Search

CN-121982486-A - Cloud management system and method for cloth detection data analysis and processing

CN121982486ACN 121982486 ACN121982486 ACN 121982486ACN-121982486-A

Abstract

The invention provides a cloud management system for cloth detection data analysis and processing, which is characterized in that an edge AI computing device is used for acquiring structured data and well-resolved voice instructions of a field cloth image, a data acquisition and transmission component is used for transmitting the acquired structured data and well-resolved voice instructions of the cloth image to a container cloud platform, the container cloud platform is used for calling an AI GC algorithm model interface and returning instruction demand results to the edge AI computing device through the data acquisition and transmission component, the container component comprises an algorithm model container component and a data processing container component, the data processing container component is used for cleaning, processing, monitoring and analyzing cloth data and voice text interaction data, and the algorithm model container component is used for storing the cloth picture data and the structured data in the management container component to train and learn model to generate algorithm model files and transmitting the algorithm model files to the edge AI computing device. The method effectively solves the problems of difficult updating, low effectiveness and low data feedback speed of the traditional intelligent cloth inspection machine model, reduces the complexity of environment deployment by using a containerized cloud management system, increases the convenience of system maintenance, improves the data analysis and processing of users, increases the AI GC dialogue interactive processing function, and improves the feedback efficiency of quality problems.

Inventors

  • XU PENG
  • XU YAFEI
  • YUE YURONG

Assignees

  • 南通莱斯特纺织有限公司

Dates

Publication Date
20260505
Application Date
20251216

Claims (10)

