CN-115758779-B - Photoelectric detection model transmission sharing method of cloud service and Internet of things monitoring and evaluating system

CN115758779BCN 115758779 BCN115758779 BCN 115758779BCN-115758779-B

Abstract

The invention discloses a cloud service photoelectric detection model transmission sharing method and an Internet of things monitoring and evaluating system, wherein the transmission sharing method comprises the steps of correcting spectrum information of an agricultural product sample by calling a temperature compensation model and a spectrum transmission model, and calculating the corrected spectrum information by calling a detection model to obtain a detection result of the agricultural product sample; and the monitoring and evaluating system of the Internet of things uses a spectrum transfer sharing method to correct spectrum information, then calls a detection model to calculate, and the detection result is transmitted back to a detection terminal in real time, so that remote monitoring and evaluating of the quality of agricultural products are realized. The invention can realize the sharing of the detection model in different detection terminals, and has wide application prospect in the sampling inspection and evaluation of the quality of agricultural products.

Inventors

GUO ZHIMING
WANG JUNYI
ZOU XIAOBO
CAI JIANRONG
SHI JIYONG
YIN LIMEI

Assignees

江苏大学

Dates

Publication Date: 20260512
Application Date: 20221129

Claims (7)

1. A photoelectric detection model transmission sharing method of cloud service is characterized in that: Acquiring spectrum information and temperature information of typical and representative samples of agricultural products in batches by using a No. 0 detection terminal, and establishing a temperature compensation model; Acquiring quality indexes of the agricultural product representativeness and representative samples, extracting characteristic wavelengths based on the spectrum after wavelength correction, and establishing a detection model by the characteristic wavelengths and the quality indexes; acquiring spectrum information of typical and representative samples of batch agricultural products by using a No. 0 detection terminal and a No. 1 detection terminal, and establishing a spectrum transfer model by using a self-encoder neural network; correcting the spectrum information of the agricultural product sample by calling a temperature compensation model and a spectrum transfer model, and calculating the corrected spectrum information by calling a detection model to obtain a detection result of the agricultural product sample; the process for establishing the spectrum transfer model by utilizing the self-encoder neural network comprises the following steps: The encoder encodes the detection terminal spectrum matrix S 1 into a hidden variable h with low dimensionality, and the decoder restores the hidden variable h of the hidden layer to the detection terminal spectrum matrix S 0 with No. 0, so that the neural network learns the difference characteristics between the matrix S 0 and the matrix S 1 , and the matrix S 1 is corrected to be a matrix S 0 ; The encoding process from the input layer to the hidden layer is: the decoding process from the hidden layer to the output layer is: Wherein: for the weight matrix of the encoding process, For the weight matrix of the decoding process, As a function of the encoding process, As a function of the decoding process, In order to provide a deviation matrix for the encoding process, In order to be a bias matrix for the decoding process, Representing a spectral matrix; The spectrum transfer model comprises an encoder and a decoder, wherein the encoder comprises a weight matrix Deviation matrix The decoder comprises a weight matrix Deviation matrix ; The photoelectric detection model transmission sharing method further comprises the steps that when a new detection terminal is added, the new detection terminal is used for acquiring spectrum information of agricultural product samples in a small batch mode, and the spectrum transmission model is updated by using a transfer learning method; Freezing a parameter matrix of a self-encoder neural network updated based on a No. 0 detection terminal and a No. 1 detection terminal, adding a decoder, and updating the parameter matrix of the added decoder by using the No. 0 terminal and the new detection terminal; generating a sample serial number of the agricultural product while obtaining a detection result, extracting a sample according to the serial number for actual measurement, calculating an error between the detection result and the actual measurement result, and if the error exceeds a set threshold value, updating a detection model; The detection model updating is carried out by using an active feedback mechanism and a passive feedback mechanism, wherein the active feedback mechanism is characterized in that representative samples are selected for carrying out the detection model active updating when key time nodes of agricultural products are after harvesting, before entering a warehouse and before marketing, the passive feedback mechanism is characterized in that sample numbering is carried out according to the detection quantity in the detection process, a certain quantity of samples are dynamically extracted and set as an independent verification set for verifying the detection model, and if the error between the detection result and the actual measurement result is larger than a preset threshold value, the independent verification set is utilized for carrying out the model updating.
2. The method for sharing the photoelectric detection model according to claim 1, wherein the model updating is performed by selecting representative agricultural product samples from independent verification sets and adding the agricultural product samples to a training set of the built detection model.
3. The method for sharing the transmission of the photoelectric detection model according to claim 1, wherein the wavelength in the spectrum information acquired by the detection terminal is corrected by a wavelength calibration equation.
4. The method for sharing the transmission of the photoelectric detection model according to claim 3, wherein the wavelength calibration equation is obtained by selecting a characteristic wavelength with a characteristic absorption peak in advance, obtaining a spectrum of the detection terminal by using a standard light source, and calibrating the characteristic wavelength.
5. The utility model provides an thing networking monitoring evaluation system which characterized in that includes: the detection terminal is used for acquiring information of an agricultural product sample to be detected, including spectrum information, temperature information and geographic information; The data server uses the transmission sharing method of any one of claims 1-4 to calibrate spectrum information, then calls a detection model to calculate, and the detection result is transmitted back to a detection terminal in real time; and the cloud data management platform is used for displaying detection results and controlling the detection terminal.
6. The internet of things monitoring and evaluating system according to claim 5, wherein the detection terminal comprises a light source, a sensor, a positioning module and a control module, the light source is used for providing an active light source required for detecting agricultural product samples, the sensor comprises a photoelectric sensor and a temperature sensor, the positioning module is used for acquiring geographic information of the position of the detection terminal, and the control module controls the operation of the whole detection terminal and has functions of photoelectric signal conversion, data display and transmission.
7. The internet of things monitoring and evaluating system of claim 5, wherein the detection terminal comprises one or more of a handheld, portable, vehicle-mounted, and online combination.

