FLANN Based Matching with SIFT Descriptors for Drowsy Features Extraction

Abstract

This paper presents an approach for extraction of drowsy features from face. Drowsiness during driving is one of the major issues of road accident. Driver drowsiness can happen due to fatigue resulting from physical or mental exertion, sedating effects of several medications, drug consumption, melancholy, or may be due to some disorders like obstructive sleep apnea. Drivers who do long distance driving mainly in tedious routes which do not require much driving input are also at a risk of getting sleepy. It is an indispensable task to detect the level of sleepiness in a person by monitoring various factors in intelligent vehicles. The current work combines Fast Library for Approximate Nearest Neighbours(FLANN) feature matching with Scale Invariant Feature Transform(SIFT) descriptors. SIFT has been widely used in face recognition and object detection tasks. SIFT algorithm is considered to be the most impervious to image deformations. The FLANN matcher matches the descriptors of features in a set with the features in the target set. The results show the superiority of FLANN-SIFT when compared with SIFT for drowsy driver detection.