Method, system and storage medium for visual expression of odor based on electronic nose technology

Abstract

The present invention discloses a method, system and storage medium for visually expressing smells based on electronic nose technology, the method comprising: obtaining category information of smells to be identified based on electronic nose technology; determining demand information based on the category information of smells to be identified; collecting corresponding relevant data to build a database based on the demand information; building a knowledge graph centered on identifying smells based on the database; and converting the structured knowledge graph into a visual node link graph. The method, system and storage medium for visualizing smells based on electronic nose technology of the present invention use visual content to display the identified smell-related information to users, and the visual content can help users have a more intuitive understanding of the smell.

Description

Technical Field

[0001] The present invention belongs to the technical field of odor visualization, and in particular relates to an odor visualization expression method, system and storage medium based on electronic nose technology. Background Art

[0002] The experience economy based on the five senses has become one of the current development trends. The sense of smell is playing an increasingly important role. Smell can subtly influence consumer behavior and enhance users' experience of products. In the existing technology, the development and popularization of electronic sensory technology can detect and identify complex gases, and the basic information of gas category components is more scientific and accurate. However, the category information display of traditional odor recognition can only be quickly understood by professionals and depends on professional knowledge and experience. It is difficult for ordinary people to understand the smell by name, which has poor versatility and large workload.

[0003] Therefore, it is necessary to design and process the identified odor information using computer technology, artificial intelligence technology, etc., to generate a comprehensive and easy-to-understand visual information report. In the existing technology, only the characteristics of the odor itself are obtained after identification, and the related information such as the relevant application and image description of the odor is not fully and clearly displayed. The existing recognition and acquisition of odors are only the areas involved by relevant researchers. It is difficult for other professionals to grasp the comprehensive information of the relevant odors in a short time, and the existing technology cannot be more intuitive and comprehensive. Summary of the invention

[0004] The present invention provides a method, system and storage medium for visually expressing smell based on electronic nose technology, which adopts the following technical solutions:

[0005] A method for visualizing smell based on electronic nose technology comprises the following steps:

[0006] Obtaining category information of the odor to be identified based on electronic nose technology;

[0007] Determine the required information according to the category information of the odor to be identified;

[0008] Build a database based on the relevant data collected according to the demand information;

[0009] Build a knowledge graph centered on odor recognition based on the database;

[0010] Convert structured knowledge graphs into visual node-link graphs.

[0011] Furthermore, category information refers to the category predicted by a trained model in pattern recognition or the best matching category in the training data set.

[0012] Furthermore, category information refers to the category to which the trained model predicts it in pattern recognition;

[0013] The specific method for obtaining the category information of the odor to be identified based on the electronic nose technology is:

[0014] Collect a number of different odor sample data to form a covariance matrix with n rows and m columns, where n is the dimension of the collected data and m is the number of odor samples;

[0015] Compute the eigenvalues ​​and eigenvectors of the covariance matrix;

[0016] Select the eigenvectors of the q dimensions with the largest eigenvalues ​​to form a projection matrix with q rows and m columns;

[0017] The classification model is trained using the data features obtained by multiplying each odor feature with the projection matrix;

[0018] The data of the odor to be identified is collected and input into the trained classification model to obtain the category information.

[0019] Furthermore, the classification model is a BP ANN model.

[0020] Furthermore, the specific method for training the classification model is:

[0021] The weights and thresholds between different layers of the BP ANN model are set to a random floating point number between -1.0 and 1.0;

[0022] Use the derivative of the Sigmoid function to calculate the error between the output value of the output layer node and the target value;

[0023] According to the Widrow and Hoff correction learning rules, the weights between nodes in each layer and the thresholds of nodes in the hidden layer and output layer are corrected until BP ANN reaches a fitting state.

[0024] Furthermore, the value of n is 8, and the value of q is 3.

[0025] Furthermore, the specific method of collecting corresponding relevant data according to the demand information to build a database is:

[0026] Crawl network data based on nutch topic information;

[0027] The crawled network data is processed to obtain the target information;

[0028] The calculated target information is saved to form a database.

[0029] Furthermore, the specific method of processing the crawled network data to obtain the target information is as follows:

[0030] The target information is obtained by calculating the crawled network data through the CTPN text detection algorithm, CRNN text recognition algorithm and SSD image detection algorithm.

[0031] An odor visualization expression system based on electronic nose technology, comprising:

[0032] A data classification unit, used for obtaining category information of odors to be identified based on electronic nose technology;

[0033] An information determination unit, used to determine required information according to the category information of the odor to be identified;

[0034] The data collection unit collects corresponding relevant data according to the demand information and builds a database;

[0035] The generation unit builds a knowledge graph centered on odor recognition based on the database;

[0036] The conversion unit converts the structured knowledge graph into a visual node-link graph.

