An interactive control method and system for an industrial vision system
By constructing a structured knowledge base and using intent parsing technology, intelligent interaction of industrial vision systems has been achieved, solving the problems of cumbersome traditional interaction methods and insufficient automation of data processing, thereby improving operational efficiency and decision-making speed.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN WEINER HUISHI TECHNOLOGY CO LTD
- Filing Date
- 2026-01-20
- Publication Date
- 2026-06-09
AI Technical Summary
Traditional industrial vision systems suffer from cumbersome interaction methods, high learning costs, low information retrieval efficiency, insufficient data processing automation, high system control thresholds, and a lack of convenient access points, resulting in low operational efficiency and delayed decision-making by management.
By building a structured knowledge base and matching core functional modules through intent parsing, standardized instructions can be issued, and execution status data can be acquired and visualized in real time, thereby reducing the interaction threshold and improving operational efficiency and automation.
It enables intelligent interaction in industrial vision systems, lowers the barrier to system interaction, improves information operation efficiency, automates data analysis and processing, and avoids information lag in management decision-making.
Smart Images
Figure CN122174821A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data analysis technology, and in particular to an interactive control method and system for an industrial vision system. Background Technology
[0002] In the field of intelligent manufacturing, industrial vision systems are core tools for quality inspection and production monitoring, but their traditional interaction and functional modes have significant bottlenecks. 1. Cumbersome interaction methods and high learning costs: Existing industrial vision systems rely on mouse and menu interfaces, requiring employees to memorize complex function paths, resulting in long training cycles for new employees and low operational efficiency. 2. Low information retrieval efficiency, relying on manual searching: Manuals, troubleshooting guides, and inspection standards for industrial vision systems are mostly stored in PDF / Word format. Employees must manually flip through dozens of pages to find relevant information, failing to meet real-time operational needs. 3. Insufficient automation in data processing, leading to delayed decision-making: Defect analysis and production reports require manual export of raw data from the system, followed by filtering, calculation, and visualization using Excel. This time-consuming process results in delayed information for management decisions. 4. High system control threshold and lack of convenient entry points: Core controls of industrial vision systems require skilled technicians, making it difficult for frontline employees to quickly intervene; furthermore, control commands must be input through a fixed interface, which is unsuitable for real-world production scenarios. Summary of the Invention
[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an interactive control method and system for an industrial vision system, which realizes the automatic execution of the entire process of data acquisition, analysis and visualization data generation, and realizes the intelligent interaction of the industrial vision system.
[0004] To address the aforementioned technical problems, this invention provides an interactive control method for an industrial vision system, the method comprising: Receive user request information and preprocess the user request information to obtain preprocessed user request information; A structured knowledge base is constructed, and intent parsing is performed on the preprocessed user request information based on the structured knowledge base to obtain intent parsing information; Based on the intent parsing information, the corresponding core functional module is matched, the instruction information is determined based on the core functional module, and the instruction information is standardized based on the parameter name mapping to obtain standardized instruction information. Based on the system interface layer, the standardized instruction information is sent to the industrial vision system, and the industrial vision system executes the standardized instruction information. During the execution of the standardized instruction information by the industrial vision system, the execution status data and execution effect data of the industrial vision system are acquired in real time, and the execution status data and execution effect data are visualized to obtain visualized data. The visualized data and intent parsing information are then stored in the data storage layer.
[0005] Optionally, the step of preprocessing the user request information to obtain preprocessed user request information includes: The user request information is subjected to noise reduction processing to obtain noise-reduced user request information; The noise-reduced user request information is cleaned to obtain pre-processed user request information.
[0006] Optionally, the step of constructing a structured knowledge base and performing intent parsing on the preprocessed user request information based on the structured knowledge base to obtain intent parsing information includes: Acquire target document information from an industrial vision system, and preprocess the target document information to obtain preprocessed target document information; Structured modeling is performed based on the preprocessed target document information to obtain structured entry information, and a structured knowledge base is constructed based on the structured entry information. The preprocessed user request information is matched with the structured entry information of the structured knowledge base to obtain the matching degree analysis results, and the intent parsing information is determined based on the matching degree analysis results.
[0007] Optionally, the step of performing structured modeling based on the preprocessed target document information to obtain structured entry information includes: Optical character recognition is performed based on the preprocessed target document information to obtain optical character recognition results, and error correction processing is performed on the optical character recognition results to obtain error-corrected optical character recognition results. Natural language processing is performed on the preprocessed target document information based on an industrial terminology dictionary to obtain natural language processing results. Structured modeling is performed based on the results of natural language processing and optical character recognition after error correction to obtain structured entry information.
[0008] Optionally, the step of matching the corresponding core functional module based on the intent parsing information and determining the instruction information based on the core functional module includes: Based on the intent parsing information, the corresponding core functional module is matched, and based on the core functional module, the intent parsing information is used to associate documents to obtain associated document information, and the instruction information is determined based on the associated document information.
