Image Segmentation-Based Bicycle Riding Side Identification Method

This paper attempts to identify which side of the road a bicycle is currently riding on using common camera for realizing an advanced navigation system and safety support system. To roadway area, proposed method performs semantic segmentation front image captured by drive recorder or smartphone. If area extends from center right, bicyclist left (i.e., correct position in Japan). In contrast, if left, right incorrect We evaluated accuracy various widths with different traffic volumes video bicycles Tsuruoka City, Yamagata Prefecture, Saitama Japan. High (>80%) was achieved any combination model, identification method, experimental conditions. Given these results, we believe that have realized effective segmentation-based on.