Near-infrared LED system to recognize road surface conditions for autonomous vehicles
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 critical safety challenge of recognizing road surface conditions (dry, wet, snowy, and icy) for autonomous vehicles (AVs). Current detection systems often rely on laser diodes or halogen lamps, which suffer from high costs, temperature sensitivity, or inability to modulate signals to compensate for ambient light. Furthermore, existing near-infrared (NIR) backscattering techniques have not been validated for large incident angles, which are necessary for long detection ranges (e.g., >33 meters at 60 km/h). The authors propose a novel system using three near-infrared light-emitting diodes (LEDs) to replace laser diodes, leveraging LEDs’ lower cost, temperature stability, and ability to illuminate larger areas. The study investigates whether the broader spectral bandwidth of LEDs impairs classification performance and validates the system’s feasibility under large incident angles. The methodology involves determining optimal wavelengths for a tri-wavelength LED source by analyzing the spectral response of asphalt under five conditions: dry, wet, water-covered, icy, and snowy. Spectral measurements were conducted using a halogen lamp and an NIR spectrometer across various incident angles (76.5° to 86.5°). Unlike previous studies assuming monochromatic sources, this work models LED spectra as normal distributions with a full width at half maximum (FWHM) of 80 nm. A two-step classification algorithm was developed: Step 1 distinguishes {dry + wet}, ice, snow, and water using the ratio of signals at wavelengths $\lambda_1$ and $\lambda_2$; Step 2 distinguishes dry from wet using the ratio of signals at $\lambda_1$ and $\lambda_3$. The optimal wavelengths were calculated by minimizing the overlap between class distributions. The theoretical influence of LED spectral bandwidth on performance was also analyzed. Finally, the system was experimentally validated using LEDs at 970, 1450, and 1550 nm and an NIR camera, testing incident angles from 78.7° to 86.2°. The results demonstrate that the proposed LED system can effectively recognize road surface conditions. The calculated optimal wavelengths for the LED sources were determined to be 1457 nm, 900 nm, and 1664 nm, though the experimental validation used 970, 1450, and 1550 nm. The classification accuracy varied by condition: the system achieved 97% accuracy in distinguishing snow, wet, and water-covered surfaces. However, distinguishing dry from wet asphalt proved more challenging, with accuracies of 73% for dry and 68% for wet conditions. This limitation is attributed to the similar spectral shapes of dry and wet asphalt observed in the measurements. The study confirms that despite the broader spectral bandwidth of LEDs compared to laser diodes, the system maintains high performance for most hazardous conditions, particularly those involving significant water or ice presence. The significance of this work lies in providing a cost-effective and robust alternative to laser-based NIR systems for AVs. By validating the use of LEDs at large incident angles, the research supports the implementation of long-range road condition detection, which is essential for safe autonomous driving. The findings suggest that while LED-based systems are highly effective for detecting snow, ice, and standing water, further refinement may be needed to improve the discrimination between dry and slightly wet road surfaces. This approach enhances the reliability of advanced driver-assistance systems (ADAS) and AVs by enabling proactive trajectory planning and speed adjustment based on real-time road surface analysis.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-20 |
| archive | success | canonical_url | — | — | 1 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-25 |
| chunk | success | chunk | — | — | 1 | 2026-06-25 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-25 |
| promote | success | — | — | — | 1 | 2026-06-20 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-25 |
| verify | success | — | — | — | 1 | 2026-06-26 |
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.