Optimization of Two-Phase Sampling Designs With Application to Naturalistic Driving Studies

• Published in: IEEE transactions on intelligent transportation systems Vol. 23; no. 4; pp. 3575 - 3588
• Main Authors: Imberg, Henrik, Lisovskaja, Vera, Selpi, Nerman, Olle
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.04.2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Amounts of traffics; Annotations; Automatic measurements; Automobiles; Brakes; Case-control studies; Driving behavior; Instance selection; Maximum likelihood estimation; naturalistic driving studies; optimal design; Optimization; pseudo-likelihood; Random sampling; Safety; Safety critical; safety critical event; Safety engineering; Sampling; Sampling designs; Segments; Simple random sampling; Traffic information; Traffic safety; Transport Systems and Logistics; Transportteknik och logistik; Two-phase samplings; unequal probability sampling; Vehicles; Video sequences
• ISSN: 1524-9050; 1558-0016; 1558-0016
• DOI: 10.1109/TITS.2020.3038180

Naturalistic driving studies (NDS) generate tremendous amounts of traffic data and constitute an important component of modern traffic safety research. However, analysis of the entire NDS database is rarely feasible, as it often requires expensive and time-consuming annotations of video sequences. We describe how automatic measurements, readily available in an NDS database, may be utilized for selection of time segments for annotation that are most informative with regards to detection of potential associations between driving behavior and a consecutive safety critical event. The methodology is illustrated and evaluated on data from a large naturalistic driving study, showing that the use of optimized instance selection may reduce the number of segments that need to be annotated by as much as 50%, compared to simple random sampling.