可預估3D環境中目標物座標的眼動儀建構

Abstract

本研究提出一種可以預估現實的3D環境中，眼睛凝視目標物之座標的眼動儀的建構方法。透過頭戴式的眼動儀設計與雙眼數據處理，達到分析3D環境的效果，同時將運算出來的結果，與由伺服馬達組成並搭配雷射筆的2軸控制平台連線，由雷射筆指出當前使用者凝視的目標。 眼動儀顧名思義即為透過偵測眼球的特徵變化與瞳孔轉動，去推斷使用者目前的凝視位置；而可預估3D目標物座標的眼動儀則是利用特殊的演算法，將2D環境(螢幕)中的預估凝視點轉為3D環境的預估凝視點，藉此預估目標物座標。本研究改善了前人提出的演算法中，有效預估距離過短及誤差過大的情形，將有效預估距離提升至100公分，平均誤差降低至2度。 除了目前頭戴式眼動儀最廣為使用2D映射法，本研究也提出一種改良的建模演算法來預估眼睛的凝視視線，本方法改善了原來建模法計算複雜的缺點，透過求出眼睛角膜上的光源反射點與瞳孔折射點，建立起整個眼球的模型，並預估出使用者眼球的光軸與視軸，達到預估凝視點的效果。

This research presents an approach to build an eye tracker for 3D target estimation. First, a wearable eye tracer, which two eye cameras are set on, is used. By analyzing the data calculated by binocular eye tracker, the 3D coordinates of the target can be estimated. Second, the two-axis control platform built by two servo motors and a laser pen are used to point the object user is looking at by receiving the 3D data obtained from the eye tracker system. Eye tracker is a device which estimates the position the user stares at by detecting the moving pupil. In this research, 2-D points of glint are transformed into 3-D points of glint to calculate the coordinates of object in real environment. Besides, this research increases the effective estimated distance by 100cm, and decreases the error by 2 degree. In addition to use 2D Mapping, this research also presents animproved 3D Modeling algorithm to estimate the gaze line (visual axis).