Automatic understanding of road signs in vehicular active night vision system

This paper proposes a supplemental mechanism to active vehicular night vision systems, which automatically identifies and reads road signs through image processing. This may add important driving aid in difficult night situations where signs can be missed or when the language is not clear to the driver. Such a solution poses a challenge, as the night vision systems produce low-resolution images with intensity flaws. To examine the validity of the proposed sign and character recognition method, we examined available samples from three classes of road sign information: English letters, Hebrew letters and traffic symbols. The obtained feature separation results show the potential of full implementation in vehicular night vision systems.