Recognizing vehicle license plate in the presence of partial occlusion

Abstract

Vehicle License Plate Recognition (VLPR) plays an important role newlinein Intelligent Transportation System. Recognition of Vehicle License Plate newline(VLP) becomes difficult in the presence of the partial occlusion such as newlinesunlight shadow, mist, tilted view and rain streaks. The efficiency in terms of newlinerecognition accuracy of VLPR depends on the robustness of License Plate newlineRecognition (LPR) process. This thesis proposes efficient Partial Occlusion newlineRemoval (POR) schemes to enhance the recognition accuracy of VLP. newline The occurrence of shadow over LP hides the alphanumeric newlinecharacters, which results in inaccurate character recognition. Thus, removal of newlineshadow from the LP is an important process. In existing shadow removal newlinemethods, such as Conditional Random Field, Bi-dimensional Empirical Mode newlineDecomposition etc., the pixel classification between shadow and non-shadow newlineregion is very difficult. These methods eliminate primitive shadow noise only. newlineExisting methods do not address the real artifact shadow problem due to newlinechange in shadow intensity at different daylight timing. This research newlineproposes the scheme named Suppression of Shadow in Partially Occluded LP newline(SSPOLP) , which suppresses the shadow noise using cepstrum approach newlinebased on gamma correction method i.e. Grassman axiom integral method for newlineshadow removal followed by LP detection using region props method. The newlineoverall recognition success rate for SSPOLP scheme is found to be 93.7% for newlineonline MNIST and MIT datasets. newline