The Research On Head-Mounted Gaze Tracking Technology

Eye, as the visual organ of human beings, is the important tool for human to perceive outside information and express the inner world to the outside. About 80% of the information in the brain is obtained by visual access. The line of sight refers to the gaze direction of human. Through gaze tracking and locating, human perception and intention to outside information can be obtained. The gaze tracking technology that using the computer to accurately track person’s visual process, has always been the hot spot of research. With further study of gaze tracking technology, the application of gaze tracking is increasingly wide. Its application field include the development of psychology, medical diagnose, commercial evaluation, entertainment, transportation, military, human-computer interaction, etc. Gaze tracking has a very good prospect in application.In view of the existing problems of gaze tracking technology, we put forward a design scheme, a kind of head-mounted gaze tracking system. The hardware structure of this system is simple and allows the head to move freely and has good tracking performance. At the same time, the software algorithm is studied and optimized. The main algorithms include the pupil algorithm, corneal reflection bright spot extraction algorithm, the calibration algorithm and the tracking algorithm. The innovation of this system is mainly in the following several aspects:1. Optimize the traditional gaze tracking system and relieve the restrictions on people: eye camera, scene camera and infrared light sources are integrated into the helmet, they remain relatively static. Thus the movement of the head will not affect the mapping relationship between scene picture and eye picture, and will not affect the tracking accuracy.2. Improve the precision of algorithm:clear images are obtained by eye camera, then we use image processing, such as binarization, Canny operator and ellipse fitting process, to position the pupil center and corneal reflection bright spot accurately.3. Improve the efficiency of algorithm and meet the real-time requirement:the algorithm use the non-iterative method, which do not need remove the corneal reflection bright spot in advance and do not do any previous processing work to image. At the same time, for the extraction of corneal reflection bright spot, we only calculate a part of eye image. So the operation speed is improved.In this paper, the error of head-mounted gaze tracking system is also calculated and the influences of various factors on gaze tracking system are analyzed. Through the experiment, the error of gaze tracking is estimated and calculated. The head-mounted gaze tracking system is proposed in this paper, its average error is around 0.4°in X direction and Y direction. The processing speed is 23 frame/s, meeting the requirement of gaze tracking system.Finally, for gaze tracking application in the field of augmented reality display, we combine gaze tracking system with helmet display system and design a set of integrated system. The system can use gaze tracking system to provide user’s gaze direction and interested area in the scene for the helmet display system. According to user’s information provided by gaze tracking system, helmet display system register the relevant augmented reality information to user’s fixation. Then 3D registration, real-time interaction and virtual-real fusion of augmented reality display system can be realized. According to the application, we design the optical system of HMD and put forward a set of design scheme. But the design is still in the preliminary stage and need further research and argument.