교차로 I2V 환경에서 노변 LiDAR 기반 객체 로그와 RSA 이벤트 로그의 위험 유형별 정합성 분석: OEC 프레임워크

초록

This study proposes an Object-Event Consistency (OEC) framework for analyzing hazard-typespecific consistency between roadside LiDAR object logs and Road Safety Alert (RSA) event logs in an intersection I2V environment. OEC is an internal consistency ratio based on predefined time, class, and ROI conditions and does not replace ground-truth-based detection metrics. One hour of data from the Cheongsa intersection in Daejeon, Korea, comprising 12,847 object observations and 42 RSA events, was analyzed. At τ=20s, OEC was 100.0% for illegal parking, 73.9% for merging caution, and 50.0% for jaywalking pedestrian/PM. An exploratory test indicated a statistically significant association between hazard type and matching status (p=0.041; Cramér’s V=0.39). The OEC estimate for jaywalking pedestrian/PM was based on eight events and had a wide Wilson 95% confidence interval of 21.5– 78.5%. The hazard-type ranking remained stable across τ=10, 20, and 30s, providing exploratory evidence for hazard-specific validation and parameter review.

키워드