[0009] Optionally, the standardization process of the instruction information based on parameter name mapping to obtain standardized instruction information includes: The instruction information is subjected to element extraction to obtain target element information; Based on a preset mapping table, the target element information is mapped by parameter name to obtain the target parameter name; The target parameter name is standardized by performing parameter value standardization processing to obtain the standardized target parameter name; Standardized instruction information is generated based on the standardized target parameter name and instruction information.
[0010] Optionally, the step of visualizing the execution status data and execution effect data to obtain visualized data includes: The execution status data and execution effect data are cleaned to obtain cleaned execution status data and execution effect data. Execution indicator information is determined based on the execution status data and execution effect data after data cleaning and processing. The execution indicator information, execution status data after data cleaning and processing, and execution effect data are converted into formats to obtain the converted execution indicator information, execution status data, and execution effect data. Determine the chart type and rendering tool, and based on the chart type and rendering tool, visualize the execution indicator information, execution status data and execution effect data after format conversion to obtain visualized data.
[0011] Optionally, the step of performing data cleaning processing on the execution status data and execution effect data to obtain cleaned execution status data and execution effect data includes: The execution status data and execution effect data are processed to unify their formats, resulting in unified execution status data and execution effect data. Perform outlier handling on the standardized execution status data and execution effect data to obtain outlier-handled execution status data and execution effect data. The execution status data and execution effect data after outlier processing are subjected to consistency verification to obtain consistency verification results. Based on the consistency verification results, the execution status data and execution effect data after outlier processing are corrected to obtain corrected execution status data and execution effect data. The modified execution status data and execution effect data are verified to obtain verification information. Based on the verification information, the modified execution status data and execution effect data are cleaned to obtain cleaned execution status data and execution effect data.
[0012] Optionally, the step of determining execution indicator information based on the execution status data and execution effect data after data cleaning includes: Based on the execution status data and execution effect data after data cleaning and processing, the basic values are determined using core calculation indicators; The summary value is determined based on the execution status data and execution effect data after data cleaning and processing; The rate of change is determined based on the execution status data and execution effect data after data cleaning and processing, and the execution indicator information is determined based on the base value, the summary value and the rate of change.
[0013] In addition, the present invention also provides an interactive control system for an industrial vision system, the interactive control system comprising: an interactive entry layer, a large model processing layer, an agent scheduling layer, a core function layer, a system interface layer, and a data storage layer, the interactive control system being configured to execute the above-described interactive control method for an industrial vision system.
[0014] In this embodiment of the invention, users can perform relevant operations simply by inputting request information, replacing traditional mouse menu operations and lowering the system's interaction threshold. Constructing a structured knowledge base and using it to parse the preprocessed user request information improves the accuracy of intent parsing. Matching the corresponding core functional modules based on the intent parsing information, determining the instruction information based on the core functional modules, and standardizing the instruction information based on parameter name mapping; the standardized instruction information is then sent to the industrial vision system via the system interface layer. The industrial vision system executes the standardized instruction information, improving information operation efficiency and automating data analysis and processing. During the execution of standardized instruction information by the industrial vision system, real-time acquisition of execution status and effect data is performed. This execution status and effect data are then visualized, and the visualized data, along with the intent parsing information, is stored in the data storage layer. This allows relevant personnel to intuitively and clearly understand the data, avoiding information lag for management decisions and realizing intelligent interaction within the industrial vision system. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 This is a flowchart illustrating the interactive control method of the industrial vision system in an embodiment of the present invention. Figure 2This is a flowchart illustrating the interactive control method of an industrial vision system according to another embodiment of the present invention. Figure 3 This is a schematic diagram of the structural composition of the interactive control system of the industrial vision system in an embodiment of the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating the interactive control method of an industrial vision system according to an embodiment of the present invention. The method includes: S11: Receive user request information and preprocess the user request information to obtain preprocessed user request information; In the specific implementation of this invention, user request information is received, and the user request information is denoised to obtain denoised user request information; the denoised user request information is then cleaned to obtain preprocessed user request information, which can improve the reliability of subsequent intent parsing.
[0019] S12: Construct a structured knowledge base, and perform intent parsing on the preprocessed user request information based on the structured knowledge base to obtain intent parsing information; In the specific implementation of this invention, target document information of an industrial vision system is acquired, and the target document information is preprocessed to obtain preprocessed target document information; structured modeling is performed based on the preprocessed target document information to obtain structured entry information, and a structured knowledge base is constructed based on the structured entry information; the preprocessed user request information is matched with the structured entry information of the structured knowledge base to obtain the matching degree analysis result, and the intent parsing information is determined based on the matching degree analysis result, so that the system can better understand the user intent and achieve accurate operation control.
[0020] S13: Match the corresponding core functional module based on the intent parsing information, determine the instruction information based on the core functional module, and standardize the instruction information based on the parameter name mapping to obtain standardized instruction information; In the specific implementation of this invention, the core functional modules are matched based on the intent parsing information. The core functional modules then use the intent parsing information to associate documents, obtaining associated document information. Instruction information is determined based on the associated document information, and element extraction is performed on the instruction information to obtain target element information. Parameter names are mapped to the target element information using a preset mapping table to obtain target parameter names. Parameter value standardization processing is performed on the target parameter names to obtain standardized target parameter names. Standardized instruction information is generated based on the standardized target parameter names and instruction information, thereby improving the accuracy of system interaction.
