A study on Application of Human Face Recognition to Protect VDT User's Eyes

指導教授 : 黃榮堂、陳正光   研究生 : 何肇偉  機電整合研究所 91年


摘要

  在本論文中,我們建立一套電腦使用者的眼睛保護系統,使用一個計時器來紀錄電腦使用者使用電腦的時間,同時經由放置在電腦螢幕上的視訊攝影機,擷取電腦使用者以及週遭工作環境的影像,應用人臉辨識的技術進行辨別,確認是否為相同的使用者在使用電腦,以客觀地估計使用者操作電腦的時間,並且利用人臉資訊來進行視距與眨眼的偵測,最後,若電腦使用者常常沒有保持適當的視距、每分鐘的眨眼太少或太多、長時間使用電腦,會使用圖片、文字和聲音來給予電腦使用者適當的警告,主動地達到保護眼睛的目的。
  在本論文中應用的技術包括人臉偵測、人臉辨識、視距與眨眼偵測等。在人臉偵測的部分中,使用建立的色彩自我校正法則,來消除環境光源對彩色影像的影響後,進行膚色抽取與膚色視窗的建立,以找出影像中的膚色區域,並在每個膚色區域中確認是否為人臉,完成影像中的人臉偵測。在人臉辨識的部分,使用影像處理均化人臉視窗後,結合邊緣偵測與二值化處理以抽取臉部特徵,由二值化的特徵影像找出臉部的中心線,沿著中心線抽取嘴巴與眼睛區域並且判斷使用者是否佩戴眼鏡,最後,將嘴巴與眼睛區域分為三個特徵區域,使用樣板比對的方式辨別出是否為不同的使用者在使用電腦。在視距與眨眼偵測部分,使用臉部的寬度來偵測視距,並且在人臉中進行動差運算,快速地偵測使用者是否眨眼,以每分鐘的眨眼次數來評估使用者的眼睛疲勞程度。

關鍵字:眼睛保護、人臉偵測、人臉辨識、視距偵測、眨眼偵測

ABSTRACT

  In this study, a system that can actively remind the VDT’s user to let his/her eyes take a break has been developed. The system uses a built-in timer to record how long the VDT user has been continuously working on the computer, and a PC camera on top of the computer monitor to simultaneously keep capturing the image of the VDT user and the environment as well. In order to objectively estimate how long a person continuously uses the computer, the developed system can determine whether the same person keeps using the computer through face recognition technology. It can detect the visual distance and eye’s blinking by the facial feature information. If the user doesn’t keep a proper visual distance, or his/her eyes blink too often or too less, or has continuously been using the computer for a long time, the system will pop up a warning message in words/picture/voice to the user. Therefore, this developed system can initiatively protect VDT user’s eyes.
  This study applies face detection, face recognition, visual distance and eye’s blinking detection. Regarding the face detection, a self-calibration algorithm of has been presented to eliminate the influence of color image caused by the light source of environment. Then, skin extraction and creating a skin window are followed to find the skin region of the image. Each skin region has to be checked to identify human face so as to complete the process of face detection. In face recognition, the image of face window is equalized by the technique of image process. The facial feature, then, can be extracted by combining edge detection and binary process. According to the features of the binary image, the central line of the face part can be determined. Then, the mouth and eyes region cam be extracted along the central line and whether the user wearing glasses can also be identified. Three distinct feature regions are taken from the mouth and eyes parts in order to identify whether the same user is using the computer by the template matching method. Regarding the visual distance, the distance is detected by the width of user’s face. In the eye’s blinking detection, the movement operation of face elements is carefully tracked to detect user’s eye blinking. How exhausted of the user’s eyes is, then, estimated by the eye’s blinking per minute.

Keywords: protecting eyes, face detection, face recognition, visual distance detection, and eye’s blinking detection