Machine Vision Based Traffic Sign Detection Methods: Review, Analyses and Perspectives

Liu, Chunsheng; Li, Shuang; Chang, Faliang; Wang, Yinhai · 2019 · OpenAlex-citations

DOI: 10.1109/access.2019.2924947

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 presents a comprehensive review and analysis of machine vision-based traffic sign detection (TSD) methods, a critical component of Advanced Driver-Assistance Systems (ADAS) and autonomous driving systems. The authors address the challenges inherent in TSD, such as varying sign types, small sizes, complex driving scenes, and occlusions. Motivated by the limitations of previous surveys—which often lacked comprehensive comparisons, excluded LIDAR-based methods, or relied on outdated literature—this study aims to categorize, compare, and provide perspectives on current TSD techniques. The authors classify reviewed detection methods into five primary categories: color-based, shape-based, color-and-shape-based, machine-learning-based, and LIDAR-based methods. To facilitate rigorous comparison, the authors reimplemented several representative methods from the literature, particularly those lacking evaluations on public datasets. These reimplemented methods were tested on standard benchmarks, including the German Traffic Sign Detection Benchmark (GTSDB), BelgiumTS, TT100K, LISA, and others. The review covers specific algorithmic approaches within each category, such as RGB and HSV thresholding for color-based detection, Hough transforms for shape detection, and neural networks, Support Vector Machines (SVM), and AdaBoost for machine-learning-based detection. Key findings from the experimental comparisons highlight the performance trade-offs of different approaches. For instance, in the analysis of color-based methods on the GTSDB dataset, the Hue-Saturation-Intensity (HSI) thresholding method achieved high detection rates (97.14% for blue and 97.23% for red) but suffered from high extraction rates (over 30%), indicating excessive background inclusion. Conversely, the Ohta-based method demonstrated a lower extraction rate (3.50%) with a high detection rate (95.56%) for red signs. The paper notes that machine learning methods, particularly deep learning, have become mainstream due to their state-of-the-art performance, while LIDAR-based methods offer distinct advantages in handling point cloud data. The significance of this work lies in its structured taxonomy and empirical comparison of TSD methods, filling gaps left by prior surveys. By providing detailed analyses of detection mechanisms and performance metrics, the paper offers clear recommendations for future research directions. It emphasizes the need for robust methods that can handle diverse environmental conditions and sign variations, thereby promoting the development of more reliable TSR systems for safety-critical applications like autonomous driving and traffic infrastructure maintenance.

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