現今以駕駛疲勞偵測是一個重要的議題。然而，利用眨眼的方式來判斷打瞌睡程度方法上可分為，直接接觸式量測與間接非接觸式量測。本研究結合即時眼動圖和影像處理，抽取相關的生理參數指標評估人清醒與否。
直接眼動圖的量測上，吾人利用表面電極與前置電路透過NI的資料擷取卡傳輸至電腦，從結果上顯示正常眨眼較緩慢眨眼動作時間較短。在間接量測上，採用網路攝影機固定在帽緣上作影像擷取，如此便不需使用臉部追蹤，再利用「閉眼百分比（Percentage of Eye Closure, PERCLOS）」演算法，分別對假體與正常受測者做測試。利用30秒重置系統的時間槽取樣方式，使生理訊號與影像能夠有80以上的相關係數。結合上述兩種方法，可計算出在影像上正常眨眼動作時間在0.22秒至0.43秒，而生理訊號上為0.21秒至0.39秒，此數據相當接近美國高速公路管理局的結果。
本論文亦分析正常光源和紅外線光源下，正常眨眼與緩慢眨眼的數據，雖然在光源仍會影響結果，但數據上顯示演算法是有一致性，因此以間接影像處理與直接量測眼動圖評估打瞌睡，此法不但簡化了演算法，克服網路攝影機傳輸延遲的問題，更提供了一種簡單而快速計算疲勞指標。

Nowadays, driver-fatigue-detection was became an important issue. However, eyelid opening and closing measurements in drowsy state evaluation were divided into direct and indirect ways. This project was combined real-time EOG and image processing methods to assess awareness by extracting related physiological parameters.
As to direct measurement, one could transfer EOG digitized signal to computer, though surface electrodes, analog circuits and NI acquisition card. The duration of eyelid-opening showed normal behavior was shorter than slow one. In indirect way, to avoid worse performance in face-tracking algorithm, we mounted webcam on the edge of hat to acquire images. Again, we adopted Percentage of Eye Closure (PERCLOS) algorithm in phantom’s and volunteer’s study. For improving the correlation coefficients above 80% from EOG and image data, we chose time-slot method, i.e. 30 seconds to reset the acquiring system. From the above two methods, one could calculate duration between two eyelid-opening about 0.22~0.43 and 0.21~0.39 second in image and signal, respectively. These data were consistent to the report of Federal Highway Administration in U.S.A.
We analyzed data of normal and slow eyelid-opening behavior in normal and infrared assisted circumstances. Although lighting problem still existed, yet there showed consistence in performance of algorithm. In this paper, we proposed characteristic methods in simplified algorithm and overcome transferring delay problem in webcam. Thus, combined real-time EOG and image processing methods, might provide an simple and quick methods to investigate drowsy state.

[1]Knipling, R.R.; Wang, J.S.; and Kanianthra, J.N., "Current NHTSA Drowsy Driver R&D". Fifteenth International Technical Conference On the Enhanced Safety of Vehicles, Melbourne, Australia, May, 1996.
[2]R. J. (Ed.). Carroll, Ocular measures of driver alertness: Technical conference proceedings," FHWA Technical Report No. FHWA-MC-99-136. Washington, DC: Federal Highway Administration, Motor Carrier and Highway Safety, 1999.
[3]Wierwille, W.W., Lewin, M.G., Fairbanks, RJ, and Neal,V.L., Research on Vehicle-Based Driver Status Performance Monitoring; 3 Volume Final Report. Vehicle Analysis and Simulation Laboratory, Virginia Polytechnic Institute and State University, VPSIU Report No. ISE 96-06 through 96-08, Contract Number DTNH 22-91-Y-07266, NHTSA Report No. TBD, September, 1996.
[4]Papanikolopoulos, N.P., Nelson, B., and Khosla, P.K., "Six degree-of-Freedom Hand Eye Visual Tracking with Uncertain Parameters", IEEE Transactions on Robotics and Automation, Volume 11, Number 5, pp 725-732, October 1995.
[5]Horng WB and Chen CY, "A Real-Time Driver Fatigue Detection System Based on Eye Tracking and Dynamic Template Matching," Tamkang Journal of Science and Engineering, Vol. 11, No. 1, pp. 65-72. (EI, NSC 92-2213-E-032-031) , March 2008
[6]R. C. Gonazlez and R. E. woods, Digital image processing, Addison-Wesley, 1992.
[7]鍾國亮，“影像處理與電腦視覺”，東華書局，2002。
[8]Rafael Barea, Luciano Boquete, Elena Lopez, Manuel Mazo, “Wheelchair Guidance Strategies Using EOG”, Journal of Intelligent and Robotic Systems, Volume 34, Number 3, 279-299,2002