License plate detection and localization in complex scenes based on deep learning

In order to solve the problem of license plate detection and localization in complex scenes, a new algorithm based on Deep Learning is proposed. Applied the model of convolutional neural network(CNN), Recognized the license plate by converted object detection problem into a binary classification problem. The candidate regions are generated on the sliding window by using selective search algorithm, and the IoU overlap of license plate boundary box and the overlap degree of the candidate regions are used to evaluate the positive and negative samples, at last Support Vector Machine is used for classification. Compared with license plate detection in traditional machine learning, it has obvious advantages in detection accuracy, especially for complex scenes, the positioning accuracy has been improved significantly.