[0037] A computer-readable storage medium stores computer instructions, which, when executed by a processor, implement any of the aforementioned methods for visualizing smells based on electronic nose technology.

[0038] The benefit of the present invention lies in the provided odor visualization expression method, system and storage medium based on electronic nose technology, which use visualized content to display the identified odor-related information to the user, and the visualized content can promote the user's more intuitive understanding of the odor.

[0039] The present invention is also beneficial in that it provides an odor visualization expression method, system and storage medium based on electronic nose technology, which obtains odor category information through electronic nose technology and collects demand information through computer crawler technology to construct a knowledge graph centered on odor identification.

[0040] The present invention is also beneficial in that the provided odor visualization expression method, system and storage medium based on electronic nose technology use natural language technology to analyze text data for odor identification, summarize and identify keywords in the odor text, and can form a database of odor categories through its own and related characteristics including but not limited to gas composition, odor category characteristics, odor descriptors, odor color, odor volatility, etc., for constructing a knowledge graph centered on odor identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a schematic diagram of a method for visually expressing odor based on electronic nose technology of the present invention;

[0042] Figure 2It is a schematic diagram of a node link graph in the odor visualization expression method based on electronic nose technology of the present invention. DETAILED DESCRIPTION

[0043] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] like Figure 1 The present application shows a method for visualizing the expression of odors based on electronic nose technology, which comprises the following steps: S1: obtaining the category information of the odor to be identified based on the electronic nose technology. S2: determining the demand information based on the category information of the odor to be identified. S3: collecting the corresponding relevant data based on the demand information to build a database. S4: constructing a knowledge graph centered on identifying odors based on the database. S5: converting the structured knowledge graph into a visual node link graph. The method for visualizing the expression of odors based on electronic nose technology uses visual content to display the identified odor-related information to the user, and the visual content can promote the user's more intuitive understanding of the odor. The above method is described in detail below.

[0045] For step S1: obtaining category information of the odor to be identified based on the electronic nose technology.

[0046] It can be understood that category information refers to the category predicted by the trained model in pattern recognition or the best suitable category matched in the training data set.

[0047] As a preferred implementation, the category information refers to the category predicted by the trained model in pattern recognition. The specific method for obtaining the category information of the odor to be identified based on the electronic nose technology is:

[0048] Several different odor sample data are collected to form a covariance matrix with n rows and m columns, where n is the dimension of the collected data and m is the number of odor samples.

[0049] Compute the eigenvalues ​​and eigenvectors of the covariance matrix.

[0050] The eigenvectors of the q dimensions with the largest eigenvalues ​​are selected to form a projection matrix with q rows and m columns.

[0051] The classification model is trained by multiplying each odor feature with the projection matrix to obtain data features. Here, the odor feature of each odor refers to the sample data collected for each odor, which is a data matrix with n rows and 1 column.

[0052] The data of the odor to be identified is collected and input into the trained classification model to obtain the category information.

[0053] Preferably, the value of n is 8 and the value of q is 3.

[0054] As a preferred implementation, the classification model is a BP ANN (back propagation neural network) model.

[0055] As a preferred implementation, the specific method for training the classification model is:

[0056] The weights and thresholds between different layers of the BP ANN model are set to a random floating point number between -1.0 and 1.0.

[0057] The derivative of the Sigmoid function is used to calculate the error between the output value of the output layer node and the target value.

[0058] According to the Widrow and Hoff correction learning rules, the weights between nodes in each layer and the thresholds of nodes in the hidden layer and output layer are corrected until BP ANN reaches a fitting state.

[0059] Specifically, a number of different odor sample data are collected through the electronic nose system. The electronic nose system includes a sensor array unit, a signal preprocessing unit, and a pattern recognition unit. Category information refers to the category to which the trained process in the pattern recognition unit predicts or the best suitable category matched in the training data set. The training process refers to the process that the electronic nose system needs to be trained before identification, that is, the process of classification using known odor categories and their characteristics. Taking the process of identifying fruit odor as an example, the training steps are as follows:

[0060] First, the four different fruits were selected to collect the odors to be analyzed through the gas collection chamber. The gas collection chamber is a closed space that can ensure the constant concentration of odors and prevent them from being easily dispersed.

[0061] The collected data of four different odor samples are formed into a matrix of 8 rows and 4 columns through the odor sensor array unit and the signal preprocessing unit, and the eigenvalues ​​and eigenvectors of the covariance matrix are calculated.

[0062] The eigenvectors of the three dimensions with the largest eigenvalues ​​are selected to form a transformation matrix, and a projection matrix with 3 rows and 4 columns is obtained.

[0063] The data features obtained by multiplying each odor feature by the projection matrix are then trained using BP ANN.

