CN-121994787-A - On-line monitoring method for biological envelope of perlechtendella salina based on hyperspectral technology

Abstract

The invention discloses a method for on-line monitoring of a biological envelope of Borowth bacteria based on a hyperspectral technology, which belongs to the technical field of food detection, and comprises the steps of firstly collecting microscopic hyperspectral data of different growth stages of the biological envelope in a pure culture system, preprocessing, and then adopting methods of analysis of variance, LASSO, SPA or PLS-VIP and the like to carry out compound feature screening to extract key feature wavelengths; and then migrating the model with effective verification to a food system, carrying out stage identification on the biological film on the surface of the food, and finally generating a visual image reflecting the spatial distribution of the biological film. The invention realizes reliable migration from a laboratory to an actual application scene, and can perform nondestructive, dynamic and visual online identification and spatial analysis on the growth stage of the biofilm.

Inventors

• CHEN QINGMIN
• WANG YILIN
• LI DONG
• LI XIN
• CHEN QUANSHENG

Assignees

• 集美大学

Dates

Publication Date

20260508

Application Date

20260130

Claims (10)

1. 1. The on-line monitoring method for the biological envelope of the pergola based on the hyperspectral technology is characterized by comprising the following steps of: s1, acquiring microscopic hyperspectral data of a biological film of the Porochalimasch in different growth stages under a pure culture system, and preprocessing; S2, performing feature screening on the preprocessed microscopic hyperspectral data, wherein the feature screening comprises performing primary screening by adopting analysis of variance, and performing secondary screening by adopting at least one of minimum absolute shrinkage and selection operator, continuous projection algorithm and variable projection importance analysis method to obtain key feature wavelengths; S3, constructing a data set and training at least one qualitative classification model based on the screened key characteristic wavelength so as to classify different growth stages of the biofilm in the pure culture system; S4, transferring the qualitative classification model which is validated by training in the step S3 to be applied to the collected biological film hyperspectral data under a food system so as to classify the growth stage of the biological film on the surface of the food; s5, fusing the classification result with the space coordinate information of the hyperspectral image to generate a visual distribution map reflecting the space distribution of the biofilm growth stage.
2. 2. The method according to claim 1, wherein in step S1, the preprocessing includes Savitzky-Golay smoothing filtering and first order differential processing.
3. 3. The method according to claim 1, wherein in step S2, the minimum absolute shrinkage and selection operator employs L1 regularization with a loss function of: Wherein, the In order to be a sample tag, Is the first Sample number The characteristics of the device are that, Is the first The coefficients of the individual features are used to determine, In order for the parameters to be regularized, In order to obtain the number of samples, Is the feature number.
4. 4. The method according to claim 1, wherein in step S2, the continuous projection algorithm iterates the selection of variables by calculating the orthogonal projection residuals of the remaining variables over the selected feature subspace, the calculation formula of the orthogonal projection residuals being: Wherein, the The selected feature vectors are spread into orthogonal bases of space, The modulus of (2) represents the variable Information content not covered by the selected feature.
5. 5. The method according to claim 1, wherein in step S2, the calculation formula of the variable importance index VIP value of the variable projection importance analysis method is: Wherein, the As a total number of variables, For the cumulative interpretation rate of all potential components versus dependent variables, As a number of potential components, Is the first The rate of interpretation of the dependent variable by the individual potential components, Is the first The variable is at the first Weights on the individual potential components, VIP values are the combined contribution of the variables to model interpretation ability.
6. 6. The method according to claim 1, wherein in step S3, the qualitative classification model comprises a support vector machine, a random forest, a nearest neighbor model, a linear discriminant analysis.
7. 7. The method according to claim 1, wherein in step S4 the food system is an aquatic product surface, in particular a large yellow croaker chunk surface.
8. 8. The method according to claim 1, wherein the black-and-white correction process is performed after the hyperspectral image is acquired in step S1 and step S4.
9. 9. The method according to claim 1, wherein in step S5, the visual distribution map indicates different growth stages and levels of biofilm in different colors.
10. 10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1-9.

Description

On-line monitoring method for biological envelope of perlechtendella salina based on hyperspectral technology Technical Field The invention belongs to the technical field of food detection, and particularly relates to an online monitoring method for a biological envelope of pergola based on a hyperspectral technology. Background In the technical field of food detection, particularly for monitoring pathogen pollution in perishable foods such as aquatic products, the monitoring of a biofilm is a key link for guaranteeing food safety and prolonging shelf life. In the prior art, conventional microbiological and biochemical methods are generally used for detecting and quantifying the biofilm, wherein the crystal violet staining quantification method is most commonly used. The method can more stably quantify the total biomass of the biological film by dyeing, eluting and photometrically measuring the biological film, and is one of standard means for evaluating the forming capacity and dynamic change of the bacterial biological film in the current laboratory and industrial scenes. The method has the advantages of standardized operation and relatively low cost, can be used in combination with traditional microorganism detection means such as culture counting, has certain applicability in scientific research and conventional detection, and provides basic data support for understanding the static formation rule of the biofilm. However, the above prior art methods have significant drawbacks. Firstly, chemical detection methods such as a crystal violet method and the like belong to in-vitro and destructive detection, sample culture is required to be stopped, and samples are consumed, so that continuous and non-invasive on-line monitoring of the same monitoring object cannot be realized, and therefore, dynamic process and spatial heterogeneity information of the evolution of a biological film on the surface of food along with time are difficult to capture. Secondly, the method can only provide a single index of the total amount of the biological film, can not distinguish different growth stages (such as adhesion, maturation, diffusion and the like) of the biological film, can not reflect the microstructure and the distribution characteristics of the biological film on the surface of the food substrate, leads to single dimension of monitoring information, and can not meet the requirements on accurate intervention and process control of the biological film. In addition, the existing method has the influence of interfering substances in complex food matrixes (such as the surface of fish meat), the detection sensitivity and specificity may be reduced, the operation flow is complex, the time consumption is long, and the technical requirements of the food processing and storage links on the rapid and automatic online monitoring are difficult to adapt. Therefore, there is a need to develop an on-line monitoring technology that can achieve nondestructive, dynamic, visual, and accurate identification of biofilm growth phases. Disclosure of Invention In order to solve the technical problems, the invention provides an online monitoring method for a biological envelope of the pergola based on a hyperspectral technology, so as to solve the problems in the prior art. To achieve the above object, in a first aspect, the present invention provides a method for on-line monitoring of a biological capsule of perusal bacteria based on hyperspectral technology, comprising: s1, acquiring microscopic hyperspectral data of a biological film of the Porochalimasch in different growth stages under a pure culture system, and preprocessing; S2, performing feature screening on the preprocessed microscopic hyperspectral data, wherein the feature screening comprises performing primary screening by adopting analysis of variance, and performing secondary screening by adopting at least one of minimum absolute shrinkage and selection operator, continuous projection algorithm and variable projection importance analysis method to obtain key feature wavelengths; S3, constructing a data set and training at least one qualitative classification model based on the screened key characteristic wavelength so as to classify different growth stages of the biofilm in the pure culture system; S4, transferring the qualitative classification model which is validated by training in the step S3 to be applied to the collected biological film hyperspectral data under a food system so as to classify the growth stage of the biological film on the surface of the food; s5, fusing the classification result with the space coordinate information of the hyperspectral image to generate a visual distribution map reflecting the space distribution of the biofilm growth stage. Preferably, in step S1, the preprocessing includes Savitzky-Golay smoothing filtering and first order differential processing. Preferably, in step S2, the minimum absolute shrinkage and selection operator is regular