[0021] S14: Based on the system interface layer, the standardized instruction information is sent to the industrial vision system, and the industrial vision system executes the standardized instruction information; In the specific implementation of this invention, the standardized instruction information is sent to the industrial vision system based on the system interface layer. The industrial vision system executes the standardized instruction information, replacing the traditional mouse menu operation, which can reduce the system interaction threshold and better adapt to the production site scenario.
[0022] S15: During the execution of the standardized instruction information by the industrial vision system, the execution status data and execution effect data of the industrial vision system are acquired in real time, and the execution status data and execution effect data are visualized to obtain visualized data. The visualized data and intent parsing information are then stored in the data storage layer.
[0023] In the specific implementation of this invention, during the execution of the standardized instruction information by the industrial vision system, the execution status data and execution effect data of the industrial vision system are acquired in real time. The execution status data and execution effect data are then cleaned to obtain cleaned execution status data and execution effect data. Execution indicator information is determined based on the cleaned execution status data and execution effect data. The execution indicator information, the cleaned execution status data, and the execution effect data are then converted to different formats to obtain format-converted execution indicator information, execution status data, and execution effect data. The chart type and rendering tool are determined, and the format-converted execution indicator information, execution status data, and execution effect data are visualized based on the chart type and rendering tool to obtain visualized data. The visualized data and intent parsing information are stored in the data storage layer, realizing the fully automated execution of the data acquisition-analysis-visualization data generation process. This intuitively displays relevant data, facilitating decision-making by management personnel.
[0024] In this embodiment of the invention, users can perform relevant operations simply by inputting request information, replacing traditional mouse menu operations and lowering the system's interaction threshold. Constructing a structured knowledge base and using it to parse the preprocessed user request information improves the accuracy of intent parsing. Matching the corresponding core functional modules based on the intent parsing information, determining the instruction information based on the core functional modules, and standardizing the instruction information based on parameter name mapping; the standardized instruction information is then sent to the industrial vision system via the system interface layer. The industrial vision system executes the standardized instruction information, improving information operation efficiency and automating data analysis and processing. During the execution of standardized instruction information by the industrial vision system, real-time acquisition of execution status and effect data is performed. This execution status and effect data are then visualized, and the visualized data, along with the intent parsing information, is stored in the data storage layer. This allows relevant personnel to intuitively and clearly understand the data, avoiding information lag for management decisions and realizing intelligent interaction within the industrial vision system.
[0025] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating an interactive control method for an industrial vision system according to another embodiment of the present invention, the method comprising: S201: Receive user request information and preprocess the user request information to obtain preprocessed user request information; In a specific implementation of the present invention, the step of preprocessing the user request information to obtain preprocessed user request information includes: performing noise reduction processing on the user request information to obtain noise-reduced user request information; and performing data cleaning processing on the noise-reduced user request information to obtain preprocessed user request information.
[0026] Specifically, the system receives user request information, which may include voice or text requests. The user request information is then subjected to noise reduction processing to obtain noise-reduced user request information. Noise reduction processing may employ an adaptive noise reduction algorithm based on spectral subtraction or a rule-based method.
[0027] The user request information after noise reduction undergoes data cleaning to obtain pre-processed user request information. If the user request information is voice information, it needs to be converted into text information before further data cleaning. The main steps of speech-to-text conversion are: 1) Format standardization: converting the denoised audio to a format supported by the automatic speech recognition system, unifying the sampling rate and channels; 2) Frame segmentation and windowing: segmenting audio at 20ms / frame and adding Hanning windows to reduce interference; 3) Feature extraction: using the MFCC / Fbank algorithm to convert the time-domain signal into a frequency-domain feature vector; 4) Model conversion: converting the acoustic model into features into phonemes, and using the industrial language model to spell text; 5) Result optimization: removing redundant interjections, correcting terminology, and outputting the text. Data cleaning removes abnormal characters and garbled text from the denoised user request information.
[0028] S202: Construct a structured knowledge base, and perform intent parsing on the preprocessed user request information based on the structured knowledge base to obtain intent parsing information; In the specific implementation of this invention, the construction of a structured knowledge base and the intention parsing of preprocessed user request information based on the structured knowledge base to obtain intention parsing information include: acquiring target document information of an industrial vision system and preprocessing the target document information to obtain preprocessed target document information; performing structured modeling based on the preprocessed target document information to obtain structured entry information, and constructing a structured knowledge base based on the structured entry information; performing a matching degree analysis between the preprocessed user request information and the structured entry information of the structured knowledge base to obtain a matching degree analysis result, and determining the intention parsing information based on the matching degree analysis result.
