used for acquiring a plurality of variety spectrum data, generating a generalization spectrum data set based on the variety spectrum data, and generating a general training set based on the vari