Real-time implementation and design of an embedded system for identifying speed limit sign

In this paper, we present an embedded system for identifying speed limit signs in real-time. The computation power of processors in embedded system is limited, so the reduction of computing for an embedded system is still an issue, especially for the real-time image processing. We proposed an efficient algorithm for identifying speed limit sign. According to the characteristics of speed limit signs, the proposed algorithm applies color clustering, dynamic thresholds and related digital image processing algorithms to propose three stages of clustering colors, searching speed limit signs and accessing the content of speed limit signs, which accesses the speed limit signs efficiently. Implementation results are conducted in different conditions, including sunny, cloudy and rainy days, which demonstrate that most speed limit signs and supplementary signs can be correctly detected on the average of 96.38%, and the average processing frame rate was 52.49fps on a 1.7 GHz dual-core ARM Cortex-A15 embedded system.