[0064] After the training is completed, when it is necessary to identify the smell of one of the fruits, the trained BP ANN model is used to make predictions and output the name of the fruit with the highest degree of fit.

[0065] For step S2: determine the required information according to the category information of the odor to be identified.

[0066] Specifically, the demand information of the smell is determined according to the identified smell category. Taking the identification of a certain fruit smell as an example, if it is identified as a strawberry smell, the demand information related to the strawberry smell is automatically generated (or artificially generated). The relevant demand information includes but is not limited to professional classification, gas composition, odor descriptors, odor color, odor volatility and other related characteristics. The above demand information can also include specific and lower-level keywords or phrases. For example, professional classifications can include: fragrant odor, resin odor, mint odor, sweet odor, chemical odor, popcorn paste odor, lemon odor, non-citrus fruit odor, irritating odor, corrupt odor, etc.

[0067] For step S3: collect corresponding relevant data according to the demand information to build a database.

[0068] The specific method of building a database based on the corresponding relevant data collected according to the demand information is:

[0069] Crawling network data based on nutch topic information. Specifically, the method for crawling network data based on nutch topic information includes the following steps: First, Nutch obtains the initial URL from the seed web page file Seeds.txt and injects it into crawldb. Then obtain the URL set to be crawled from crawldb, create a new segment, and generate Fetchlist. When the crawler crawls the web page, it stores the web page data in the segment and then parses the content, extracts the new URL and updates crawl1db. Finally, obtain the new URL set from the updated crawlc to start crawling network data.

[0070] The crawled network data is processed to obtain target information. Specifically, the crawled network data is calculated by the CTPN text detection algorithm, the CRNN text recognition algorithm and the SSD image detection algorithm to obtain the target information.

[0071] Finally, the calculated target information is saved to form a database.

[0072] For step S4: construct a knowledge graph centered on odor identification based on the database.

[0073] The method for constructing a knowledge graph includes the following steps: establishing a basic knowledge ontology of the odor category according to the category image, wherein the basic knowledge ontology of the odor category has multiple entities and connection relationships, wherein the connection relationships can establish different connections according to the attributes between the entities, and further filling the above target information into the basic knowledge ontology of product design. The knowledge graph construction also includes the extraction of entities, wherein the entity extraction also includes the extraction of entity concepts and entity relationships. The extracted entity concepts and entity relationships are filled into the basic knowledge ontology of the odor category to form a knowledge graph.

[0074] For step S5: convert the structured knowledge graph into a visual node-link graph.

[0075] The constructed odor category knowledge graph is structured data, which needs to be converted into a visual representation of a node-link graph. The node-link graph represents the ontology as interconnected nodes, and the edges between nodes are represented by lines. Figure 2 As shown. Among them, the focus + context / distortion technology is a method of expressing a node-link graph. The node on the focus is the central node (i.e., the odor category information), and the remaining nodes appear around it (i.e., the target information). The display size of the associated nodes decreases with the progression of the inheritance relationship until it is invisible, and finally forms a visualized odor category knowledge graph. In the process of using this technology to visualize the knowledge graph, it is necessary to focus on specific nodes for zooming in. This method makes room for the focus by distorting the geometric shape of the graph and compressing the context area, so that the node of interest to the user can be easily moved to the display center as the focus, so as to display more details while maintaining the context display of the focus-related node.

[0076] The present application also discloses a smell visualization expression system based on electronic nose technology. The smell visualization expression system based on electronic nose technology can execute the above-mentioned smell visualization expression method based on electronic nose technology. Specifically, the smell visualization expression system based on electronic nose technology includes: a data classification unit, an information determination unit, a data acquisition unit, a generation unit and a conversion unit.

[0077] The data classification unit is used to obtain the category information of the odor to be identified based on the electronic nose technology. The specific method for the data classification unit to obtain the category information of the odor to be identified based on the electronic nose technology is: collect a number of different odor sample data to form a covariance matrix of n rows and m columns, n is the dimension of the collected data, and m is the number of odor samples; calculate the eigenvalues ​​and eigenvectors of the covariance matrix; select the eigenvectors of q dimensions with the largest eigenvalues ​​to form a projection matrix of q rows and m columns; train the classification model with the data features obtained by multiplying each odor feature with the projection matrix; collect the data of the odor to be identified and input it into the trained classification model to obtain the category information. The information determination unit is used to determine the required information according to the category information of the odor to be identified. The data collection unit collects the corresponding relevant data according to the required information to build a database. The generation unit builds a knowledge graph centered on identifying odors according to the database. The conversion unit converts the structured knowledge graph into a visual node link graph. Among them, the specific execution content of each unit refers to the content disclosed by the aforementioned odor visualization expression method based on electronic nose technology, which will not be repeated here.

