A Hybrid Technique for Real Time License Plate Localization with the aid of FFBPNNAPSO

摘要

Vehicle License Plate Recognition (VLPR) is an imperative constituent in Intelligent Transportation Systems (ITS), which encircles three foremost phases essentially License Plate Localization (LPL), Character Segmentation (CS), Character Recognition (CR). In this paper, we have intended to introduce a novel License Plate Localization algorithm subjected to Artificial Neural Networks (ANN). This proposed scheme involves distinct phases of pre-processing, image de-noising and enhancement, feature extraction, Neural Network training and License Plate detection. Followed by the mining of assorted statistical features, geometrical features, edge features and texture features from the vehicular image, they are given as the input to Feed Forward Back Propagation Neural Network (FFBPNN) in order to localize the License Plate. During the training process, the parameters of the FFBPNN will be optimized using the eminent Adaptive Particle Swarm Optimization (APSO) algorithm in order to improve the Neural Network convergence performance. The License Plate Localization of our proposed technique is analyzed with simple Feed Forward Back propagation Neural Network (FFBPNN) in terms of accuracy, sensitivity and specificity. The experimental outcomes demonstrate that the proposed procedure proficiently accomplishes an extremely high localization rate with elevated specificity (91.3%).