Mobile system for road sign detection and recognition with template matching

Michał, Maćkowski; Michał, Sawiski; Wojciech, Walczyszyn · 2019 · DOAJ

DOI: 10.1051/matecconf/201925203014

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of implementing a real-time road sign detection and recognition system on mobile devices, specifically Android smartphones. The motivation stems from the high cost and limited availability of advanced driver assistance systems in older vehicles, contrasted with the widespread ownership of mobile phones equipped with cameras and powerful processors. The authors aim to determine if mobile devices are robust enough to perform automated sign recognition in real-time, thereby enhancing driving safety without requiring expensive external hardware. The proposed system utilizes the Open Source Computer Vision Library (OpenCV) and consists of two primary modules: detection and recognition. The detection module involves image acquisition, pre-processing, and region-of-interest identification. Image pre-processing employs a modified histogram equalization technique in the HSV color model, which excludes the brightest pixels to improve contrast in poorly lit conditions. Sign detection relies on color thresholding for yellow, red, and blue signs, followed by the Canny edge detector and Hough transform to identify geometric shapes (triangles, rectangles, circles, and octagons). The recognition module uses template matching, comparing detected signs against a database of 152 binary pattern images categorized by shape. The system was tested using nearly 15,000 image frames from a standard mobile phone camera, with an optimal frame size of 840x630 pixels. The results indicate that the modified histogram equalization significantly improves detection performance, increasing the average number of detected signs per frame from 6.39 to 50.83. This enhancement leads to a higher overall recognition accuracy of 88.76% for equalized images, compared to 79.64% for non-equalized images, yielding a total average correctness of 83.59%. However, this increased accuracy comes at the cost of processing time; the execution time for images requiring histogram equalization was nearly five times higher than for those that did not. The system operates at an average speed of 4 frames per second, which the authors deem sufficient for vehicles traveling up to 130 km/h in built-up areas and 260 km/h on highways. Detection efficiency varied by shape, with octagonal "STOP" signs showing faultless recognition, while triangles and rectangles suffered from false positives due to excessive line detection. The study concludes that while the algorithm achieves satisfactory recognition rates and meets real-time requirements for typical driving speeds, it faces significant practical limitations regarding power consumption. The continuous processing of image data causes rapid battery depletion, limiting operational time to 2–3 hours, and leads to device overheating. Consequently, while the technical feasibility of mobile-based sign recognition is demonstrated, further optimization is required to address energy efficiency and thermal management for practical deployment.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• sign visibility legibility