Related Experiment Video
Updated: Sep 22, 2025
12:39
A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers
Published on: January 18, 2020
7.8K
Driver's Head Pose and Gaze Zone Estimation Based on Multi-Zone Templates Registration and Multi-Frame Point Cloud
Yafei Wang1, Guoliang Yuan1, Xianping Fu1
1School of Information Science and Technology, Dalian Maritime University, Dalian 116026, China.
Sensors (Basel, Switzerland)
|May 20, 2022
Summary
This study introduces a new method for accurately estimating driver head pose and gaze zone using RGB-D cameras. The approach enhances point cloud data fusion and registration, improving driver visual attention analysis in complex driving scenarios.
Area of Science:
- Computer Vision
- Automotive Safety
- Human-Computer Interaction
Background:
- Driver head pose and eye gaze are critical for analyzing visual attention.
- Existing methods struggle with naturalistic driving due to illumination variations and occlusions.
- Point cloud data is susceptible to errors from partial facial occlusion and incorrect feature extraction.
Purpose of the Study:
- To propose a novel method for accurate driver head pose and gaze zone estimation.
- To address challenges in naturalistic driving scenes, including non-uniform illumination and head rotation.
- To improve the robustness of visual attention analysis using RGB-D cameras.
Main Methods:
- A point cloud fusion and registration strategy using continuous multi-frame data.
- Utilizing nearest neighbor gaze zone point clouds as templates for registration.
- Employing Normal Distributions Transform (NDT) with particle filter for coarse-to-fine transformation.
- Training a gaze zone estimator by combining head pose and multi-scale sparse coding eye image features.
Main Results:
- The proposed strategy significantly improves head pose tracking accuracy.
- The method demonstrates a low error rate in gaze zone classification.
- Enhanced robustness against partial facial occlusion and illumination variations was observed.
Conclusions:
- The novel method effectively estimates driver head pose and gaze zone in complex driving conditions.
- The point cloud fusion and registration strategy enhances the reliability of visual attention analysis.
- This approach offers a promising solution for advanced driver assistance systems.