[0029] Specifically, the target document information of the industrial vision system is acquired, including documents such as operation manuals, defect standards, and fault code tables. This target document information is then preprocessed to obtain preprocessed target document information. Preprocessing includes deduplication, high-resolution enhancement, and classification labeling.
[0030] Structured modeling is performed based on the preprocessed target document information to obtain structured entry information. This structured entry information can be obtained by performing optical character recognition and natural language processing on the preprocessed target document information. A structured knowledge base is then constructed based on the structured entry information, and the structured entry information is stored in the data storage layer to complete the construction of the knowledge base.
[0031] The preprocessed user request information is matched with the structured entries in the structured knowledge base to obtain the matching degree analysis results. The preprocessed user request information and each structured entry in the structured knowledge base are then vectorized to obtain a first vector corresponding to the preprocessed user request information and a second vector corresponding to each structured entry in the structured knowledge base. The similarity between the first and second vectors is calculated, and this similarity is the matching degree analysis result. Based on the matching degree analysis result, intent parsing information is determined. Structured entries with similarity reaching a preset threshold are filtered out, and the structured entry with the highest similarity is selected as the intent parsing information.
[0032] Furthermore, the step of performing structured modeling based on the preprocessed target document information to obtain structured entry information includes: performing optical character recognition based on the preprocessed target document information to obtain optical character recognition results, and performing error correction processing on the optical character recognition results to obtain error-corrected optical character recognition results; performing natural language processing on the preprocessed target document information based on an industrial terminology dictionary to obtain natural language processing results; and performing structured modeling based on the natural language processing results and the error-corrected optical character recognition results to obtain structured entry information.
[0033] Specifically, optical character recognition (OCR) is performed on the preprocessed target document information. An industrial-grade OCR tool is used to extract the text from the preprocessed target document information; the extracted text is the OCR result. The OCR result is then subjected to error correction processing to obtain an error-corrected OCR result. The error correction can be reviewed and corrected by relevant professionals.
[0034] Natural language processing is performed on the preprocessed target document information based on the industrial terminology dictionary. The preprocessed target document information is then optimized by word segmentation using the industrial terminology dictionary. Core entities (such as fault code E302 and defect judgment conditions) and related relationships (such as E302 corresponding to camera communication interruption) are extracted. The resulting core entities and related relationships are the natural language processing results.
[0035] Structured modeling is performed based on the results of natural language processing and optical character recognition after error correction to obtain structured entry information. Fields are designed according to document type (e.g., "code, description, solution" for fault types) based on the results of natural language processing and optical character recognition after error correction, and information is filled in to generate standard entries, thus obtaining standardized entries.
[0036] S203: Match the corresponding core functional module based on the intent parsing information, determine the instruction information based on the core functional module, and standardize the instruction information based on the parameter name mapping to obtain standardized instruction information; In a specific implementation of the present invention, the step of matching the corresponding core functional module based on the intent parsing information and determining the instruction information based on the core functional module includes: matching the corresponding core functional module based on the intent parsing information, using the intent parsing information to perform document association based on the core functional module to obtain associated document information, and determining the instruction information based on the associated document information.
[0037] Specifically, based on the intent parsing information, the corresponding core functional modules are matched. These core functional modules include the Intelligent Inquiry Module, Intelligent Data Module, Intelligent Control Module, and Intelligent Assistance Module. The Intelligent Inquiry Module enables Artificial Intelligence (AI) interaction for document retrieval, troubleshooting, and standard queries. It can generate structured answers: first, it outputs a solution (e.g., "1. Check network cable connection; 2. Restart camera controller"), then associates relevant documents (e.g., click to view Chapter 5 of the "Camera Troubleshooting Manual"). The method for associating documents is as follows: 1. Preprocess documents: classify them by fault type, label the associated fault entities, chapters, and storage paths / URLs, and store them in the system in a structured manner; 2. Build a fault-document mapping table: bind faults (e.g., E302) with corresponding document information in the knowledge base; 3. Generate interactive links: when outputting solutions, attach hyperlinks pointing to the document paths (UNC paths for local use, shared URLs for cloud use); 4. Regularly verify link validity and user access permissions to ensure normal access.
[0038] The intelligent data module implements an intelligent workflow for defect analysis, report statistics, and decision support. The specific process is as follows: 2.1 Workflow Configuration: Provides visual templates (such as daily defect percentage analysis and monthly production line pass rate reports). Users can adjust parameters via text commands (e.g., reports grouped by production lines A / B / C, with a time range of the past 7 days). Configures data sources: Obtains raw data from the industrial vision system (defect type, inspection time, production line number) and production data from the production execution system from the system interface layer. 2.2 Automatic Data Processing: The intelligent agent automatically executes according to the workflow template: Data acquisition (synchronized once per hour) → Data cleaning (removing outliers, such as data with empty inspection times) → Analysis and calculation (e.g., scratch defect percentage = number of scratches / total number of defects) → Visualization generation (automatically generating line charts / bar charts). Supports follow-up analysis: Users can provide further commands (e.g., why did the pass rate of production line A decrease by 5% this week), and the system automatically correlates the defect type change data for that production line, outputting a cause analysis (e.g., crack defects increased by 30%, suspected mold wear).
