License Plate Detection using NMF with Sparseness constraints through still Images

Real time license plate detection and recognition from CCTV videos is an active research problem. Most of the existing solutions are reasonably successful and efficiently fast but most of them are effective only in controlled environments where light intensity, illumination, orientation of plate do not vary much and image resolution is not too low. Two crucial image processing steps in an LPDR system are: (a) localization of license plates within an image and (b) recognition of license plate using an OCR system. The aim of this paper is to address the localization problem for low quality images. We use a novel, and robust framework to build a license plate detection system. Implemented system is intelligent enough to tackle varying environment conditions automatically with low hardware requirements and less complex algorithm. Experimental results on database collected under varying conditions demonstrate the robustness of the proposed approach.