Description

Photoelectric detection model transmission sharing method of cloud service and Internet of things monitoring and evaluating system Technical Field The invention relates to the field of agricultural product quality sorting monitoring, in particular to a cloud service photoelectric detection model transmission sharing method and an Internet of things monitoring and evaluating system. Background With the upgrading of consumption demands, people pay more attention to the quality safety of agricultural products after meeting the demand of the quantity of the agricultural products. Because agricultural products have certain natural characteristics, obvious quality differences exist among individuals. If mixed sales are adopted, the multi-level demand preference of consumers cannot be met, and the high-quality price of agricultural products cannot be realized. And in the product homogenization environment, malignant competition among enterprises is easier to cause, and the healthy development of the whole agricultural product industry is not facilitated. Therefore, the quality classification of the agricultural products can realize the quality and price of the agricultural products and meet the multi-level demands of consumers, and has important significance for promoting the effective matching of the supply and demand of the agricultural products and reducing the resource waste. In terms of quality detection inside agricultural products, spectroscopic analysis techniques exhibit great advantages. The spectrum analysis technology has the advantages of no damage, high detection efficiency, low cost, good reproducibility, no pretreatment for sample measurement, suitability for field detection and online analysis, and the like. However, the spectral analysis first requires the establishment of a spectral detection model between the spectral matrix X and the response matrix Y. However, the spectra collected on different instruments are different due to the difference of the spectrometers themselves, the influence of the aging of the instruments or the environmental change. The established quantitative analysis model has the following challenges in practical application that (1) the analysis model established on one instrument cannot be used for a long time, (2) the analysis model established on one instrument cannot directly conduct quantitative analysis on a spectrum acquired on the other instrument, and (3) a large amount of manpower, material resources and financial resources are consumed in reestablishing a new model due to the fact that certain samples are toxic, unstable in chemical properties, high in price and the like. At present, most of agricultural product sorting equipment designed based on a spectrum analysis technology is online, monitoring requirements are difficult to meet, and a single equipment is independently modeled, so that a large amount of manpower and material resources are wasted. With the rise of the Internet of things, big data and cloud computing, the spectrometer is taken as a photoelectric sensor to be a natural component part of the Internet of things, and the spectrum analysis technology and the information technology are combined, so that the quality in-situ detection or monitoring of agricultural products can be realized, and the system can serve the national agricultural departments, industry taps, national supervision departments and the like. In addition, the Internet of things provides an implementation way for sharing the detection model, and becomes an important motive power for commercialized popularization of detection equipment. In addition, along with the development of futures markets, the futures army of China is increasingly rich and various varieties are increasingly wide, and the futures varieties of agricultural products are not lacked in all fields. For example, apple futures are put on sale and traded in Zhengzhou commodity exchange, and become the first fresh fruit futures in China and the world. The method has the advantages that the method has a plurality of reasons for pushing out fruit futures, china is the country with the largest production and consumption of the fruits, the planting area is wider, the yield is larger, but the situation that the fruits hurt farmers easily caused by huge supply is also positive, the fruit futures are marketed, related enterprises and growers can carry out set-time value preservation, price discovery and risk avoidance as much as possible through the apple futures, meanwhile, the price fluctuation of the fruits has certain periodicity, the amplitude is obvious, the market scale is larger, the method has the basic advantage of pushing out the futures, and the method is also an important factor of successful fruit futures marketing and stable operation in China. However, fruit futures are currently standardized to a lower degree. For example, apple appearance phase is not exactly positively correlated with internal quality, but