RETRACTED: Research on vehicle license plate recognition technology based on deep convolutional neural networks

Abstract

Convolutional Neural Networks (CNN) are used for visual type recognition. This is a first-order learning method. Types of Deep Neural Networks (DNNs) training to improve vehicle identification algorithms routinely. Symptoms are difficult to diagnose accurately. Recognition is done approximately in harsh environments. The recognition accuracy of Convolutional Neural Networks (CNN) is relatively high, which requires a large number of samples. A Vehicle Logo is the most explicit vehicle manufacturer indicator and is based on the Vehicle Logo Recognition (VMR) system. Mark recognition can still be a challenge because it is difficult to pinpoint the vehicle's identity in the image and seek stability in different imaging situations simultaneously. Vehicle Manufacturer Recognition (VMR) is proposed to eliminate the requirements for accurate identification and analysis of a Convolutional Neural Network (CNN) system. The lining is used based on information obtained from a CNN source. To train the neural network use the question expansion strategy to increase the training set of the synthetic transformation and to increase the rate at which the recall. This method can be maintained with high accuracy under light changes and noise conditions and is not adaptable to the environment. Experimental results show that data sets, which show that the algorithm is capable of strong generalizations, do not have high classification accuracy. The class's average accuracy will be higher than other methods; to show that CNN is better at recognizing vehicle identification with the results of comparative tests.