[0039] The intelligent control module enables the integration of voice / text commands with the industrial vision system. The specific process is as follows: 3.1 Command Parsing and Verification: The module receives user commands (e.g., starting a chip appearance inspection task on production line B with an exposure time of 10ms). The large model parses the controlled object (production line B), control action (starting inspection), and parameters (exposure time 10ms). The selected large model can be CodeQwen1.5-7B. The parsing steps are: 1) Preprocessing: Loading an industrial vision domain dictionary into CodeQwen1.5-7B to adapt to the control command scenario; 2) Entity Recognition: Inputting user commands, the model extracts relevant entities of the controlled object through industrial customization capabilities; 3) Element Classification: Classifying the extracted entities into production line identifiers and task types to clarify the core dimensions of the controlled object; 4) Ambiguity Removal: If the command has ambiguous expressions, the model is calibrated using a preset production line list to confirm the unique controlled object; 5) Outputting Structured Results: Generating structured information of the controlled object, production line B, for subsequent parsing of control actions and parameters. 3.2 Control Command Conversion and Execution: Natural language commands are converted into standardized commands recognizable by the industrial vision system (e.g., converting an exposure time of 10ms into the system API requirement of exposure_time=10); commands are sent to the target system through the system interface layer, execution status is obtained in real time, and feedback is provided to the user via voice / text. 3.3 Control Log Recording: All control commands (command content, execution time, executor, execution result) are stored in the interactive log library of the data storage layer, supporting traceability and auditing (e.g., querying the detection task start record of production line B on March 15th).
[0040] The intelligent assistance module integrates human resources / finance / administration with industrial vision business. The specific process is as follows: 4.1 Employee Profiling and Function Adaptation: Obtain employee information (position, department, permissions) from the HR system and construct employee profiles (e.g., "Production line inspector - can query attendance / leave, cannot approve expense reports"). Employee profile construction methods include: 1) Data Extraction: Through interface integration with the HR system (e.g., Yonyou U8), obtain employee ID, name, position (e.g., production line inspector), department, permissions, etc., and ensure data consistency through a synchronous + real-time trigger mechanism; 2) Data Standardization: Unify position names (e.g., "inspection position changed to production line inspector"), integrate permission fields, and filter out resigned / test accounts; 3) Dimensional Modeling: Construct a profile framework and populate data based on basic identifiers (ID / name), position attributes (position / department), permission scope (can query attendance / cannot approve expense reports), and service requirements (high-frequency functions); 4) Rule Mapping: Use preset position-permission-... 5) Dynamic updates: When employees are transferred or their permissions change, the profile is updated synchronously to ensure that permissions and functions match. Based on the profile, the adapted functions are displayed on the unified desktop (such as the system client homepage) (such as displaying attendance inquiry and office supply application for inspectors, and displaying expense approval and department reports for management). 4.2 Automated processing of transactions: Human resources transactions: User instructions (such as querying my attendance record in March), the system connects to the human resources system to obtain data, outputs a text summary (22 days of attendance in March, 1 late arrival) and detailed attachments; Financial transactions: User instructions (such as the progress of my travel expense reimbursement), the system connects to the financial system, provides feedback on the status (approved, expected to arrive in 3 working days), and prompts that a hotel bill needs to be supplemented, asking whether to upload it now; Administrative transactions: User instructions (such as applying for 2 pens), the system automatically generates an application form, pushes it to the administrative approval process, provides feedback that the application form has been submitted, the approver is Manager Zhang, and it is expected to be completed in 1 working day. 4.3 Business-Service Linkage: Supports business-triggered services: For example, after a user completes a production line inspection task, the system automatically prompts them that they need to work 2 hours of overtime today and asks if they want to apply for overtime pay. Clicking the prompt will take them directly to the overtime application process without requiring manual system switching. Based on the core functional modules, the system uses the intent parsing information to associate documents and obtain related document information. Instruction information is determined based on the associated document information, which can be generated in conjunction with real-time user feedback.
[0041] Furthermore, the standardization process of the instruction information based on parameter name mapping to obtain standardized instruction information includes: extracting elements from the instruction information to obtain target element information; mapping the target element information to parameter names based on a preset mapping table to obtain target parameter names; standardizing the parameter values of the target parameter names to obtain standardized target parameter names; and generating standardized instruction information based on the standardized target parameter names and instruction information.
[0042] Specifically, the instruction information is processed to extract the target element information. The instruction information (such as exposure time 10ms) can be parsed using the CodeQwen1.5-7B tool to extract the parameter type (exposure time) and parameter value (10ms), which is the target element information.
[0043] Based on a preset mapping table, the target element information is mapped by parameter name to obtain the target parameter name. Then, a preset natural language parameter name-API parameter name mapping table (e.g., exposure time corresponds to exposure_time) is called to match the parameter name required by the system API.
