A Gaussian mixture model and support vector machine approach to vehicle type and colour classification. Issue 2 (1st March 2014)

• Record Type: Journal Article
• Authors: Chen, Zezhi; Pears, Nick; Freeman, Michael; Austin, Jim
• Is Part Of: IET intelligent transport systems. Volume 8:Issue 2(2014)
• Page Start: 135
• Page End: 144
• Publication Date: 2014-03-01
• Subjects: cameras -- closed circuit television -- Gaussian processes -- image classification -- image colour analysis -- image motion analysis -- image segmentation -- object detection -- support vector machines -- traffic engineering computing -- transforms -- video signal processing; Gaussian mixture model‐support vector machine approach -- vehicle type -- colour classification -- moving road vehicle segmentation -- colour video data -- stationary roadside CCTV camera -- pragmatic vehicle classification applications -- multidimensional smoothing transform -- camera vibration -- shadow detection approach -- background subtraction -- kernelised support vector machine; Intelligent transportation systems -- Periodicals; Electronics in transportation -- Periodicals; 388.31205
• DOI: 10.1049/iet-its.2012.0104
• Languages: English
• ISSNs: 1751-956X

Abstract

Abstract : The authors describe their approach to segmenting moving road vehicles from the colour video data supplied by a stationary roadside closed‐circuit television (CCTV) camera and classifying those vehicles in terms of type (car, van and heavy goods vehicle) and dominant colour. For the segmentation, the authors use a recursively updated Gaussian mixture model approach, with a multi‐dimensional smoothing transform. The authors show that this transform improves the segmentation performance, particularly in adverse imaging conditions, such as when there is camera vibration. The authors then present a comprehensive comparative evaluation of shadow detection approaches, which is an essential component of background subtraction in outdoor scenes. For vehicle classification, a practical and systematic approach using a kernelised support vector machine is developed. The good recognition rates achieved in the authors' experiments indicate that their approach is well suited for pragmatic vehicle classification applications.