Research of license plate character features extraction and recognition

Abstract

A license plate character cascading recognition algorithm based on structure and gray pixel features is presented in this paper. First, it extracts character's stroke features, inflection point features, contour accumulation features from the preprocessed license plate characters; and then, extracts the pixel features from the character's gray images with PCA algorithm; finally, uses support vector machine (SVM) to respectively construct structure feature classifiers and gray pixel feature classifiers and recognize characters, at the same time, takes the results with higher confidence level as the first classification results from the both classifications. If the results are "8", "B" and the other easily confused characters, then reclassifies characters using the gray pixel classifier and yields the final results; if they are "0","D"and the other characters with similar structure features, and then uses the structure classifiers to classify again.