Evolving GMMs for road-type classification

Mohammad, Mahmud Abdulla, Kaloskampis, Ioannis ORCID: https://orcid.org/0000-0002-4450-4935 and Hicks, Yulia ORCID: https://orcid.org/0000-0002-7179-4587 2015. Evolving GMMs for road-type classification. Presented at: 2015 IEEE International Conference on Industrial Technology (ICIT), Seville, Spain, 17-19 March 2015. Industrial Technology (ICIT), 2015 IEEE International Conference on. IEEE, pp. 1670-1673. 10.1109/ICIT.2015.7125337

Full text not available from this repository.

Abstract

In this paper, a new online vision-based road-type classification method is proposed. The method uses video captured by a single video camera and takes into account the visual information of the whole scene by segmenting the video frames into temporally consistent frame segments. To this end, we use a video segmentation algorithm based on evolving Gaussian mixture models (GMMs). Our method consists of two stages. In the first stage, we build a priori statistical models of different road types, one model per road type under consideration. For this purpose, we use GMMs produced by the video segmentation algorithm applied to the training video data offline. In the second stage, new video frames are segmented and classified into one of several possible road types on the basis of the Bhattacharyya distance between the Gaussians produced from the new video frame and the Gaussians from the a priori models representing the different road types. Experimental results on real-world data indicate that our method outperforms the state of the art method in this area in both classification accuracy per road type and overall classification accuracy.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Publication
Status:	Published
Schools:	Engineering
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
Publisher:	IEEE
Last Modified:	28 Oct 2022 09:19
URI:	https://orca.cardiff.ac.uk/id/eprint/74106

Citation Data

Cited 3 times in Scopus. View in Scopus. Powered By Scopus® Data

Actions (repository staff only)

Edit Item