A Lightweight IMU-Only Pipeline for Road Hazard Detection
Bildiriler > Tam Metin Bildiri
Yazarlar: Hasan BULUT, Huseyin Karacali, Cagla Karahan
2025 9th International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT), , , Uluslararası, Tam Metin Bildiri, ss 1-6
DOI: 10.1109/ismsit67332.2025.11268251
Özet
Road-surface anomalies such as bumps and potholes pose a safety and comfort risk, especially for two-wheeled and micromobility users. Although visual perception provides a wealth of information, it is resource-intensive and inconsistent under adverse conditions. Inertial measurement units (IMUs), on the other hand, are inexpensive, ubiquitous and weather-independent, making them attractive for wide-area, permanent monitoring. This paper presents a lightweight, modular pipeline for IMU-only hazard detection consisting of candidate segmentation, feature extraction, rule-based classification and standardized evaluation. Experiments confirm the approach: the hazard detection achieved a precision of 0.91, a recall of 0.71 and an F1 score of 0.80, with a high specificity for normal driving (F1 = 0.91). While bumps and potholes were consistently detected, the classification between them reached 50% accuracy, which is due to overlapping signatures. The results provide a transparent basis for IMU-based hazard detection and highlight future opportunities for hybrid rule/learning approaches to improve fine-grained classification.