🤖 AI Summary
To address perspective distortion induced by arbitrary camera angles, difficulties in rectifying single- and double-line Chinese license plates, and computational constraints on edge devices, this paper proposes a lightweight end-to-end network jointly optimizing geometric rectification and recognition. The method integrates a differentiable Perspective Transformation Rectification (PTR) module directly into the recognition backbone, enabling weakly supervised geometric correction guided solely by recognition outputs. It further incorporates an enhanced AFLNet architecture, a strengthened channel-spatial attention mechanism, and a weighted Focal Loss to improve discrimination among visually similar Chinese characters (e.g., “川”/“川”, “京”/“津”). Evaluated on standard benchmarks, the model achieves state-of-the-art accuracy while processing each frame in under 10 ms on GPU hardware—demonstrating both high precision and efficiency for real-time deployment on resource-constrained platforms.
📝 Abstract
Chinese License Plate Recognition (CLPR) faces numerous challenges in unconstrained and complex environments, particularly due to perspective distortions caused by various shooting angles and the correction of single-line and double-line license plates. Considering the limited computational resources of edge devices, developing a low-complexity, end-to-end integrated network for both correction and recognition is essential for achieving real-time and efficient deployment. In this work, we propose a lightweight, unified network named LPTR-AFLNet for correcting and recognizing Chinese license plates, which combines a perspective transformation correction module (PTR) with an optimized license plate recognition network, AFLNet. The network leverages the recognition output as a weak supervisory signal to effectively guide the correction process, ensuring accurate perspective distortion correction. To enhance recognition accuracy, we introduce several improvements to LPRNet, including an improved attention module to reduce confusion among similar characters and the use of Focal Loss to address class imbalance during training. Experimental results demonstrate the exceptional performance of LPTR-AFLNet in rectifying perspective distortion and recognizing double-line license plate images, maintaining high recognition accuracy across various challenging scenarios. Moreover, on lower-mid-range GPUs platform, the method runs in less than 10 milliseconds, indicating its practical efficiency and broad applicability.