Corona segmentation for nighttime brake light detection

Brake light detection of front cars has become a very important issue in safety of transport systems in recent years. As an adjunct component of automatic braking or warning systems, recognition and discrimination of the brake lights using vehicle‐mounted cameras provides early warning to avoid rear‐end collisions for the vehicles. Therefore in this paper a single camera‐based segmentation method is introduced for detecting the brake lights in nighttime cruising and discriminating them from the other lights, such as tail lights and turn lights. Basically, a novel system is put forward for discriminating brake lights which is initialised with capturing the frames of front car having the tail lights on, with a mounted camera. Subsequent to acquisition, image enhancement is applied to frames for whitening the red corona and darkening the rest including the centre of the light sources. Region of interests are determined using the cumulative contrast differences as well as rear light positions with calculation of white and black pixel ratios in coronas. Yet, the tail lights have the approximately same ratio for all distances, ratios of the brake lights are significantly high, resulting in discrimination of brake lights from others, for the vehicles cruising in the dark.