CN-121476478-B - Multi-source data fusion tobacco shred detection method, device, equipment, medium and product

CN121476478BCN 121476478 BCN121476478 BCN 121476478BCN-121476478-B

Abstract

The application relates to a tobacco shred detection method, device, equipment, medium and product for multi-source data fusion. The method comprises the steps of obtaining first detection data and second detection data of tobacco shreds to be detected, wherein the first detection data are data obtained by detecting through a gas chromatography-surface acoustic wave detector, the second detection data are data obtained by detecting through an electronic nose sensor array, feature fusion is conducted on the basis of the first detection data and the second detection data to obtain fusion features, the fusion features are used as input of a target classification model, and tobacco shred detection is conducted through the target classification model to obtain tobacco shred detection results. The method can improve the efficiency of detecting the smell of the cut tobacco.

Inventors

ZHOU TAO
CHEN LONG
TANG NI
MA JIAN
HE ZONGWEI
WU XIAO
HE CHUAN
XIE LIHUA
WANG YANG

Assignees

四川中烟工业有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260107

Claims (10)

1. A method for detecting cut tobacco by multi-source data fusion, which is characterized by comprising the following steps: acquiring first detection data and second detection data of tobacco shreds to be detected, wherein the first detection data are data obtained by detecting through a gas chromatograph-surface acoustic wave detector, and the second detection data are data obtained by detecting through an electronic nose sensor array; Performing feature fusion based on the first detection data and the second detection data to obtain fusion features; taking the fusion characteristic as input of a target classification model, and detecting tobacco shreds by the target classification model to obtain a tobacco shred detection result; the training process of the target classification model comprises the following steps: Acquiring sample fusion characteristics of sample tobacco shreds and a reference tobacco shred detection result, wherein the sample fusion characteristics are obtained by fusion based on first detection data and second detection data of the sample tobacco shreds; Taking the sample fusion characteristics as the input of a plurality of classification models, and respectively detecting tobacco shreds by the plurality of classification models to obtain predicted tobacco shred detection results corresponding to the classification models, wherein the plurality of classification models correspond to different model parameters; determining the classification accuracy of each classification model based on the reference tobacco shred detection result and the prediction tobacco shred detection result corresponding to each classification model; iterating based on the classification accuracy of each classification model and a plurality of groups of model parameters corresponding to a plurality of classification models to obtain the target classification model; The obtaining the sample fusion characteristic of the sample tobacco shreds comprises the following steps: Acquiring first detection data and second detection data of the sample cut tobacco; performing feature sampling in data features corresponding to the first detection data and the second detection data to obtain a first feature population, wherein the first feature population comprises a plurality of feature subsets, and each feature subset comprises a plurality of data features; And carrying out feature iteration by a plurality of feature subsets in the first feature population according to the classification fitness of the feature subsets to obtain the sample fusion features of the sample tobacco shreds.
2. The method according to claim 1, wherein the iterating based on the classification accuracy of each classification model and a plurality of sets of model parameters corresponding to a plurality of classification models to obtain the target classification model includes: Constructing a1 st model population based on a plurality of the classification models; determining a genetic classification model of a kth round from a plurality of classification models based on classification accuracy of each classification model in a kth model population, wherein k is a positive integer; Performing parameter crossing and parameter variation based on model parameters of the k-th round genetic classification model to obtain a k+1th model population; if k is equal to M, determining a target classification model with highest classification accuracy from the Mth model population, wherein M is a positive integer greater than 1; If k is smaller than M, adding 1 to k, and returning to execute classification accuracy based on each classification model in the kth model population, and determining a kth round of genetic classification model from a plurality of classification models.
3. The method of claim 1, wherein the performing feature iteration from the plurality of feature subsets in the first feature population according to the classification fitness of the feature subsets to obtain the sample fusion feature of the sample cut tobacco comprises: determining a genetic feature subset of the ith round from a plurality of feature subsets based on the classification fitness of the plurality of feature subsets in the ith feature population; Performing feature crossing and feature mutation operation based on the genetic feature subsets to obtain an (i+1) th feature population; If i is equal to N, determining a target feature subset with the highest corresponding classification fitness based on the classification fitness of a plurality of feature subsets in the Nth feature population, and obtaining sample fusion features of the sample tobacco shreds; And if i is smaller than N, adding 1 to i, and returning to execute classification fitness based on a plurality of feature subsets in the ith feature population, determining a genetic feature subset of the ith round from the plurality of feature subsets, wherein i is a positive integer, and N is a positive integer greater than 1.
4. The method of claim 1, wherein the first detection data is a response value sequence detected by a surface acoustic wave sensor after chromatographic column separation by the gas chromatograph-surface acoustic wave detector, and the second detection data is an electrical signal response map generated by the electronic nose sensor array.
5. A multi-source data fusion cut tobacco detection device, the device comprising: The device comprises an acquisition module, a detection module and a detection module, wherein the acquisition module is used for acquiring first detection data and second detection data of tobacco shreds to be detected, the first detection data are data obtained by detection through a gas chromatograph-surface acoustic wave detector, and the second detection data are data obtained by detection through an electronic nose sensor array; The feature fusion module is used for carrying out feature fusion based on the first detection data and the second detection data to obtain fusion features; The detection module is used for taking the fusion characteristics as the input of a target classification model, and detecting tobacco shreds by the target classification model to obtain a tobacco shred detection result; The model training module is used for acquiring sample fusion characteristics of sample tobacco shreds and reference tobacco shred detection results, wherein the sample fusion characteristics are obtained by fusion based on first detection data and second detection data of the sample tobacco shreds, the sample fusion characteristics are used as input of a plurality of classification models, tobacco shred detection is carried out by the classification models respectively to obtain predicted tobacco shred detection results corresponding to the classification models, and the classification models correspond to different model parameters; the characteristic selection module is used for acquiring first detection data and second detection data of the sample tobacco shreds, performing characteristic sampling on data characteristics corresponding to the first detection data and the second detection data respectively to obtain a first characteristic population, wherein the first characteristic population comprises a plurality of characteristic subsets, each characteristic subset comprises a plurality of data characteristics, and the characteristic subsets in the first characteristic population perform characteristic iteration according to classification fitness of the characteristic subsets to obtain sample fusion characteristics of the sample tobacco shreds.
6. The device of claim 5, wherein the model training module is further configured to construct a1 st model population based on a plurality of the classification models, determine a kth round of genetic classification models from the plurality of classification models based on classification accuracy of each classification model in the kth model population, wherein k is a positive integer, perform parameter crossover and parameter mutation based on model parameters of the kth round of genetic classification models to obtain a k+1th model population, determine a target classification model with highest classification accuracy from the Mth model population if k is equal to M, wherein M is a positive integer greater than 1, add 1 to k if k is less than M, and return to perform classification accuracy based on each classification model in the kth model population to determine a kth round of genetic classification models from the plurality of classification models.
7. The device of claim 5, wherein the feature selection module is further configured to determine an ith round of genetic feature subsets from the feature subsets based on classification fitness of the feature subsets in the ith feature population, perform feature crossover and feature mutation operations based on the genetic feature subsets to obtain an ith+1th feature population, determine a target feature subset with a highest corresponding classification fitness based on the classification fitness of the feature subsets in the nth feature population if i is equal to N, obtain a sample fusion feature of the sample tobacco, and add 1 if i is less than N, and return to perform classification fitness based on the feature subsets in the ith feature population to determine the ith round of genetic feature subsets from the feature subsets, where i is a positive integer, and N is a positive integer greater than 1.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 4 when the computer program is executed.
9. 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 of any of claims 1 to 4.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the method of any of claims 1 to 4.

