Automatic number plate recognition plays a vital role in an intelligent transportation system. The intelligent transportation system enforced the government's traffic control law. The enforcement of automatic number plate recognition reduces the traffic load and accident recovery on the roadside. Several emerging machine learning algorithms increase the accuracy of number plate recognition, but realistic problems still exist. The performance of accuracy depends on several factors, like feature extraction, character segmentation, and training of characters for recognition. In this paper, we study several algorithms of advanced machine learning, such as support vector machines, deep neural networks, convolutional neural networks, and recurrent neural networks. The support vector machine overcomes the limitation of overfitting problems in classification algorithms. The other three algorithms, DNN, CNN, and RNN, are categories of deep learning algorithms that boost the performance of classification of number plate characters. The classified characters increase the automatic number plate recognition. The deep neural network is a multiple hidden layer-based network that improves the rate of recognition. the counter-race between CNN and RNN algorithms in concern for accuracy in recognition. For the validation of algorithms, three standard datasets were employed, such as the FZU cars dataset, the Stanford cars dataset, and the HUMAin2019 dataset. For the testing, we created five sets of car image groups, numbered 100 to 500. The performance of algorithms is estimated as accuracy, precision, and recall. The analysis of the results suggests that CNN's algorithms are better than other existing algorithms for number plate recognition.

NPR, Deep Learning, Machine Learning, image processing, DNN