不均勻照度環境的駕駛昏睡偵測與警示; Driver Drowsiness Detection and Warning under Various Illumination Conditions

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/9581

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/9581

Title:	不均勻照度環境的駕駛昏睡偵測與警示;Driver Drowsiness Detection and Warning under Various Illumination Conditions
Authors:	曹力仁;Li-Jen Tsao
Contributors:	資訊工程研究所
Keywords:	眼睛偵測;支援向量機;眼睛閉合偵測;昏睡判定;eye detection;support vector machine;eye open/closed detection;drowsiness discrimination
Date:	2008-06-26
Issue Date:	2009-09-22 11:51:02 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	昏睡與疲勞會降低對外界的反應、注意力，與警覺度。為了保障駕駛人和行人的安全，我們提出了一個駕駛昏睡偵測和警示系統；我們架設一部紅外線相機於駕駛座前方擷取駕駛人的臉部影像，分析駕駛人是否昏睡並提出警示。此系統包含了主動式取像設備、眼睛偵測、眼睛追蹤、眼睛閉合與視線偵測，及昏睡判定與警示。為了使系統能也夠在夜晚或是亮度不足的環境下運作，我們使用一組紅外線相機配合紅外線打光器來取得駕駛人的臉部影像。針對亮度不均勻的影像，我們根據影像的邊強度將影像劃分為多個區域，再對影像各區域各別二值化，經由幾何限制條件擷取可能的眼睛區塊，再利用支援向量機 (Support Vector Machine, SVM) 辨識是否為眼睛，最後再進行雙眼的驗證。根據雙眼的位置估計臉部範圍。當連續三張影像偵測成功，進入追蹤模式。在追蹤模式，我們在預測的區域內做三階段的眼睛偵測。在眼睛閉合偵測部份，我們測試了兩種方法並比較其準確性。在視線方向偵測，我們根據瞳孔的位置來判斷視線的左右方向。在昏睡判定方面，我們根據單位時間內閉眼張數所佔的百分比來判斷駕駛人是否陷入昏睡。我們在不同照度環境下測試我們的系統。實驗數據提示，眼睛偵測的偵測率為88.6%，誤判率為1.16%；眼睛閉合偵測的正確率為93.5%，誤判率為6.39%；平均偵測時間為0.057045秒，每秒約可處理18張影像。我們的方法能夠正確的偵測到駕駛人的眼睛和閉合狀態，並在偵測到駕駛昏睡的0.9秒內發出警示 To insure the safety of the driver and pedestrians on the road, we propose a vision based driver drowsiness detection and warning system. We use a camera mounted on the vehicle to capture the driver’s face images for eye detection and drowsiness discrimination. The system consists of five parts: Image acquisition system, eye detection, eye tracking, eye open/close and gaze direction estimation, and drowsiness discrimination. In the image acquisition system, we use an IR camera with two illuminators to capture driver’s face images in poor illuminated conditions. In order to deal with the uneven illumination, we propose a local thresholding method to divide the image into several partitions based on the strong edges then iteratively threshold each partition. We use connected-component and support vector machine (SVM) to verify eyes. If there are fixed numbers of frames succeeded in detection mode, we alternate the processing to tracking mode. In tracking mode, we detect eyes in the predicted region. We extract eye open/closed statuses and gaze directions information as our visual cues. In eye open/closed statues determination, we consider two criteria and compare their performance. In gaze direction estimation, we divide the eye region into three equal-sized subregions, then determine the pupil location in which subregion for the estimation. In drowsiness discrimination, we use PERCLOS measurement to judge whether the driver is drowsy. We test our system on our experimental car in various illumination conditions such as sunny day, cloudy day, at night, uneven illuminated conditions, with/without glasses. From the experimental results, we find that the proposed approach can stably detect the eyes and give a warning if drowsiness is detected.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...