駕駛者臉部定位

Locating Driver’s Face During Driving

何礎安(Chu-An Ho)

指導教授 ： 陳世旺

國立臺灣師範大學/理學院/資訊工程研究所/碩士(2011年)

摘要

交通事故死亡率在國內死亡率排名總是居高不下，其中肇事的主因多來自於駕駛者精神狀態不好所造成，因此有一部分的視覺式駕駛監控系統中，嘗試利用攝影機拍攝駕駛者的臉部狀態，利用臉部特徵的擷取，來進行其精神狀態的分析。為了能使這樣的系統在不同光照環境中也能穩定的運作，本論文主要在研究視覺式駕駛監控系統中，以影像補償的方式使影像回覆影像原始色彩，使系統在不受光源的影響下亦可快速即時的進行人臉定位。
本研究使用一般攝影機進行拍攝，本研究首先會對攝影機拍攝的影像序列選取參考影像，其目的是為了提供給稍後影像補償的動作使用，在參考影像的選取上，利用了五種特徵：邊緣的空間分佈(Compactness of Spatial Distribution of Edges)、色調統計(Hue Count)、膚色統計(Skin Color)、對比度(Contrast)和模糊程度(Blur)，來進行參考影像的選取。接著對影像序列進行分鏡偵測，藉由兩張前後相鄰影像間相關的趨勢，判斷場景是否發生變化，若產生變化。則進入的影像將利用K-L transform的方式將需補償影像之色彩分佈，轉換至參考影像的色彩分佈。最後利用Adaboost的方式進行人臉偵測和以粒子群最佳化為基礎的粒子濾波器(Particle swarm optimization- based particle filter)進行追蹤，並將偵測和追蹤結合，以追蹤輔助偵測、偵測確認追蹤的方式來輸出人臉定位的結果。

關鍵字

駕駛人臉定位 ； 影像評估 ； 分鏡偵測 ； 影像補償 ； 人臉偵測 ； 人臉追蹤

並列摘要

Facial expressions convey rich inward feelings, including both psychological (e.g., cheer, anger, delight, frustration, disgust, fear, and surprise) and physiological (e.g., vitality, fatigue, drowsiness, attention, and distraction) reactions. Humans can easily identify the inward reactions based on facial expressions. A system that can sense the inward feelings of a driver will be of great help for driving safety. To this end, the driver’s face should first be located. In this paper, we focus on a vision-based detection and tracking of the driver’s face in the input video sequence while driving.
The major difficulty with the above task is illumination variations resulting from sunshine, shadows, environmental lights, underground passages, overheads, and tunnels. To deal with this difficulty, we develop a process that consists of three steps: reference image selection, illumination variation detection, and lighting compensation. The process keeps eye on the input video sequence in order to maintain to some extent its image quality. The driver’s face is then detected using the Adaboost technique and is tracked using the particle swarm optimization method applied to the resultant video sequence. The proposed technique was shown to work well in a number of experimental video sequences with different conditions of illumination, driver, gender, and wearing. A high face location rate around 98% has been achieved.

並列關鍵字

Driver’s face localization ； Reference image selection ； Illumination variation detection ； Lighting compensation ； Adaboost face detection ； Particle swarm optimization tracking

參考文獻

[Lan71] Edwin H. Land, John J. McCANN, “Lightness and Retinex Theory.” Journal OF THE OPTICAL SOCIETY OF AMERICA, Vol. 61. No. 1, 1-11, January 1971.

[Hor04] W. B. Horng, C. Y. Chen , Y. Chang and C. Hai Fan, “Driver Fatigue Detection Based on Eye Tracking and Dynamic, Template Matching,” Proc. of IEEE Int’l Conf. on Networking, Sensing and Control, vol. 1, pp. 7 -12, Mar. 2004.

[Can02] J. L. Cantero, M. Atienza and R. M. Salas, “Human Alpha Oscillations in Wakefulness, Drowsiness Period, and REM Sleep: Different Electroencephalographic Phenomena within the Alpha Band,” Neurophysiol Clin 32:54–71, 2002.

[Gaf03] Philipp P. Caffier, Udo Erdmann and Peter Ullsperger, ”Experimental evaluation of eye-blink parameters as a drowsiness measure,” EUROPEAN JOURNAL OF APPLIED PHYSIOLOGY Volume 89, Numbers 3-4, 319-325. 2003.

[Ji04] Q. Ji, Z. Zhu, P. Lan, “Real-time Nonintrusive Monitoring and Prediction of Driver Fatigue,” Trans. of IEEE on Vehicular Technology, vol. 53, no. 4, pp. 1052 -1068, Jul. 2004.