Search

License plate recognition is a significant technology to build a smart city transportation system. The current license plate recognition systems have achieved a good recognition and application effect for single-row license plates, nonetheless, these can’t satisfy the recognition need for license plates which contain stacked characters. Based on this problem, this paper proposed a method based on deep learning and independent of character segmentation to identify the license plates containing stacked characters. Firstly, the method corrected the tilted or distorted license plate, and used MobileNet-SSD algorithm to detect single-row characters and stacked characters in the license plate. Then, it sent stacked characters to the stacked character recognition network based on CTC loss and conducted the end-to-end identification without character segmentation. Result of the experiment indicates that this algorithm not only has high recognition accuracy for the license plate with stacked characters, but also adapts to the license plate under the complex environment such as tilt and distortion, greatly improves the generality of the license plate recognition system. [ABSTRACT FROM AUTHOR]