Misshaped boundary classifier model for license plate detection in haze weather using entropy CNN

Abstract

Weather that creates haze can cover up car license plates, creating warped lines that make it difficult to see and identify them. This paper suggests a novel Primitive Boundary Classifier Model (PBCM) that uses the unique properties of bright and dark boundaries to solve this problem. Iteratively extracting characteristics from the input image, the PBCM draws volatile borders and ends linearity at particular pixel positions. To detect irregular boundaries in the hidden layers through changes in entropy and regularity terminations, this procedure is combined with linear entropy learning, which is accomplished by altering a convolutional neural network. Identifying the license plate area and its related embedding is possible by finding these terminating border pixel locations. The model evaluates its performance during validation by considering similarity and false rate metrics. The comparative analysis, this model improves the 7.34% detection precision with 15.98% high similarity and 8.95% less false rate for the maximum epochs performance ratio of 90.1% and error rate of 11.2%.