A Novel License Plate Detection Approach for an Embedded System

초록

This paper proposes a new license plate detection (LPD) algorithm that is suitable for operation in an embedded system where computational resources are limited. Unlike the conventional approaches where images are scanned multiple times to investigate the edge density, we propose a novel method where we utilize the parallel horizontal edge to reduce the search area. To reduce frequent memory accesses and the processing time, the proposed algorithm operates using strong horizontal edge position information. We evaluated the detection accuracy and processing speed of the proposed system in comparison with other conventional edge density– based methods. We also implemented our algorithm in a real embedded system with an ARM Cortex-A9 processor and 1 GB RAM. The average success rate was 93.12%, and the average processing speed was 45 ms per frame.

키워드