利用多亮度紅外線打光器之夜間車牌偵測

Abstract

本論文的主要目的，是利用週期性改變亮度之紅外線打光器，拍攝不同亮度之車牌影像，以提升夜間遠距離車牌影像的清晰度，以及減少近距離車牌影像的過曝情形。針對週期性光源變化之車牌影像分析，我們提出利用車牌字元筆觸寬度，偵測不同亮度影像中車牌位置。由於不同亮度影像偵測結果不同，且各亮度影像掌握不同資訊，因此挑選各亮度影像中最佳之車牌偵測結果與整合各亮度之影像，以合成一張高動態範圍影像，達成偵測環境中方便觀察不同深度、角度車牌之目的。

The goal of this thesis is to enhance the visibility of the images of license plates at far distances, and to reduce the over-exposure defects for similar images taken near a camera with an illuminator with fixed intensity, by applying an innovative, multi-intensity IR illuminator for night-time video surveillance. Based on an analysis of the stroke width of characters of the license ID, license plates at various distances can be detected under different illumination intensities. Since images obtained with different illumination conditions may contain useful information for different regions interest, best images of license plates are identified for each illumination condition and integrated into a high dynamic range (HDR) representation for easy inspection and examination. Experimental results show that the proposed approach can increase the monitored area both in depth and in width, and enhance the security level of night-time surveillance.