  1. 1. The cloud management system for cloth detection data analysis and processing is characterized by being constructed by a container cloud technology through a container cloud platform, and comprises edge AI computing equipment (1), a data acquisition and transmission assembly (2) and the container cloud platform (5); the edge AI computing equipment (1) is used for acquiring structured data of the cloth image on site and the parsed voice command, and periodically receiving and updating an algorithm model file transmitted by the container cloud platform; The data acquisition and transmission assembly (2) is used for transmitting the structured data of the acquired cloth image and the parsed voice command to the container cloud platform (5); The container cloud platform (5) is used for calling AIGC an algorithm model interface and returning an instruction demand result to the edge AI computing equipment through a data acquisition and transmission component; The container cloud platform is also used for deploying, dispatching and managing containerized application programs and services and controlling the operation of the mirror warehouse and the container assembly; the mirror warehouse (51) is used for storing and managing mirror files in the container assembly to organize, share and distribute container mirrors so as to facilitate the deployment of the same container environment on multiple hosts; the container assembly (52) is a dock assembly for implementing environmental isolation and quick deployment functions, packaging, distributing and running applications; The container assembly (52) includes an algorithm model container assembly (521), a data processing container assembly (522); The data processing container assembly (522) is used for cleaning, processing, monitoring and analyzing cloth data and voice text interaction data; The algorithm model container component (521) is used for storing and managing cloth picture data and structured data in the container component to train a learning model to generate an algorithm model file, and transmitting the algorithm model file to the edge AI computing equipment.
  2. 2. A cloud management system for cloth inspection data analysis and processing according to claim 1, wherein said algorithmic model container assembly (521) comprises a cloth data preprocessing algorithm module (5211), a breadth analysis algorithm module (5212), a warp and weft density analysis algorithm module (5213), a cloth flaw detection and classification algorithm module (5214), a cloth quality evaluation algorithm module (5215) and a AIGC big model algorithm module (5216); The cloth data preprocessing algorithm module (5211) is used for analyzing cloth structured data acquired from the edge AI computing equipment, decoding and restoring a picture format into an original picture, and storing the original picture into a database of the data storage and management container component (523); The width analysis algorithm module (5212) is used for calculating the width of the cloth, calling cloth sample pictures collected by 8 cameras stored in the container cloud platform for width analysis, and calculating the width of a target area; the warp density and weft density analysis algorithm module (5213) is used for calling camera pictures of 4 positions of a non-corner area of a scene stored to the cloud, and intercepting a cloth target area at a central point to calculate the warp density and weft density; The cloth flaw detection and classification algorithm module (5214) is used for training YoloV and ShuffleNet models to detect and classify cloth flaws; The cloth quality evaluation algorithm module (5215) is used for extracting cloth structured data to train a cloth quality decision tree evaluation model to classify three quality grades of high, medium and low of cloth, and then score performance of textile personnel; The AIGC big model algorithm module (5216) is used for fine tuning ChatGLM the language model framework with live voice instructions and expected answer demand instructions, realizing that expected instruction words are input and output according to given instruction words and returned to the edge AI computing equipment.
  3. 3. A cloud management system for cloth inspection data analysis and processing as claimed in claim 2, wherein said data collection and transmission component employs a custom restful API to convert pictures into base64 format and picture related structured data into json string format as input to restful AP I.
  4. 4. A cloud management system for cloth detection data analysis and processing as claimed in claim 3, wherein the cloth quality decision tree evaluation model adopts a XGBoost classification model, discrete and standardized data are used as input of a XGBoost classification model, manually calibrated quality evaluation categories are used as target output of the model, selected features with higher importance coefficients are used as input of a XGBoost model, a cloth quality evaluation model is trained, and after an accuracy threshold set by classification of a test set is met, the trained XGBoost model is saved for cloth quality evaluation; And the high, medium and low of the cloth quality evaluation correspond to the performance coefficients of the textile personnel respectively, and the performance coefficients of the personnel in the performance period are determined according to the cumulative average value of the performance coefficients.
  5. 5. A cloud management system for cloth inspection data analysis and processing as claimed in claim 4, wherein the step of calculating the size of the web in the web analysis algorithm module comprises: Step1, acquiring an image, in particular an off-line cloth image which is acquired by calling 8 cameras of a container cloud platform at the same time; Step2, image stitching and repeated area removing, which specifically comprises the steps of extracting features of images acquired by 8 cameras, performing feature matching according to the trend of cloth lines, finding out corresponding feature point pairs, matching according to the feature point pairs, aligning the images, removing the repeated area according to the matching result, enabling the feature point pairs to overlap the corresponding positions, and stitching the aligned images to form a complete cloth image; step3, performing edge detection on the complete cloth image formed by splicing, specifically detecting the edge of the cloth image by using a Canny edge detection algorithm; step4, carrying out edge identification and measurement according to the detected edges of the cloth image, wherein the method specifically comprises the following steps: identifying and analyzing the detected cloth edges to find the upper edge and the lower edge of the cloth; Converting the pixel difference value of the upper edge and the lower edge with the proportional size of the corresponding known object to obtain the final width of the door; And comparing the calculated final width of the door with the width of the door which is actually and manually measured, and verifying the effectiveness of the algorithm.
  6. 6. The cloud management system for cloth inspection data analysis and processing according to claim 5, wherein said warp and weft density calculation step in said warp and weft density analysis algorithm module (5213) comprises: Step1, acquiring cloth images acquired by cameras at 4 positions of a non-corner area of a site stored by a container cloud platform, acquiring the center point position of each image, and cutting square RO I areas with unit length areas of the same length by taking the position as a midpoint, wherein the unit length area is an imaging pixel area obtained by scaling an actual 1 square inch; step2, preprocessing the RO I area obtained by cutting, namely performing image filtering and histogram equalization on the RO I area; Step3, calculating the weft density and the weft density, namely detecting transverse lines and vertical lines in the cloth image by using a straight line detection algorithm, namely counting the number of the weft lines and the warp lines of 4 ROI areas respectively, and taking the average value as a final calculated value of the warp density and the weft density.
  7. 7. A cloud management system for cloth inspection data analysis and processing as claimed in claim 6, wherein said container assembly further comprises a data storage and management container assembly and a log and monitoring container assembly; The data storage and management container component is used for storing and managing collected and generated data; the log and monitoring container assembly comprises a monitoring tool and a log recording tool, and is used for monitoring the running state, performance indexes and recording log information of the system; the cloud management system further comprises a front-end display interaction component and a user terminal; The user terminal is connected with the data acquisition and transmission assembly and the container cloud platform through the front-end display interaction assembly to interact, and is used for displaying and controlling the operation of the whole system, checking the data of the container cloud platform and monitoring logs, and training and periodically updating the algorithm model.
  8. 8. A cloud management method for cloth detection data analysis and processing is characterized in that the cloud management system for cloth detection data analysis and processing is adopted, and the management method comprises the following steps: S100, acquiring structural data of a cloth image on site by using an edge AI computing device, and analyzing a voice command into voice text data; s200, uploading the obtained structured data of the field cloth image and the voice text data to a container cloud platform by utilizing a data acquisition and transmission assembly; S300, the container cloud platform receives and analyzes the structured data and the voice text data through a cloth data preprocessing algorithm module in the data processing container assembly, and stores the data into the data storage and management container assembly; S400, extracting structured data of cloth pictures in the data storage and management container assembly, training a learning model to generate an algorithm model file, and transmitting the algorithm model file to the edge AI computing equipment.
  9. 9. The cloud management method for cloth inspection data analysis and processing as claimed in claim 8, wherein the step S400 specifically comprises: Analyzing the obtained cloth structured data, decoding and restoring the picture format into an original picture, and storing the original picture into a database of a data storage and management container component (523); calculating the width of cloth, and calculating the width of a target area by calling cloth sample pictures collected by 8 cameras stored in a container cloud platform to perform width analysis; Calling camera pictures of 4 positions of a non-corner area of a scene stored to a cloud end, and intercepting a cloth target area at a central point to calculate warp density and weft density; Training YoloV and ShuffleNet models to detect and classify cloth flaws; the cloth quality decision tree evaluation model is trained by extracting cloth structured data to classify the three quality grades of high, medium and low of the cloth, and then the performance of textile personnel is scored.
  10. 10. A cloud management system for cloth inspection data analysis and processing as claimed in claim 9 further comprising a framework for fine tuning ChatGLM the language model for live voice commands and desired answer demand commands to output desired command words based on given command words input and return to the edge AI computing device.

