Machine vision measurement method for trouser workshop station working hours

Abstract

The invention discloses a machine vision measurement method for trouser workshop station working hours, and relates to the technical field of workshop station working hour measurement, comprising the following steps: erecting a camera above a trouser production line to collect a worker operation video; carrying out segmentation processing on the collected process video by using industrial video analysis software; decomposing the segmented process video into pictures, labeling the pictures by using labeling software LabelImg, and outputting a data set; performing machine learning training on a Faster-RCNN-based target detection algorithm by using the data set to obtain a target detection model; the trained target detection model is adopted to recognize the working state of a worker, and the specific working hours of all procedures are calculated; a camera is erected above a trousers production line to collect operation videos, industrial video analysis, image segmentation, deep learning target detection model training and process recognition technologies are combined, automatic recognition and accurate calculation of operation time of each process of a worker are achieved, and a time sequence is introduced in the method to analyze and position a processing cycle.

Description

Technical Field

[0001] This invention relates to the field of workshop workstation time measurement technology, and in particular to a machine vision method for measuring workstation time in a trousers workshop. Background Technology

[0002] Workshop timekeeping technology refers to the measurement and recording of the time required to complete a specific process at each workstation through various means. Its purpose is to optimize production processes, improve production efficiency, ensure product quality, and rationally plan production schedules. Therefore, how to utilize advanced technologies to improve the intelligence and safety of workshop timekeeping has become one of the most pressing issues to be addressed.

[0003] In the field of workshop workstation time measurement, in traditional trouser production workshops, workstation time and work quality mainly rely on manual observation and recording. This method not only consumes a lot of time and manpower, but also makes it difficult to guarantee the accuracy and consistency of data due to human factors. Furthermore, trouser production involves many and complex processes, and the production processes and procedures for different styles of trousers may vary. Traditional methods are difficult to adapt to these changes quickly, and there are also significant limitations in quality inspection of each process, making it impossible to fully cover all key processes. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a machine vision method for measuring workstation hours in trouser production workshops. This addresses the problem that in traditional trouser production workshops, workstation hours and work quality mainly rely on manual observation and recording. This method not only consumes a lot of time and manpower, but also makes it difficult to guarantee the accuracy and consistency of data due to human factors. Furthermore, trouser production involves numerous and complex processes, and the production processes and procedures for different styles of trousers may vary. Traditional methods are difficult to adapt to these changes quickly, and there are also significant limitations in quality inspection for each process, failing to comprehensively cover all key processes.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a machine vision method for measuring work hours at workstations in a trouser manufacturing workshop, comprising:

[0008] Cameras were installed above the trouser production line to collect videos of workers at their work.

[0009] Industrial video analytics software was used to segment and process the collected process videos.

[0010] The segmented process video is decomposed into images, and the images are labeled using the labeling software LabelImg to output the dataset;

[0011] The target detection algorithm based on Faster-RCNN was trained using a dataset to obtain a target detection model;

[0012] A trained target detection model is used to identify the workers' working status and calculate the specific working hours for each process.

[0013] As a preferred embodiment of the machine vision method for measuring work hours at workstations in a trousers production line according to the present invention, the step of setting up a camera above the trousers production line to collect video of workers' work includes the following steps:

[0014] For each workstation, the specific process steps are determined based on the operating procedures on the production line.

[0015] Time series analysis is used to mark the time points of each process step to form process progress information;

[0016] The processing cycle in the video is located based on the process step information, specifically:

[0017] Collect and record product information and time information (T) for each process at each workstation;

[0018] Based on the time information T, the start and end positions of each process are accurately located in the video, and the total number of frames N for each process is calculated.

[0019] As a preferred embodiment of the machine vision method for measuring workstation hours in a trouser workshop according to the present invention, the step of segmenting the collected process video using industrial video analysis software includes the following steps:

[0020] Based on the process flow information, key frames for each process in the video are extracted using image processing techniques.

