Abstract

Aim: Vertical edge extraction plays a very important role in license plate processing. In complex scenarios, unwanted vertical edges can increase complexity and trigger false detections. Vertical character edge detection highlights the specific vertical edges belonging to characters, which is essential for accurate license plate localization. In this paper, a lightweight Character Edge Detection Algorithm based on binary pixel differentiation is proposed.
Methods: Following binarization, morphological operations are used to extract image boundaries, highlight character-specific details through strategic intersection. Subsequently, resulting candidate regions are then validated to extract license plate area.
Results: Experimental results demonstrate that the proposed model minimizes the trade-off between localization accuracy and algorithmic complexity. The method achieved approx. 2x faster edge extraction than Sobel (82ms vs 171ms) with lower asymptotic growth and improved robustness in night-time captures, yielding localization accuracy of 85.44% and complexity to the power of 2.
Conclusion: The model achieves a significant reduction in processing latency compared to the traditional Sobel operator. While accuracy might be lower than some heavy models, the speed gains are the primary contribution to real-time applications on resource-constrained hardware.

Key words: license plate detection, character edge, vertical edge, resource efficient, embedded system.