THE DETECTION METHOD OF ABNORMAL VEHICLE BEHAVIOUR USING POINT SEQUENCE VEHICLE TRAJECTORY DATA

Abstract

Probe data is attracting attention to detect vehicles with abnormal behaviour caused by disasters or traffic accidents. This is because probe data can collect data over a wider area and narrower streets than conventional installed sensors. This study uses commercial vehicle probe data in dot format, which does not require map-matching. A method is proposed for detecting vehicles whose behaviour differs from daily life in terms of similarity calculated by converting the trajectory vehicle dot data into a string. As an example of a situation where abnormal behaviour is likely to occur, the proposed method was applied to actual expressway road closures and lane restrictions. The analysis results showed that similarity based on the Levenshtein distance index of the fifth mesh detected abnormal behaviour, including detours and speed changes with high accuracy.