[0021] Feature extraction is performed on keyframes to generate a dataset describing the characteristics of the process.

[0022] The target detection model is further optimized using the above dataset to improve the recognition accuracy for specific processes, specifically as follows:

[0023] The LabelImg tool was used to define the label categories to be identified in the dataset, and the main processes were divided into three categories: material picking idle, positioning alignment and sewing connection.

[0024] After annotation is complete, generate an XML file in PASCAL VOC format, ensuring that the ImageSets folder contains images from both the test and training sets.

[0025] As a preferred embodiment of the machine vision method for measuring workstation hours in a trousers workshop according to the present invention, the step of decomposing the segmented process video into images, labeling the images using the labeling software LabelImg, and outputting a dataset includes the following steps:

[0026] Each frame in the video is classified using a combination of manual and automatic annotation.

[0027] For the three main processes of material picking, positioning, and sewing, clear operating standards and time quotas are defined for each, including T d Material picking and positioning T d and sewing T d ;

[0028] All idle time is included in the material handling operation time, simplifying the labeling workload without affecting the accuracy of the overall working time. Specifically:

[0029] Calculate the actual material collection time T a =T q +Idle time, where idle time is T a -T q .

[0030] As a preferred embodiment of the machine vision method for measuring work hours in a trouser workshop according to the present invention, the step of using a dataset to train a Faster-RCNN-based object detection algorithm to obtain an object detection model includes the following steps:

[0031] A trained object detection model is used to monitor the worker's operational behavior in the video in real time;

[0032] Each detected operation is timestamped to form a detailed operation time series T. i ;

[0033] By statistically analyzing the time series of each operational action, the total operation time and the working hours of each specific process are calculated, as follows:

[0034] Let the total operation time be For a certain process e, the total number of frames is N. e Then the working time of this process FPS stands for video frame rate.

[0035] As a preferred embodiment of the machine vision method for determining work hours at workstations in a trousers workshop according to the present invention, the step of using a trained target detection model to identify the worker's working status and calculate the specific work hours for each process includes the following steps:

[0036] Based on the calculated total operation time and the working hours of specific processes, combined with the pre-set material handling time quota T q After deducting positioning time and sewing time;

[0037] The remaining time is calculated as the actual material collection time plus the idle time, specifically:

[0038] T a =∑T i (material collection time) + ∑T i (Free time);

[0039] Where, ∑T i (Materials collection time) = T q ,∑T i (Idle time) = T a -T q ;

[0040] Operations that affect working hours are divided into three categories: material retrieval, positioning, and sewing. Idle time plus the quota material retrieval time is included in the material retrieval time and marked accordingly.

[0041] Based on the statistical results of process time, a dual-label network planning diagram is constructed to optimize the production line layout.

[0042] As a preferred embodiment of the machine vision method for measuring workstation hours in a trouser workshop according to the present invention, the method for optimizing the production line layout includes the following steps:

[0043] Generate a human-machine workstation diagram based on the double-label network planning diagram to optimize the production line layout;

[0044] Generating a human-machine workstation diagram for the ZD17804 jeans production line based on the network planning diagram;

[0045] After optimization, the production line layout reduces material handling distances, lowers physical labor consumption, and saves handling time. Specifically:

[0046] By reducing material handling distances and minimizing workers' physical exertion, logistics between each workstation becomes more efficient, thereby improving overall production efficiency.

[0047] Based on the human-machine workstation diagram, the machine vision system is used to measure the working time of repetitive work processes.

[0048] As a preferred embodiment of the machine vision method for measuring work time at workstations in a trousers workshop according to the present invention, the step of measuring the work time of repetitive work processes using a machine vision system based on a human-machine workstation diagram includes the following steps:

[0049] After determining the human-machine workstation diagram, fix the work scene, and use the machine vision system to measure the working time of repetitive work processes.

[0050] By analyzing repetitive work processes in the video, the efficiency and effectiveness of the work are evaluated, specifically as follows:

[0051] The confusion matrix is ​​used to evaluate model performance, and the detection accuracy, recall, and mean precision are calculated.

