以嵌入式系統實作螢幕眼動儀
No Thumbnail Available
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本研究以Zedboard 嵌入式開發板，實作螢幕眼動儀。使用的硬體設備為ZedBoard、USB OTG、USB滑鼠、USB視訊攝影機、HDMI顯示器。嵌入式平台的嵌入式作業系統Linaro Ubuntu 12.04，系統移植OpenCV與Qt Creator，可提供螢幕眼動儀所需要使用的函式庫與人機介面環境。系統功能為偵測眼睛之瞳孔影像，找出瞳孔中心位置，使用二階多項式完成螢幕眼動儀的校正後，可透過HDMI螢幕畫面上得知使用者凝視螢幕上的位置。用意是記錄使用者凝視軌跡，計算出人眼凝視點的位置。嵌入式環境中使用暗瞳孔法取得較深色的瞳孔影像資訊，攝影機取得的影像資料，濾除眉毛、眼皮等影像資訊，使用標籤演算法計算出瞳孔中心的位置較為準確。進行九點校正的過程中，操作者保持頭部固定，避免頭部移動所產生的系統誤差。研究成果為使用者與電腦螢幕距離40公分處，凝視誤差為0.69度，而距離更改為80公分處，凝視誤差為0.44度。所得到的凝視誤差結果與其他嵌入式螢幕眼動儀比較凝視點誤差，結果為本研究製作的嵌入式螢幕眼動儀之凝視誤差較好，完成ZedBoard實作螢幕眼動儀。
In this study, used Zedboard completed embedded system implementation screen-based eye tracker. Hardware device used ZedBoard, USB OTG, USB mouse, USB video camera, HDMI display. Embedded platform is Linaro Ubuntu 12.04 embedded operating system, transplantation OpenCV and Qt Creator. Procedure for the screen-based eye tracker is eigenvalue of the pupil to complete the mapping of the human eye and the computer screen, using second order polynomial calibration. The user can know the gaze on the HDMI display coordinate position. Dark pupil method made darker pupil image information for embedded environment. Camera image obtained filtered off eyebrows, eyelids and other image information. Use labeling algorithms to calculate the center of the pupil position more accurate. Research results for the user and the computer screen at a distance of 40 cm, gaze error of 0.69 degrees, and the distance is changed to 80 centimeters, gaze error of 0.44 degrees. And other embedded system screen-based eye tracker comparison gaze point error, the results of this study gaze point error better.
In this study, used Zedboard completed embedded system implementation screen-based eye tracker. Hardware device used ZedBoard, USB OTG, USB mouse, USB video camera, HDMI display. Embedded platform is Linaro Ubuntu 12.04 embedded operating system, transplantation OpenCV and Qt Creator. Procedure for the screen-based eye tracker is eigenvalue of the pupil to complete the mapping of the human eye and the computer screen, using second order polynomial calibration. The user can know the gaze on the HDMI display coordinate position. Dark pupil method made darker pupil image information for embedded environment. Camera image obtained filtered off eyebrows, eyelids and other image information. Use labeling algorithms to calculate the center of the pupil position more accurate. Research results for the user and the computer screen at a distance of 40 cm, gaze error of 0.69 degrees, and the distance is changed to 80 centimeters, gaze error of 0.44 degrees. And other embedded system screen-based eye tracker comparison gaze point error, the results of this study gaze point error better.
Description
Keywords
ZedBoard, 螢幕眼動儀, 嵌入式系統, ZedBoard, eye tracker, embedded system