A subway driver standard finger difference calling behavior recognition method based on machine vision

By using machine vision technology and front and left-side camera equipment to collect video streams, key nodes of the subway driver are identified, and the angle and holding time are calculated. This solves the problem that the driver's call behavior depends on human judgment, improves the accuracy of recognition and the standardization of actions, and ensures the safety of subway operation.

CN115512338BActive Publication Date: 2026-06-19BEIHANG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
BEIHANG UNIV
Filing Date
2022-08-20
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Subway drivers' call-out behavior relies on human judgment and lacks corresponding safeguards, leading to inaccurate identification and non-standard actions, which affects the safe operation of the subway.

Method used

Using a machine vision-based approach, video streams are captured by cameras positioned in front of and to the left of the driver. These streams are then stitched together to form a single-source video. Key driver nodes are identified, and the angle between the arm and the vertical line of the body, as well as the arm bending angle, are calculated. Combined with a predetermined error range and holding time, the system determines whether the standard call-in action has been completed.

Benefits of technology

This improved the accuracy of recognizing drivers' finger-pointing and calling behaviors, ensuring that actions meet standards and enhancing the safety and orderliness of subway operations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115512338B_ABST
    Figure CN115512338B_ABST
Patent Text Reader

Abstract

This invention discloses a machine vision-based method for recognizing standard finger-pointing call behavior of subway drivers, comprising: real-time acquisition of video streams using cameras deployed at the front and sides; identification of key nodes based on the video streams; determination of the slope 'a' of the line connecting the nose node and the midpoints of the left and right shoulder nodes in the front camera's view, the slope 'b' of the line connecting the left wrist node and the left shoulder node in the side camera's view, and the slope 'c' of the line connecting the left shoulder node and the left ankle node in the side camera's view; and obtaining the included angle θ from the slopes 'a' and 'b'. arm1 The included angle θ is obtained from the slopes b and c. arm2 Determine the included angle θ respectively. arm1 With the included angle θ arm2 Whether it meets the preset error range; if it does, determine the action holding time according to the target tracking algorithm; if the action is held for more than the preset standard time, then the standard finger call action is determined to be completed. This invention assists the driver in making a standard action when making a finger call.
Need to check novelty before this filing date? Find Prior Art