[0052] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the machine vision method for measuring workstation hours in a trouser workshop as described in the first aspect of the present invention.

[0053] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein, when the computer program is executed by a processor, it implements any step of the machine vision method for measuring workstation hours in a trouser workshop as described in the first aspect of the present invention.

[0054] The beneficial effects of this invention are as follows: By installing cameras above the trouser production line to collect work videos, and combining industrial video analysis, image segmentation, deep learning target detection model training, and process recognition technology, the automatic identification and accurate calculation of the operation time of workers in each process are realized. The method introduces time series analysis to locate the processing cycle, uses the Faster-RCNN model to identify key actions such as material picking, positioning, and sewing, and reasonably classifies idle time to improve annotation efficiency. At the same time, the layout of human-machine workstations is optimized based on the statistical results of process time, reducing material handling distance and physical consumption, which significantly improves production management efficiency and time measurement accuracy. By constructing a standardized, automated, and quantifiable time measurement system, it provides effective support for lean management and intelligent upgrading of garment manufacturing workshops. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0056] Figure 1 This is a flowchart of the machine vision method for measuring work hours at workstations in the trouser workshop in Example 1.

[0057] Figure 2 This is a network diagram of the ZD17804 style jeans in Example 2.

[0058] Figure 3 This is a diagram of the human-machine interface on the YD17802SC jeans production line in Example 2.

[0059] Figure 4 This is a diagram illustrating the time measurement process in Example 2.

[0060] Figure 5 This is a video cycle diagram of the processed product in Example 3.

[0061] Figure 6 This is a schematic diagram of the Faster-RCNN algorithm network architecture in Example 3.

[0062] Figure 7 This is a schematic diagram of the camera installation on the trouser production line in Example 2.

[0063] Figure 8 This is a flowchart of the image library construction process in Example 2. Detailed Implementation

[0064] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0065] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0066] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0067] Example 1, referring to Figures 1 to 8 This is the first embodiment of the present invention, which provides a machine vision method for measuring work hours at workstations in a trousers workshop, including the following steps:

[0068] S1. Install cameras above the trouser production line to collect video of workers' operations;

[0069] Furthermore, for each workstation, the specific process steps are determined based on the operating procedures on the production line;

[0070] Time series analysis is used to mark the time points of each process step to form process progress information;

[0071] The processing cycle in the video is located based on the process step information, specifically:

[0072] Collect and record product information and time information (T) for each process at each workstation;

[0073] Based on the time information T, the start and end positions of each process are accurately located in the video, and the total number of frames N for each process is calculated.

[0074] It should be noted that the camera installation position has been optimized through on-site testing to ensure complete coverage of the workstation area and avoid obstruction. The video sampling frequency is consistent with the production line rhythm to ensure the accuracy of process time measurement. The process over-point information comes from manual records or process flow data in the MES system, which is used to align the video frame sequence with the actual operation time node, thereby realizing the foundation for high-precision time calculation based on vision.

[0075] S2. Use industrial video analysis software to segment the collected process videos;

[0076] Furthermore, based on the process step information, image processing technology is used to extract key frames for each process in the video;

[0077] Feature extraction is performed on keyframes to generate a dataset describing the characteristics of the process.

[0078] The target detection model is further optimized using the above dataset to improve the recognition accuracy for specific processes, specifically as follows:

[0079] The LabelImg tool was used to define the label categories to be identified in the dataset, and the main processes were divided into three categories: material picking idle, positioning alignment and sewing connection.

[0080] After annotation is complete, generate an XML file in PASCAL VOC format, ensuring that the ImageSets folder contains images from both the test set and the training set.

[0081] It should be noted that the video segmentation process adopts a method based on a combination of keyframe extraction and action recognition to ensure that the segmentation results correspond one-to-one with the process steps. Image feature extraction includes dimensions such as color, texture, and motion trajectory to enhance the representativeness of the model training samples. The generated dataset strictly follows the PASCAL VOC standard format, which facilitates the training and evaluation of the subsequent object detection model, while ensuring that there is no data leakage between the training set and the test set.