Description

Cloud management system and method for cloth detection data analysis and processing Technical Field The invention relates to the technical field of cloth detection, in particular to a cloud management system and method for cloth detection data analysis and processing. Background The intelligent cloth inspection machine adopts a fixed model for identification, and is difficult to update and upgrade in time. If a new cloth flaw type or an improved algorithm model occurs, the model needs to be manually updated. The intelligent cloth inspection machine has the advantages that the process is relatively complex, professional staff is required to operate the intelligent cloth inspection machine, production efficiency can be influenced for a long time, the use flow of the intelligent cloth inspection machine is relatively complex, operators are required to have higher expertise and skills, and certain time and resource investment are required for cultivating such professionals. Meanwhile, the service data feedback speed is low, the feedback of the cloth processing result of the intelligent cloth inspection machine to the actual production link is slow, which means that a long time is required to obtain the cloth quality feedback result, and corresponding measures are taken, so that the quality problem is delayed from the production link, and the quality and the production efficiency of the final product are affected. In addition, the current intelligent cloth inspection machine has no voice interaction function, and a worker needs to communicate with the cloth inspection machine in other ways, such as input and operation through a keyboard, a mouse or a touch screen. This approach may be inadequate in view and convenience, increasing the learning costs of operators familiarity with workflow and parameter settings, and reducing work efficiency. Disclosure of Invention The invention aims to provide a cloud management system and a cloud management method for cloth detection data analysis and processing, which are used for solving the problems in the background technology. In order to achieve the above object, one of the purposes of the present invention is that a cloud management system for cloth detection data analysis and processing is constructed by a container cloud technology through a container cloud platform, and the system comprises an edge AI computing device, a data acquisition and transmission assembly and the container cloud platform; the edge AI computing equipment is used for acquiring structured data of the cloth image on site and analyzing the voice command; the data acquisition and transmission component is used for transmitting the structural data of the acquired cloth image and the parsed voice command to the container cloud platform; the container cloud platform is used for calling AIGC an algorithm model interface and returning an instruction demand result to the edge AI computing equipment through the data acquisition and transmission component; The container cloud platform is also used for deploying, scheduling and managing containerized application programs and services and controlling the operation of the mirror warehouse and the container assembly; The mirror warehouse is used for storing and managing mirror files in the container assembly to organize, share and distribute container mirrors so as to facilitate the deployment of the same container environment on multiple hosts; The container component is a docker component and is used for realizing the functions of environment isolation and rapid deployment, packaging, distributing and running application programs; the container assembly comprises an algorithm model container assembly and a data processing container assembly; the data processing container component is used for cleaning, processing, monitoring and analyzing cloth data and voice text interaction data; the algorithm model container component is used for training the learning model to generate an algorithm model file by storing and managing cloth picture data and structured data in the container component, and transmitting the algorithm model file to the edge AI computing equipment. As a preferred mode which can be realized, the algorithm model container component comprises a cloth data preprocessing algorithm module, a breadth analysis algorithm module, a warp density and weft density analysis algorithm module, a cloth flaw detection and classification algorithm module, a cloth quality evaluation algorithm module and a AIGC big model algorithm module; The cloth data preprocessing algorithm module is used for analyzing cloth structured data acquired from the edge AI computing equipment, decoding and restoring a picture format into an original picture, and storing the original picture into a database of the data storage and management container assembly; the width analysis algorithm module is used for calculating the width of the cloth, calling cloth sample pictures collected by 8