License plate detection method based on multi-task cascaded convolutional neural network

Abstract

The invention relates to a license plate detection method based on a multi-task cascaded convolutional neural network. Aiming at the problems of poor scalability and generalization ability of traditional license plate detection methods, and the decrease of license plate recognition rate caused by poor license plate detection results, a multi-task cascaded convolutional neural network license plate detection method was proposed. In order to obtain the precise location of the license plate in the complex image, the method builds a new multi-task cascaded convolutional neural network model and conducts large-scale training on the CCPD vehicle data, wherein the convolutional layer of the network model of the present invention adopts and BN layer Combined method to improve the feature extraction ability of license plate, speed up model training, use Relu activation function to increase the nonlinear ability of the model, use multi-task loss function to improve the network's classification of license plate and the accuracy of regression box, and regression feedback Four key points of the license plate information. Finally, the trained model is applied to license plate detection. The method is simple and flexible, and has strong practical applicability.

Description

technical field [0001] The invention relates to the fields of pattern recognition and computer vision, in particular to a license plate detection method based on a multi-task cascaded convolutional neural network. Background technique [0002] With the rapid development of public transportation systems, smart technologies are playing an increasingly important role in many applications. Among them, the most prominent ones are video surveillance, pattern recognition, image processing and automatic detection technology, which are also attracting more and more attention. Cars can be seen everywhere around us, and they have become the most important travel tool for people. As we all know, every car is equipped with a unique "identity" document, also known as the vehicle's license plate information. In efficient license plate management, automatic collection and automatic identification of a large number of license plate information is an extremely critical link. Therefore, the...

AI Technical Summary

Problems solved by technology

[0004] In view of this, the object of the present invention is to provide a license plate detection method based on a multi-task cascaded convolutional neural network to solve the problem of license plate detection and positioning in roads based on global eye video surveillance in complex scenarios

Images