[0044] The target parameter name is standardized to obtain a standardized target parameter name. The standardization process may involve removing units (e.g., ms, when the system API defaults to units) or converting units (e.g., 1s to 1000). The numerical compliance is verified (e.g., exposure time ≥ 1ms). Based on the standardized target parameter name and instruction information, standardized instruction information is generated and concatenated according to the system application programming interface syntax to generate standardized instruction information.
[0045] S204: Based on the system interface layer, the standardized instruction information is sent to the industrial vision system, and the industrial vision system executes the standardized instruction information; In the specific implementation of this invention, the standardized instruction information is sent to the industrial vision system based on the system interface layer. The system interface layer connects the industrial vision system and the enterprise service system to realize data communication and instruction issuance. The industrial vision system executes the standardized instruction information.
[0046] S205: During the execution of the standardized instruction information by the industrial vision system, the execution status data and execution effect data of the industrial vision system are acquired in real time, and the execution status data and execution effect data are cleaned to obtain the cleaned execution status data and execution effect data. In a specific implementation of this invention, the step of performing data cleaning on the execution status data and execution effect data to obtain cleaned execution status data and execution effect data includes: performing format unification processing on the execution status data and execution effect data to obtain format-unified execution status data and execution effect data; performing outlier processing on the format-unified execution status data and execution effect data to obtain outlier-processed execution status data and execution effect data; performing consistency verification on the outlier-processed execution status data and execution effect data to obtain consistency verification results, and performing correction processing on the outlier-processed execution status data and execution effect data based on the consistency verification results to obtain corrected execution status data and execution effect data; performing result verification on the corrected execution status data and execution effect data to obtain result verification information, and performing data cleaning on the corrected execution status data and execution effect data based on the result verification information to obtain cleaned execution status data and execution effect data.
[0047] Specifically, during the execution of the standardized instruction information by the industrial vision system, the execution status data and execution effect data of the industrial vision system are acquired in real time. The execution status data includes task progress data, production line data, etc., while the execution effect data is the result of instruction execution, such as visual defect detection results.
[0048] The execution status data and execution effect data are processed to unify the format, resulting in unified execution status data and execution effect data. This means unifying the field formats such as time (e.g., YYYY-MM-DD HH:MM) and numerical values.
[0049] Outlier handling is performed on the execution status data and execution effect data after the format is standardized to obtain outlier-handled execution status data and execution effect data. Missing key fields (detection time, production line number) are directly removed. Anomalies (such as negative defect count) are found using rule / statistical methods. Erroneous data is removed and special cases are marked.
[0050] The execution status data and execution effect data after outlier processing are subjected to consistency verification to obtain consistency verification results. Based on the consistency verification results, the execution status data and execution effect data after outlier processing are corrected to obtain corrected execution status data and execution effect data. The data logic is checked (e.g., number of defects ≤ total number of detections). If there is any inconsistency, the data from the production execution system shall prevail for correction.
[0051] The modified execution status data and execution effect data are verified to obtain result verification information. Result verification is carried out by sampling 10% of the data and can be verified by auditors. Based on the result verification information, the modified execution status data and execution effect data are cleaned to obtain cleaned execution status data and execution effect data.
[0052] S206: Determine execution indicator information based on the execution status data and execution effect data after data cleaning and processing; In the specific implementation of this invention, the determination of execution indicator information based on the execution status data and execution effect data after data cleaning includes: determining a basic value using core calculation indicators based on the execution status data and execution effect data after data cleaning; determining a summary value based on the execution status data and execution effect data after data cleaning; determining a rate of change based on the execution status data and execution effect data after data cleaning; and determining the execution indicator information based on the basic value, summary value, and rate of change.
[0053] Specifically, based on the cleaned and processed execution status and performance data, core calculation indicators are used to determine baseline values. These core indicators include the number of scratch defects, the total number of defects in production line A, and the statistical period. Baseline values include, for example, the daily scratch percentage = number of scratches on that day / total number of defects on that day. A summary value is then determined based on the cleaned and processed execution status and performance data. A summary value is given, for example, the weekly average percentage = sum of daily percentages / 7.
[0054] The rate of change is determined based on the execution status data and execution effect data after data cleaning and processing. The rate of change is such as (this week's percentage - last week's percentage) / last week's percentage × 100%. The execution indicator information is determined based on the base value, the summary value and the rate of change. That is, the execution indicator information consists of the base value, the summary value and the rate of change.
[0055] S207: Perform format conversion processing on the execution indicator information, execution status data after data cleaning and processing, and execution effect data to obtain the format-converted execution indicator information, execution status data, and execution effect data; In the specific implementation of this invention, the execution indicator information, execution status data after data cleaning and processing, and execution effect data are converted into formats to obtain the execution indicator information, execution status data, and execution effect data after format conversion, that is, the analyzed data is converted into a format supported by visualization tools (such as Matplotlib).
[0056] S208: Determine the chart type and rendering tool, and based on the chart type and rendering tool, perform visualization processing on the execution indicator information, execution status data and execution effect data after format conversion to obtain visualization data, and store the visualization data and intent parsing information in the data storage layer.
