Enhanced danger detection of headlight through vision estimation and vector magnitude

Abstract

Dangerous glare detection during night driving is able to reduce car accidents because the driver has time to prepare for upcoming situations. Even though increasing driving concentration might be a good solution for this issue, wrong detection would still practically be an essential factor decreasing driver's visual performance. Furthermore, loss of driving vision could cause car accidents. This research intends to reduce wrong detection in means of false positive and negative using methods based on an embedded system. Two techniques including HSV and the magnitude of vector are implemented to find the warning threshold of dangerous headlight glare with the least possible false positive and negative detection.