高精確度眼角偵測用於可見光眼動儀之頭動補償

Abstract

人類接收資訊的來源有80%以上來自視覺。眼動儀透過追蹤眼球運動來獲得凝視點。透過分析視覺行為過程的資訊，可以應用於認知心理學、商業、醫學和人機介面等。為了更方便有效地蒐集眼動資訊，眼動追蹤技術也不斷的發展與創新。但是可見光眼動儀在支援頭部移動仍有許多困難與挑戰，例如：在影像中如何穩定地追蹤眼睛，如何建立空間中眼睛、攝影機與螢幕之間的相對位置關係，還有每個人主視眼不一的情況。本文提出的高精確度眼角偵測，用於定位與眼球的相對位置，在頭動補償中眼角偵測的座標也用於估計頭部在影像中的相對移動，以校正程序建立空間中眼睛、攝影機與螢幕之間的相對位置關係，並偵測主視眼。透過眼角偵測與頭動補償的方法，進一步改善了凝視點估計的精確度。

More than 80% of people get information from the visual. The gaze tracker obtains gaze points by tracking eye movements. By analyzing visual behavior, it can be applied to cognitive psychology, business, medicine, and human-machine interface. In order to collect eye movement information more conveniently and effectively, eye tracking technology continues to develop and innovate. However, visible eye trackers still have many difficulties and challenges in supporting head movements, such as how to stably track the eyes in the image, how to position the eyes in space, the relative positional relationship between the camera and the screen, and the situation in which each person has a different main eye. Inconsistent situations. The high-precision canthus detection proposed in this paper is used to locate the relative position of the eyeball. canthus detection is also used to estimate the motion of the head in the image. In order to obtain the position of the eyes, camera and screen in the space, and to detect the main eye. The accuracy of the gaze point estimation is further improved by the method of canthus detection and head motion compensation.