[0057] In the specific implementation of this invention, the chart type and rendering tool are determined. The chart type includes line charts for time series and bar charts for category comparison. The parameters of the rendering tool include axis labels, legends, colors, and titles. Based on the chart type and rendering tool, the execution indicator information, execution status data, and execution effect data after format conversion are visualized to obtain visualized data. The visualized data is embedded in the system desktop or as an attachment, and supports clicking to view specific values. The visualized data and intent parsing information are stored in the data storage layer. The data storage layer is used to store user interaction records, intelligent knowledge base, intelligent data workflow templates, system control parameters, etc.
[0058] In this embodiment of the invention, users can perform relevant operations simply by inputting request information, replacing traditional mouse menu operations and lowering the system's interaction threshold. Constructing a structured knowledge base and using it to parse the preprocessed user request information improves the accuracy of intent parsing. Matching the corresponding core functional modules based on the intent parsing information, determining the instruction information based on the core functional modules, and standardizing the instruction information based on parameter name mapping; the standardized instruction information is then sent to the industrial vision system via the system interface layer. The industrial vision system executes the standardized instruction information, improving information operation efficiency and automating data analysis and processing. During the execution of standardized instruction information by the industrial vision system, real-time acquisition of execution status and effect data is performed. This execution status and effect data are then visualized, and the visualized data, along with the intent parsing information, is stored in the data storage layer. This allows relevant personnel to intuitively and clearly understand the data, avoiding information lag for management decisions and realizing intelligent interaction within the industrial vision system.
[0059] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structural composition of the interactive control system of the industrial vision system in an embodiment of the present invention, as shown below. Figure 3 As shown, the interactive control system includes: an interactive entry layer 31, a large model processing layer 32, an agent scheduling layer 33, a core function layer 34, a system interface layer 35, and a data storage layer 36. The interactive control system is configured for the interactive control method of the above-mentioned industrial vision system.
[0060] In the specific implementation of this invention, the interaction entry layer 31 is used to receive user voice or text requests and perform preprocessing. The large model processing layer 32 adopts an industrial-customized large language model to parse user intent and professional terminology and generate an initial response. The intelligent agent scheduling layer 33 is used to schedule core functional modules according to intent and handle multiple concurrent requests. The core functional layer 34 includes an intelligent inquiry module, an intelligent data module, an intelligent control module, and an intelligent assistance module, which respectively realize information query, data processing, system control, and employee service. The system docking layer 35 is used to dock with the original industrial vision system and enterprise service system to realize data interoperability and command issuance. The data storage layer 36 is used to store knowledge base, interaction logs, and workflow configuration data. The intelligent inquiry module includes a knowledge base construction unit and an intent response unit: the knowledge base construction unit collects industrial vision documents and historical questions and answers to generate a structured knowledge base; the intent response unit matches the knowledge base, outputs structured answers, and associates them with documents. The intelligent data module includes a workflow configuration unit, an automatic processing unit, and a push unit: the workflow configuration unit provides visual templates and parameter adjustment functions; the automatic processing unit executes data collection, cleaning, analysis, and visualization; the push unit pushes reports to management at preset times. The intelligent control module includes an instruction parsing unit, an access verification unit, and an instruction conversion unit: the instruction parsing unit extracts control objects, actions, and parameters; the access verification unit queries user permissions; and the instruction conversion unit converts natural language instructions into standardized instructions for the industrial vision system. The intelligent assistance module includes an employee profiling unit, a transaction processing unit, and a linkage unit: the employee profiling unit is based on HR data adaptation; the transaction processing unit connects to HR / finance / administration systems to process transactions; and the linkage unit provides automatic prompts for business-triggered services.
[0061] In this embodiment of the invention, users can perform relevant operations simply by inputting request information, replacing traditional mouse menu operations and lowering the system's interaction threshold. Constructing a structured knowledge base and using it to parse the preprocessed user request information improves the accuracy of intent parsing. Matching the corresponding core functional modules based on the intent parsing information, determining the instruction information based on the core functional modules, and standardizing the instruction information based on parameter name mapping; the standardized instruction information is then sent to the industrial vision system via the system interface layer. The industrial vision system executes the standardized instruction information, improving information operation efficiency and automating data analysis and processing. During the execution of standardized instruction information by the industrial vision system, real-time acquisition of execution status and effect data is performed. This execution status and effect data are then visualized, and the visualized data, along with the intent parsing information, is stored in the data storage layer. This allows relevant personnel to intuitively and clearly understand the data, avoiding information lag for management decisions and realizing intelligent interaction within the industrial vision system.
[0062] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.