[0078] The present application embodiment provides a storage medium storing executable instructions, wherein executable instructions are stored, and when the executable instructions are executed by a processor, the processor will be caused to execute the odor visualization expression method based on the electronic nose technology provided by the present application embodiment. In some embodiments, the storage medium can be a memory such as a flash memory, a magnetic surface memory, an optical disk, or an optical disk memory; or it can be a variety of devices including one or any combination of memories. In some embodiments, the executable instructions can be in the form of a program, software, a software module, a script or a code, written in any form of programming language (including a compiled or interpreted language, or a declarative or procedural language), and it can be deployed in any form, including being deployed as an independent program or being deployed as a module, component, subroutine or other unit suitable for use in a computing environment. As an example, the executable instructions may but do not necessarily correspond to a file in a file system, and may be stored in a part of a file that stores other programs or data, for example, stored in one or more scripts in a Hypertext Markup Language (HTML) document, stored in a single file dedicated to the program in question, or stored in multiple collaborative files (for example, files storing one or more modules, subroutines or code portions). As an example, the executable instructions may be deployed to be executed on one onboard computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed at multiple sites and interconnected by a communication network.

[0079] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for visualizing smell based on electronic nose technology, characterized in that: The following steps are involved: Obtaining category information of the odor to be identified based on electronic nose technology; Determine the required information according to the category information of the odor to be identified; Collect relevant data corresponding to the demand information to build a database; Constructing a knowledge graph centered on odor recognition based on the database; Convert structured knowledge graphs into visual node-link graphs; The category information refers to the category predicted by the trained model in pattern recognition; The specific method for obtaining the category information of the odor to be identified based on the electronic nose technology is: Collect a number of different odor sample data to form a covariance matrix with n rows and m columns, where n is the dimension of the collected data and m is the number of odor samples; Calculating the eigenvalues ​​and eigenvectors of the covariance matrix; Select the eigenvectors of the q dimensions with the largest eigenvalues ​​to form a projection matrix with q rows and m columns; The classification model is trained using the data features obtained by multiplying each odor feature with the projection matrix; Collecting data of the odor to be identified and inputting it into a trained classification model to obtain the category information; The specific method of collecting corresponding relevant data according to the demand information to build a database is: Crawl network data based on nutch topic information; Process the crawled network data to obtain the target information; The target information obtained by calculation is saved to form the database; The specific method of processing the crawled network data to obtain the target information is: The target information is obtained by calculating the crawled network data through the CTPN text detection algorithm, the CRNN text recognition algorithm and the SSD image detection algorithm.

2. The method for visualizing smell based on electronic nose technology according to claim 1, characterized in that: The classification model is a BP ANN model.

3. The method for visualizing smell based on electronic nose technology according to claim 2, characterized in that: The specific method for training the classification model is: The weights and thresholds between different layers of the BP ANN model are set to a random floating point number between -1.0 and 1.0; Use the derivative of the Sigmoid function to calculate the error between the output value of the output layer node and the target value; The weights between nodes in each layer and the thresholds of nodes in the hidden layer and the output layer are corrected according to the Widrow and Hoff correction learning rules until the BP ANN reaches a fitting state.

4. The method for visualizing smell based on electronic nose technology according to claim 2, characterized in that: The value of n is 8 and the value of q is 3.

5. An odor visualization expression system based on electronic nose technology, characterized in that: Include: A data classification unit, used for obtaining category information of odors to be identified based on electronic nose technology; An information determination unit, used to determine required information according to the category information of the odor to be identified; A data collection unit collects corresponding relevant data according to the demand information to build a database; A generation unit, which constructs a knowledge graph centered on odor recognition based on the database; The conversion unit converts the structured knowledge graph into a visual node-link graph; The category information refers to the category predicted by the trained model in pattern recognition; The specific method for the data classification unit to obtain the category information of the odor to be identified based on the electronic nose technology is: Collect a number of different odor sample data to form a covariance matrix with n rows and m columns, where n is the dimension of the collected data and m is the number of odor samples; Calculating the eigenvalues ​​and eigenvectors of the covariance matrix; Select the eigenvectors of the q dimensions with the largest eigenvalues ​​to form a projection matrix with q rows and m columns; The classification model is trained using the data features obtained by multiplying each odor feature with the projection matrix; Collecting data of the odor to be identified and inputting it into a trained classification model to obtain the category information; The specific method for the data collection unit to collect corresponding relevant data according to the demand information to build a database is: Crawl network data based on nutch topic information; Process the crawled network data to obtain the target information; The target information obtained by calculation is saved to form the database; The specific method of processing the crawled network data to obtain the target information is: The target information is obtained by calculating the crawled network data through the CTPN text detection algorithm, the CRNN text recognition algorithm and the SSD image detection algorithm.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by the processor, the method for visualizing smell based on electronic nose technology as described in any one of claims 1 to 4 is implemented.

Images