本研究以手機鏡頭取代額外的攝影機，在擷取影像後以美國國家儀器公司LabVIEW®之開發環境下實現以眼睛閉合時間為理論基礎之疲勞駕駛辨識系統。此系統在獲得影像後從三原色(RGB)影像中萃取飽和度(Saturation)與色調(Hue)影像，以可降低光線影響之自動二值化之尼布雷克法(NiBlack Threshold)做為基礎，將影像經過多重演算法並堆疊，最後再以凸包法(Convex Hull)萃取完整臉部面積，此時系統亦從RGB影像中萃取亮度(Luminance)影像，以相同二值化方法，將影像經過多重演算，最後再以海伍德圓度法(Heywood Circularity)過濾出眼睛部分並判別眼睛粒子數量(NoP)。隨後將萃取後的完整人臉的上半部面積做為眼睛萃取的區域，進一步降低雜訊的出現。最後以ㄧ組布林累積器(Boolean Accumulator)做為疲勞駕駛判斷的依據，並降低系統因高速擷取布林訊號而導致誤判的機率。系統在NoP小於2時，以電腦同時釋放視覺與聽覺的警告訊息，提醒駕駛人避免疲勞駕駛事件的發生。;In this study, a smartphone lens substituted additional camera in capturing images under National Instrument LabVIEW® development environment to achieve a driving fatigue supervision system based on eye blink analysis. First, the system extracts Saturation and Hue plane from RGB image. In order to reduce the light impact, an automatic binarization with NiBlack threshold was introduced to stack images through multiple algorithms before extract complete face area with Convex Hull. The system also extracted Luminance plane afterward from RGB image with the same binarization through multiple algorithms to filter eye part out and discriminate number of particles (NoP) with Heywood Circularity. Then the system extracted eyes region further on upper half of face for reducing possible noises after a complete face extraction. Subsequently, a set of Boolean Accumulator was developed for judging fatigue driving, and reduce possible failures due to high-speed Boolean signal capturing. At last, the system released both visual and auditory warnings simultaneously by a computer to alert drivers of avoiding fatigue driving when NoP is less than 2.