[0082] S3. Decompose the segmented process video into images, and use the labeling software LabelImg to label the images, and output the dataset;

[0083] Furthermore, a combination of manual and automatic annotation is used to classify each frame in the video;

[0084] For the three main processes of material picking, positioning, and sewing, clear operating standards and time quotas are defined for each, including T dMaterial picking and positioning T d and sewing T d ;

[0085] All idle time is included in the material handling operation time, simplifying the labeling workload without affecting the accuracy of the overall working time. Specifically:

[0086] Calculate the actual material collection time T a =T q +Idle time, where idle time is T a -T q ;

[0087] It should be noted that time quotas are introduced as an auxiliary basis for judgment during the annotation process. By setting operational standards for three main behaviors—material picking, positioning, and sewing—the annotation deviation caused by subjective judgment is reduced. Idle time is uniformly included in the material picking time, which simplifies the annotation complexity without affecting the accuracy of the overall working time statistics, improves the efficiency of data preparation, and is suitable for the rapid construction of large-scale video data.

[0088] S4. Use the dataset to train the Faster-RCNN-based object detection algorithm using machine learning to obtain the object detection model;

[0089] Furthermore, a trained object detection model is used to monitor the workers' operational behavior in the video in real time;

[0090] Each detected operation is timestamped to form a detailed operation time series T. i ;

[0091] By statistically analyzing the time series of each operational action, the total operation time and the working hours of each specific process are calculated, as follows:

[0092] Let the total operation time be For a certain process e, the total number of frames is N. e Then the working time of this process FPS stands for video frame rate;

[0093] It should be noted that the object detection model incorporates a transfer learning strategy during training to improve its generalization ability under small sample conditions. After deployment, the model can parse video streams in real time and form a structured operation behavior log by recording the timestamps of the target box coordinates and category labels. The statistics of the total number of frames in the process depend on the continuity judgment of the object detection results. The specific working time is calculated by combining the video frame rate, providing a quantitative basis for subsequent human-machine collaboration optimization.

[0094] S5. Use a trained target detection model to identify the worker's working status and calculate the specific working hours for each process.

[0095] Furthermore, based on the calculated total operation time and the working hours of specific processes, combined with the pre-set material handling time quota T... q After deducting positioning time and sewing time;

[0096] The remaining time is calculated as the actual material collection time plus the idle time, specifically:

[0097]

[0098] in,

[0099] Operations that affect working hours are divided into three categories: material retrieval, positioning, and sewing. Idle time plus the quota material retrieval time is included in the material retrieval time and marked accordingly.

[0100] Based on the process time statistics, a dual-label network planning diagram is constructed to optimize the production line layout;

[0101] Generate a human-machine workstation diagram based on the double-label network planning diagram;

[0102] Generating a human-machine workstation diagram for the ZD17804 jeans production line based on the network planning diagram;

[0103] After optimization, the production line layout reduces material handling distances, lowers physical labor consumption, and saves handling time. Specifically:

[0104] By reducing material handling distances and minimizing workers' physical exertion, logistics between each workstation becomes more efficient, thereby improving overall production efficiency.

[0105] Based on the human-machine workstation diagram, the machine vision system is used to measure the working time of repetitive work processes.

[0106] After determining the human-machine workstation diagram, fix the work scene, and use the machine vision system to measure the working time of repetitive work processes.

[0107] By analyzing repetitive work processes in the video, the efficiency and effectiveness of the work are evaluated, specifically as follows:

[0108] The confusion matrix is ​​used to evaluate model performance, and the detection accuracy, recall, and mean precision are calculated.

