Research on Pattern Recognition-Based Detection Method for Sightline Estimation
Wenbo Huang, Mingwei Zhao, Shuai Wang, Heng Zhang, Lingping Kong · Advances in transdisciplinary engineering · 2025
To overcome the limitations of imprecise spatial line-of-sight (LOS) gaze point calculation and target detection, this study proposes an innovative data fusion method that integrates both head and eye movement information. Traditional gaze tracking methods often suffer from accuracy issues due to the inability to effectively combine head and eye movement data. This study introduces a data fusion neural network model, which establishes a complex relationship between head and eye movement and the LOS direction. The model enhances spatial gaze computation by enabling the precise alignment of gaze direction with the actual target in space. By combining binocular LOS data with environmental factors such as lighting and the orientation of the cockpit, the model is capable of determining the exact focal point of the gaze in real-time. This approach improves both gaze accuracy and target tracking performance. The method is evaluated through simulations of flight cockpit environments, where visual targets are analyzed for characteristics such as type, size, and motion patterns. The results demonstrate that the proposed method significantly enhances LOS-based target detection, providing improved accuracy and robustness over traditional methods. This data fusion approach offers a powerful framework for applications in aviation, human-computer interaction, and other fields requiring precise gaze tracking and target detection. The framework also has the potential to enhance situational awareness and decision-making processes in dynamic and high-stakes environments, making it an essential tool for future human-computer interface developments.