Description

Multi-source data fusion tobacco shred detection method, device, equipment, medium and product Technical Field The application relates to the technical field of gas detection, in particular to a method, a device, equipment, a medium and a product for detecting cut tobacco by multi-source data fusion. Background The odor quality of the finished cut tobacco is one of the core indexes for determining the final sensory quality and brand style of the cigarette product. At present, the evaluation and classification of the smell of the finished tobacco shreds mainly depend on a subjective sensory evaluation method, wherein the method relies on professional evaluation staff trained for a long time to describe and score the smell of the tobacco shreds in a manual sniffing mode. However, the subjective sensory evaluation method has the problems of strong subjectivity, high cost and low efficiency, so that a tobacco shred detection method is needed to realize quick detection of tobacco shreds. Disclosure of Invention In view of the foregoing, it is desirable to provide a tobacco shred detection method, apparatus, device, medium and product with multi-source data fusion that can improve tobacco shred smell detection efficiency. In a first aspect, the present application provides a method for detecting cut tobacco by multi-source data fusion, including: acquiring first detection data and second detection data of tobacco shreds to be detected, wherein the first detection data are data obtained by detecting through a gas chromatograph-surface acoustic wave detector, and the second detection data are data obtained by detecting through an electronic nose sensor array; Performing feature fusion based on the first detection data and the second detection data to obtain fusion features; And taking the fusion characteristic as input of a target classification model, and detecting the cut tobacco by the target classification model to obtain a cut tobacco detection result. In one embodiment, the training process of the object classification model includes: Acquiring sample fusion characteristics of sample tobacco shreds and a reference tobacco shred detection result, wherein the sample fusion characteristics are obtained by fusion based on first detection data and second detection data of the sample tobacco shreds; Taking the sample fusion characteristics as the input of a plurality of classification models, and respectively detecting tobacco shreds by the plurality of classification models to obtain predicted tobacco shred detection results corresponding to the classification models, wherein the plurality of classification models correspond to different model parameters; determining the classification accuracy of each classification model based on the reference tobacco shred detection result and the prediction tobacco shred detection result corresponding to each classification model; and iterating based on the classification accuracy of each classification model and a plurality of groups of model parameters corresponding to a plurality of classification models to obtain the target classification model. In one embodiment, the performing iteration based on the classification accuracy of each classification model and a plurality of sets of model parameters corresponding to a plurality of classification models to obtain the target classification model includes: Constructing a1 st model population based on a plurality of the classification models; determining a genetic classification model of a kth round from a plurality of classification models based on classification accuracy of each classification model in a kth model population, wherein k is a positive integer; Performing parameter crossing and parameter variation based on model parameters of the k-th round genetic classification model to obtain a k+1th model population; if k is equal to M, determining a target classification model with highest classification accuracy from the Mth model population, wherein M is a positive integer greater than 1; If k is smaller than M, adding 1 to k, and returning to execute classification accuracy based on each classification model in the kth model population, and determining a kth round of genetic classification model from a plurality of classification models. In one embodiment, the obtaining the sample fusion feature of the sample tobacco shred includes: Acquiring first detection data and second detection data of sample cut tobacco; performing feature sampling in data features corresponding to the first detection data and the second detection data to obtain a first feature population, wherein the first feature population comprises a plurality of feature subsets, and each feature subset comprises a plurality of data features; And carrying out feature iteration by a plurality of feature subsets in the first feature population according to the classification fitness of the feature subsets to obtain the sample fusion features of the sample tobacco