[0109] It should be noted that the time calculation method is based on the classification and statistics of process behavior. By independently timing the three types of operations—material picking, positioning, and sewing—and combining the preset time quota with the deduction of non-effective operation time, refined time management is achieved. The dual-label network planning diagram is constructed based on this time data to guide the optimization of human-machine workstation layout, ultimately achieving the goal of reducing handling distance and improving logistics efficiency. The measurement of repetitive operation time further verifies the stability and reliability of the system, and the confusion matrix evaluation index provides an objective evaluation basis for the model performance.

[0110] This embodiment also provides a computer device suitable for the machine vision measurement method of workstation hours in a trouser workshop, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the machine vision measurement method of workstation hours in a trouser workshop as proposed in the above embodiment.

[0111] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0112] This embodiment also provides a storage medium on which a computer program is stored. When executed by a processor, the program implements the machine vision method for measuring workstation hours in a trouser workshop as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0113] In summary, this invention achieves automatic identification and accurate calculation of workers' operation time for each process by installing cameras above the trouser production line to collect operation videos and combining industrial video analysis, image segmentation, deep learning object detection model training, and process recognition technology. The method introduces time series analysis to locate the processing cycle, uses the Faster-RCNN model to identify key actions such as material picking, positioning, and sewing, and rationally classifies idle time to improve annotation efficiency. At the same time, it optimizes the layout of human-machine workstations based on process time statistics, reduces material handling distance and physical exertion, and significantly improves production management efficiency and time measurement accuracy. By constructing a standardized, automated, and quantifiable time measurement system, it provides effective support for lean management and intelligent upgrading of garment manufacturing workshops.

[0114] Example 2, refer to Figure 2 - Figure 8 This is the second embodiment of the present invention. This embodiment provides a method for generating human-machine workstation diagrams and analyzing work time composition in a machine vision method for measuring workstation hours in a trousers workshop, including the following steps:

[0115] S1. Generate human-machine workstation diagram

[0116] Based on the process breakdown and structure plan of ZD17804 denim trousers, the process tasks of 36 workers were combined and allocated according to experience, working hours, machine type and worker expertise.

[0117] Based on an average flow time of 23.933 seconds per second, processes with a flow time greater than this value (such as "upper waist" with 38 seconds) are decomposed into multiple workstations, while processes with a flow time less than this value (such as "connecting the horse belt" with 4 seconds) are combined and optimized.

[0118] Based on the double-label network diagram ( Figure 2 The process logic is analyzed to reduce material handling distances, generating a human-machine interface diagram for the production line. Figure 3 Specific optimization measures include:

[0119] Arrange closely connected upstream and downstream processes (such as "combining small parts" and "sealing parts") adjacent to each other;

[0120] Large-scale fabrication and handling processes (such as "attaching back pockets") are concentrated in the front and back fabrication work areas;

[0121] Processes using the same sewing machine (such as "pressing the waistband" and "attaching the washing label" on a single-needle sewing machine) are assigned to the same worker.

[0122] S2. Time Composition Analysis and Visual Measurement Preparation

[0123] Workstation hours are divided into three categories of operations ( Figure 4 ):

[0124] Material retrieval time: The quota is set to a fixed value (e.g., 5 seconds / time), and idle time is counted as material retrieval operation;

[0125] Positioning time: The time it takes for the cut piece and accessories to align with the machine feet, measured through visual recognition;

[0126] Sewing time: Machine sewing connection time, measured by visual recognition.

[0127] A single camera is installed above the front of the workstation. Figure 7 To ensure coverage of the worker's operating area, the video sampling frequency is synchronized with the production line cycle time (e.g., 30fps).

[0128] S3. Video Acquisition and Process Status Labeling

[0129] Industrial video analytics software (such as ECRS) is used to segment the acquired video by process, extract keyframes, and generate an image library. Figure 8 ).

[0130] Three states are defined using the LabelImg annotation tool:

[0131] Material handling idle time: Workers do not come into contact with the cut pieces or machines;

[0132] Positioning and alignment: The action of aligning the cut piece with the machine foot;

[0133] Sewing connection: Machine sewing operation status;

[0134] The labeled data was generated in PASCAL VOC format, containing 1683 training images and 421 test images.