[0063] Furthermore, the interactive control method and system of an industrial vision system provided by the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. An interactive control method for an industrial vision system, characterized in that, The method includes: Receive user request information and preprocess the user request information to obtain preprocessed user request information; A structured knowledge base is constructed, and intent parsing is performed on the preprocessed user request information based on the structured knowledge base to obtain intent parsing information; Based on the intent parsing information, the corresponding core functional module is matched, the instruction information is determined based on the core functional module, and the instruction information is standardized based on the parameter name mapping to obtain standardized instruction information. Based on the system interface layer, the standardized instruction information is sent to the industrial vision system, and the industrial vision system executes the standardized instruction information. During the execution of the standardized instruction information by the industrial vision system, the execution status data and execution effect data of the industrial vision system are acquired in real time, and the execution status data and execution effect data are visualized to obtain visualized data. The visualized data and intent parsing information are then stored in the data storage layer.
2. The interactive control method for an industrial vision system according to claim 1, characterized in that, The step of preprocessing the user request information to obtain preprocessed user request information includes: The user request information is subjected to noise reduction processing to obtain noise-reduced user request information; The noise-reduced user request information is cleaned to obtain pre-processed user request information.
3. The interactive control method for the industrial vision system according to claim 1, characterized in that, The construction of a structured knowledge base, and the subsequent intent parsing of preprocessed user request information based on the structured knowledge base to obtain intent parsing information, includes: Acquire target document information from an industrial vision system, and preprocess the target document information to obtain preprocessed target document information; Structured modeling is performed based on the preprocessed target document information to obtain structured entry information, and a structured knowledge base is constructed based on the structured entry information. The preprocessed user request information is matched with the structured entry information of the structured knowledge base to obtain the matching degree analysis results, and the intent parsing information is determined based on the matching degree analysis results.
4. The interactive control method for an industrial vision system according to claim 3, characterized in that, The structured modeling based on the preprocessed target document information to obtain structured entry information includes: Optical character recognition is performed based on the preprocessed target document information to obtain optical character recognition results, and error correction processing is performed on the optical character recognition results to obtain error-corrected optical character recognition results. Natural language processing is performed on the preprocessed target document information based on an industrial terminology dictionary to obtain natural language processing results. Structured modeling is performed based on the results of natural language processing and optical character recognition after error correction to obtain structured entry information.
5. The interactive control method for an industrial vision system according to claim 1, characterized in that, The process of matching the corresponding core functional module based on the intent parsing information and determining the instruction information based on the core functional module includes: Based on the intent parsing information, the corresponding core functional module is matched, and based on the core functional module, the intent parsing information is used to associate documents to obtain associated document information, and the instruction information is determined based on the associated document information.
6. The interactive control method for an industrial vision system according to claim 1, characterized in that, The standardization process based on parameter name mapping to obtain standardized instruction information includes: The instruction information is subjected to element extraction to obtain target element information; Based on a preset mapping table, the target element information is mapped by parameter name to obtain the target parameter name; The target parameter name is standardized to obtain a standardized target parameter name; Standardized instruction information is generated based on the standardized target parameter name and instruction information.
7. The interactive control method for an industrial vision system according to claim 1, characterized in that, The step of visualizing the execution status data and execution effect data to obtain visualized data includes: The execution status data and execution effect data are cleaned to obtain cleaned execution status data and execution effect data. Execution indicator information is determined based on the execution status data and execution effect data after data cleaning and processing. The execution indicator information, execution status data after data cleaning and processing, and execution effect data are converted into formats to obtain the converted execution indicator information, execution status data, and execution effect data. Determine the chart type and rendering tool, and based on the chart type and rendering tool, visualize the execution indicator information, execution status data and execution effect data after format conversion to obtain visualized data.
8. The interactive control method for an industrial vision system according to claim 7, characterized in that, The step of performing data cleaning processing on the execution status data and execution effect data to obtain cleaned execution status data and execution effect data includes: The execution status data and execution effect data are processed to unify their formats, resulting in unified execution status data and execution effect data. Perform outlier handling on the standardized execution status data and execution effect data to obtain outlier-handled execution status data and execution effect data. The execution status data and execution effect data after outlier processing are subjected to consistency verification to obtain consistency verification results. Based on the consistency verification results, the execution status data and execution effect data after outlier processing are corrected to obtain corrected execution status data and execution effect data. The modified execution status data and execution effect data are verified to obtain verification information. Based on the verification information, the modified execution status data and execution effect data are cleaned to obtain cleaned execution status data and execution effect data.
9. The interactive control method for an industrial vision system according to claim 7, characterized in that, The determination of execution indicator information based on the execution status data and execution effect data after data cleaning and processing includes: Based on the execution status data and execution effect data after data cleaning and processing, the basic values are determined using core calculation indicators; The summary value is determined based on the execution status data and execution effect data after data cleaning and processing; The rate of change is determined based on the execution status data and execution effect data after data cleaning and processing, and the execution indicator information is determined based on the base value, the summary value and the rate of change.
10. An interactive control system for an industrial vision system, characterized in that, The interactive control system includes: an interactive entry layer, a large model processing layer, an agent scheduling layer, a core function layer, a system interface layer, and a data storage layer. The interactive control system is configured to execute the interactive control method of the industrial vision system according to any one of claims 1-9.