Research And Implement Of Eye Tracking System Based On Dual Cameras And Dual Light Sources

Vision is one of the important ways for humans to interact with the outside world.More than 80% of external information is obtained through vision.Gaze estimation technology calculates the observer's gaze point through optical,electronic,mechanical and other methods.It not only provides a new visual-based human-computer interaction method,but also provides a method to quantify the human eye vision and gaze law.It has great research value.This paper deeply studies the core technology of gaze tracking,learns and summarizes the achievements of the predecessors,explores the design and implementation of eye tracking system under low-cost hardware conditions from the perspective of macro system design.This system adopts the hardware design of dual cameras and dual light sources.The gaze estimation model is based on the pupil center and corneal reflection method.This system uses a low-cost infrared camera without industrial lens.This paper designs each step of the system in a targeted manner and balances the contradiction between hardware cost and accuracy of gaze estimation.The main work is as follows:1.The design of the hardware system.It includes selection of infrared point source and camera,installation and debugging and to achieve the best image quality.2.Extraction of gaze feature parameters,extraction of the pupil center and the corneal reflection center.The low-cost camera leads to low image quality,low resolution of the target area image.There are many other factors that determine like change of the illumination and human eye movement.The system proposes a method for extracting gaze feature parameters and takes the accuracy of the extraction and the robustness of the method into account.This method includes video acquisition,pupil localization,the corneal reflection center extraction,the pupil center extraction,and the post-processing to deal with the abnormal situation.3.Design of the gaze estimation algorithm.According to the physiological structure and visual mechanism of the human eye,the human eye model is abstracted.The gaze estimation algorithm based on the pupil center and corneal reflection method is derived.4.Coordinate system transformation and parameter calibration.It involves the converting between the local coordinate system and the world coordinate system,the calibration of camera parameters and camera poses,the derivation and optimization of the algorithm parameter calibration process.5.Implementation of the eye tracking system.It includes the implementation and integration of each aspect of the system,optimization and acceleration of the complete system.This paper finally completes an eye tracking system based on dual cameras and dual light sources.The experiment shows that the average root mean square error of the system is 12 mm,which meets the basic needs of some applications.The system reduces the hardware requirements,and has a great room for improvement.The system has certain reference value in practical applications.