[0135] S4. Time Measurement and Error Analysis

[0136] Using the Faster-RCNN model ( Figure 6 Detect the process status in the video frames, count the number of frames for each type of operation, and calculate the working time according to the formula:

[0137]

[0138] Compared with the manual measurement results of industrial video analysis software, the relative error of visual recognition time is controlled within 10%, which meets production requirements.

[0139] In summary, the human-machine workstation diagram reduces material handling distance by 30%, reduces worker physical exertion, introduces quota material retrieval time combined with visual recognition to improve time statistics efficiency, the single-camera solution is suitable for complex workshop environments, and the Faster-RCNN model achieves a detection speed of 15fps, meeting real-time requirements.

[0140] Example 3

[0141] Reference Figure 5 and Figure 6 This is the third embodiment of the present invention, which provides a method for workstation time status detection and multi-batch video analysis based on the Faster-RCNN model, including the following steps:

[0142] S1. Video periodic positioning and point-crossing information matching

[0143] Based on the characteristics of subcontracted production, the processing cycle in the video is located using the time-lapse information (product batch, process start / end time) from the MES system. Figure 5 ).For example:

[0144] The video segment of the "grouping" process for the first batch of products (n1 pieces) is aligned with the time when the cut pieces enter / leave the workstation using timestamps;

[0145] The video is divided into a “material picking-positioning-sewing” loop unit, with each unit corresponding to the processing of one product.

[0146] S2 and Faster-RCNN Model Training and Optimization

[0147] Hardware configuration: Intel i7-12700H, NVIDIA RTX 3060 GPU, 16GB RAM;

[0148] Software environment: Windows 11, Python 3.7, CuDNN 8.0.5, OpenCV 4.2.0;

[0149] Training parameters: Batch size = 2, 600 iterations, loss function converges to below 0.05;

[0150] Performance metrics: Precision 92%, Recall 88%, mAP 85%.

[0151] S3, Multi-batch Time Statistics and Production Optimization

[0152] Statistical analysis of the time measurement results from multiple consecutive batches of videos (e.g., 4 batches of products) revealed the following:

[0153] The time spent on the "waist-up" process fluctuates significantly (±15%) because the placement of the cut pieces is not fixed.

[0154] The "seam pressing" process is efficient with an error of only ±5%.

[0155] Adjust the production plan based on the results:

[0156] Add a cutting piece positioning bracket to the "upper waist" workstation to reduce positioning time;

[0157] Increase the workload of the "pressing seam" process by 10% to balance the production line pace.

[0158] In summary, the system enables precise correlation between work hours and production batches through time-based information, supporting quality traceability; it identifies bottleneck processes based on multi-batch data, allowing for targeted improvements to the production process; and the model supports integration with the MES system, enabling automatic uploading and analysis of work hour data.

[0159] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A machine vision method for measuring work hours at workstations in a trousers manufacturing workshop, characterized by: include: Cameras were installed above the trouser production line to collect videos of workers at their work. Industrial video analytics software was used to segment and process the collected process videos. The segmented process video is decomposed into images, and the images are labeled using the labeling software LabelImg to output the dataset; The target detection algorithm based on Faster-RCNN was trained using a dataset to obtain a target detection model; A trained target detection model is used to identify the workers' working status and calculate the specific working hours for each process.

2. The machine vision method for measuring work hours at workstations in a trouser workshop as described in claim 1, characterized in that: The method of installing cameras above the trouser production line to collect video of workers' operations includes the following steps: For each workstation, the specific process steps are determined based on the operating procedures on the production line. Time series analysis is used to mark the time points of each process step to form process progress information; The processing cycle in the video is located based on the process step information, specifically: Collect and record product information and time information (T) for each process at each workstation; Based on the time information T, the start and end positions of each process are accurately located in the video, and the total number of frames N for each process is calculated.

