Rethinking and Designing a High-performing Automatic License Plate Recognition Approach

TL;DR

This paper introduces VSNet, a novel real-time ALPR system that combines vertex detection and recognition with innovative neural network techniques, achieving high accuracy and speed, and demonstrating strong generalization on multiple datasets.

Contribution

The paper presents a new ALPR approach, VSNet, featuring a cascaded CNN framework with vertex information and weight-sharing classifiers, improving accuracy and efficiency over existing methods.

Findings

Achieves over 99% recognition accuracy on CCPD and AOLP datasets.

Outperforms state-of-the-art methods by more than 50% in error rate.

Runs at 149 FPS, demonstrating real-time performance.

Abstract

In this paper, we propose a real-time and accurate automatic license plate recognition (ALPR) approach. Our study illustrates the outstanding design of ALPR with four insights: (1) the resampling-based cascaded framework is beneficial to both speed and accuracy; (2) the highly efficient license plate recognition should abundant additional character segmentation and recurrent neural network (RNN), but adopt a plain convolutional neural network (CNN); (3) in the case of CNN, taking advantage of vertex information on license plates improves the recognition performance; and (4) the weight-sharing character classifier addresses the lack of training images in small-scale datasets. Based on these insights, we propose a novel ALPR approach, termed VSNet. Specifically, VSNet includes two CNNs, i.e., VertexNet for license plate detection and SCR-Net for license plate recognition, integrated in a resampling-based cascaded manner. In VertexNet, we propose an efficient integration block to extract the spatial features of license plates. With vertex supervisory information, we propose a vertex-estimation branch in VertexNet such that license plates can be rectified as the input images of SCR-Net. In SCR-Net, we introduce a horizontal encoding technique for left-to-right feature extraction and propose a weight-sharing classifier for character recognition. Experimental results show that the proposed VSNet outperforms state-of-the-art methods by more than 50% relative improvement on error rate, achieving > 99% recognition accuracy on CCPD and AOLP datasets with 149 FPS inference speed. Moreover, our method illustrates an outstanding generalization capability when evaluated on the unseen PKUData and CLPD datasets.