圓形可變樣板應用於眼睛張開程度偵測

Abstract

在許多交通事故中，疲勞駕駛導致意外發生的事件層出不窮。於駕駛環境下，使用不妨礙駕駛人的視訊設備，觀測駕駛者眼睛的狀況，以達成瞌睡偵測是最直接有效的。亦即，經由偵測人眼開闔狀態來判斷是否出現瞌睡情形是相當準確可靠之方法。在此篇論文中，我們以彩色CCD 攝影機做為影像輸入來源，經過膚色判定來區分出人臉區域，再以主成份分析法擷取出眼部區域，再利用圓形可變樣版搜尋此區域，找出瞳孔的位置並加以分析之，以此判定眼睛此時的張開狀態。經實驗證明，我們提出的方法於判定眼睛狀態的準確度相當高，將有助於提高昏睡偵測系統的成功率。

Sleepiness and driving is a dangerous combination, drowsy driving can be just as fatal. Accordingly, it is necessary to develop a drowsy driver awareness system. To avoid interrupting the driver, it is necessary to build a system under non-invasive and non-contact condition. Image process system suits to achieve such a request. Hence, it is recommended to judge the drowsiness state by observing eye status of operators via eye video. In this thesis, we use a CCD camera as the image sources and use skin color map to segment skin region. Then we use PCA algorithm to find the eye region for circular template searching. The circular template will locate the iris region and finally we can analyze this region to classify the eye openness state. By numerical simulation, we have obtained a high accuracy on eye openness detection and it would be helpful for the drowsy detection system.