3. The machine vision method for measuring work time at workstations in a trouser workshop as described in claim 2, characterized in that: The process of segmenting the acquired process videos using industrial video analytics software includes the following steps: Based on the process flow information, key frames for each process in the video are extracted using image processing techniques. Feature extraction is performed on keyframes to generate a dataset describing the characteristics of the process. The target detection model is further optimized using the above dataset to improve the recognition accuracy for specific processes, specifically as follows: The LabelImg tool was used to define the label categories to be identified in the dataset, and the main processes were divided into three categories: material picking idle, positioning alignment and sewing connection. After annotation is complete, generate an XML file in PASCAL VOC format, ensuring that the ImageSets folder contains images from both the test and training sets.

4. The machine vision method for measuring work time at workstations in a trouser workshop as described in claim 3, characterized in that: The process of decomposing the segmented video into images, labeling the images using LabelImg software, and outputting a dataset includes the following steps: Each frame in the video is classified using a combination of manual and automatic annotation. For the three main processes of material picking, positioning, and sewing, clear operating standards and time quotas are defined for each, including T d Material picking and positioning T d and sewing T d ; All idle time is included in the material handling operation time, simplifying the labeling workload without affecting the accuracy of the overall working time. Specifically: Calculate the actual material collection time T a =T q +Idle time, where idle time is T a -T q .

5. The machine vision method for measuring work time at workstations in a trouser workshop as described in claim 4, characterized in that: The process of training a Faster-RCNN-based object detection algorithm using a dataset to obtain an object detection model includes the following steps: A trained object detection model is used to monitor the worker's operational behavior in the video in real time; Each detected operation is timestamped to form a detailed operation time series T. i ; By statistically analyzing the time series of each operational action, the total operation time and the working hours of each specific process are calculated, as follows: Let the total operation time be ∑T i For a certain process e, the total number of frames is N. e Then the working time of this process FPS stands for video frame rate.

6. The machine vision method for measuring work hours at workstations in a trouser workshop as described in claim 5, characterized in that: The process of using a trained target detection model to identify the worker's working status and calculate the specific working hours for each process includes the following steps: Based on the calculated total operation time and the working hours of specific processes, combined with the pre-set material handling time quota T q After deducting positioning time and sewing time; The remaining time is calculated as the actual material collection time plus the idle time, specifically: T a =∑T i (material collection time) + ∑T i (Free time); Where, ∑T i (Materials collection time) = T q ,∑T i (Idle time) = T a -T q ; Operations that affect working hours are divided into three categories: material retrieval, positioning, and sewing. Idle time plus the quota material retrieval time is included in the material retrieval time and marked accordingly. Based on the statistical results of process time, a dual-label network planning diagram is constructed to optimize the production line layout.

7. The machine vision method for measuring work time at workstations in a trouser workshop as described in claim 6, characterized in that: The optimization of the production line layout includes the following steps: Generate a human-machine workstation diagram based on the double-label network planning diagram to optimize the production line layout; Generating a human-machine workstation diagram for the ZD17804 jeans production line based on the network planning diagram; After optimization, the production line layout reduces material handling distances, lowers physical labor consumption, and saves handling time. Specifically: By reducing material handling distances and minimizing worker physical exertion, logistics between each workstation becomes more efficient, thereby improving overall production efficiency.

8. The machine vision method for measuring work hours at workstations in a trouser workshop as described in claim 7, characterized in that: The method of determining the working time of repetitive work processes based on human-machine workstation diagrams and using a machine vision system includes the following steps: After determining the human-machine workstation diagram, fix the work scene, and use the machine vision system to measure the working time of repetitive work processes. By analyzing repetitive work processes in the video, the efficiency and effectiveness of the work are evaluated, specifically as follows: The confusion matrix is ​​used to evaluate model performance, and the detection accuracy, recall, and mean precision are calculated.

9. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the machine vision method for measuring work hours at workstations in a trouser workshop as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the machine vision method for measuring work time at workstations in the trouser workshop as described in any one of claims 1 to 8.

Images