Job track compliance monitoring method based on reinforcement learning tracking mechanism

By analyzing the CAD spatial topology of a textile factory and combining the lightweight FootNet regression network and the DQN tracker, the problems of insufficient positioning accuracy and semantic understanding in textile factory operation trajectory monitoring are solved, achieving high-precision, low-false-report trajectory compliance monitoring, supporting intelligent manufacturing and safe production.

CN122157168APending Publication Date: 2026-06-05HUAQIAO UNIVERSITY +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUAQIAO UNIVERSITY
Filing Date
2026-05-06
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing methods for monitoring the operational trajectory in textile factories suffer from problems such as insufficient accuracy in foot positioning, lack of spatial semantic understanding, and rigid tracking strategies in complex production environments, leading to unstable detection, discontinuous trajectories, and high recognition errors.

Method used

A reinforcement learning-based method for monitoring work trajectories in compliance mode is adopted. By analyzing the CAD spatial topology of a textile factory and combining the lightweight FootNet regression network and DQN tracker, high-precision foot projection point localization and trajectory tracking are achieved. Reinforcement learning is used to adjust the tracking strategy to adapt to complex environments.

Benefits of technology

It achieves high-precision, low-false-alarm-rate, and robust trajectory compliance monitoring in complex textile factory environments, supporting intelligent manufacturing and safe production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122157168A_ABST
    Figure CN122157168A_ABST
Patent Text Reader

Abstract

The application relates to a kind of job track compliance monitoring methods based on reinforcement learning tracking mechanism, with CAD drawing as data source, analysis constructs factory space topology network;Uniform CAD engineering coordinates and video physical coordinates, with YOLOv3 detection frame as the region of interest input lightweight foot bottom key point regression network FootNet and output foot bottom subpixel level coordinates, on the job identification adopts whether foot bottom projection point falls into station area and YOLOv3 confidence degree is jointly constrained, when condition satisfies, trigger track, the deviation between foot bottom projection point and standard track point is main feedback basis, by reconstructing reward function guide DQN tracker automatically adjust tracking strategy, keep track continuity, realize the stable matching of multi-target identity.The application constructs complete closed-loop control link by "space topology modeling-YOLOv3 detection trigger-FootNet foot bottom point extraction-tracking optimization-track compliance determination", can realize high-precision, low false alarm rate, strong robustness of track compliance monitoring in complex textile factory environment.
Need to check novelty before this filing date? Find Prior Art