PURPOSES : This paper is aimed at suggesting a novel approach for determining the pavement condition rating based on the tire-surface friction noise using a machine learning algorithm as a low-end pavement condition monitoring system.
METHODS : Vehicle on-board type noise measurement system according to the ISO11819-2, and the K-nearest neighbors with dynamic time warping algorithm were applied. The system and algorithm were empirically tested with a field study.
RESULTS : The developed AI- and noise-based pavement condition monitoring system demonstrated significantly positive results with a precision 90.8%, recall 84.8%, and f1-score 86.1%.
CONCLUSIONS: We herein confirmed that the acoustic property between the tire and road surface can be used for monitoring pavement conditions. It is believed this finding presented a new paradigm for monitoring pavement conditions based on visual information. However, extensive studies focused on the practical application of this method are required.

ABSTRACT
1. 서론
2. 문헌고찰
3. 타이어/노면 마찰소음 측정 및 분석시스템구축
    3.1. 타이어/노면 마찰소음 측정시스템 구축
    3.2. 포장상태등급 판단 인공지능 구축
4. 포장상태등급 판정 인공지능 네트워크 구축
    4.1. 포장상태등급체계 및 데이터 측정/가공
    4.2. 인공지능 학습과 정확도 테스트
5. 결론
REFERENCES

• 한대석(한국건설기술연구원) | Daeseok HAN (Korea Institute of Civil Engineering and Building Technology) 교신저자