Driver fatigue state image recognition method based on facial video stream analysis

By mapping facial video stream analysis to a Riemannian manifold space and constructing an attention entropy field model, the problem of fatigue detection failure caused by head posture deflection and illumination changes in existing technologies is solved, enabling accurate identification and predictive intervention of driver fatigue.

CN121999472BActive Publication Date: 2026-06-23GUIZHOU INST OF TECH
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
GUIZHOU INST OF TECH
Filing Date
2026-04-10
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing driver fatigue detection technologies are prone to failure when faced with head posture changes and sudden changes in lighting, and they are difficult to capture the early subtle characteristics and trends of drivers' efforts to combat fatigue, resulting in false positives and false negatives.

Method used

A method based on facial video stream analysis is adopted to construct a covariance matrix and map it to a symmetric positive definite Riemannian manifold space. The geodesic distance is calculated using affine invariant Riemannian metric, and an attention entropy field model is constructed. The fatigue state is identified by the potential energy well threshold, and the Euclidean distance tracking mode is switched under abnormal working conditions.

Benefits of technology

It effectively separates rigid head movements from facial expression changes, accurately captures fatigue characteristics, reduces the misjudgment rate, and enables refined classification assessment and predictive intervention of fatigue status, ensuring high availability and real-time performance of the system in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121999472B_ABST
    Figure CN121999472B_ABST
Patent Text Reader

Abstract

The present application relates to intelligent driving assistance and computer vision technical field, specifically for driver fatigue state image recognition method based on face video stream analysis, including the following steps: step one, collecting the face video stream of the driver, extracting the face key point sequence of continuous frame, and forming the manifold feature point;Step two, in the symmetric positive definite Riemann manifold space, the affine invariant Riemann metric is introduced, the geodesic distance of manifold feature point relative to the preset reference clear state point is calculated, and the expression deformation feature is obtained;Step three, the attention entropy field model is constructed, the drift speed of manifold feature point on the surface of manifold is calculated, and is converted into the entropy increase rate of system;Step four, the potential well threshold is preset, the fatigue state evaluation result is obtained, and the corresponding driver state instruction is generated.The present application can still accurately extract the real fatigue feature under the condition that the driver is not looking straight ahead or the head posture frequently changes, and significantly reduces the misjudgment rate caused by posture change.
Need to check novelty before this filing date? Find Prior Art