Abstract:
The Cyclists and Motorcyclists detection in streaming traffic video has been a challenging task due to irregular movement within the road and smaller image size within the frame. This paper is proposed a novel method of image segmentation of cyclists and motor cyclists and subsequent detection with image moments in traffic video footage. However, isolation of pedal cycles and motorcycles with perfect object boundaries has been a challenging problem with respect to other vehicle categories in the context of image segmentation. Irregular shape image segmentation for pedal cyclists and motor cyclists using a novel recursive image segmentation algorithm is proposed in this work The recursive image segmentation algorithm is applied to extract image pixels of a moving object in the binary image. The extraction of all the pixels of a bicycle could be accomplished successfully using the proposed algorithm. Subsequently, pixel count, height, width and the image Hu moments are recorded and used to identify the motorcycle category. An accuracy of 91.2% was obtained for video footage duration of 6 minutes video sequence for the detection of cyclists and motor cyclists. This recursive image segmentation method has successfully been applied in identification of motorcycles in traffic video sequences.
Citation:
M. Eeshwara, R. Thilakumara and N. Amarasingha, "Cyclists and Motorcyclists Detection in Traffic Video Footage," 2022 Moratuwa Engineering Research Conference (MERCon), 2022, pp. 1-5, doi: 10.1109/MERCon55799.2022.9906193.