Vehicle Recognition

We have worked on roadside vehicle recognition, with a view to embedding this functionality inside roadside ANPR (Automatic Number Plate Recognition) cameras.

Our approach has been to develop methods to detect and segment the vehicle from the rest of the scene using adaptive background segmentation. We then use support vector machine (SVM) classifiers which operate on both the colour of the vehicle to give a colour category classification and the size/shape of the vehicle to give a vehicle type classification, such as 'HGV', 'light van', or 'car'. Other work that we have developed aims to classify the make and model of the vehicle if it is a car.

In the video below (YouTube hosted), the image frame is divide into four quadrants. Top-left shows the raw video stream. Bottom left shows the segmentation. Yellow pixels are the foreground object (the car), green pixels are the car's shadow on the road. Bottom-right shows the car boundary, with shadows removed and top-right shows the segmented car pixels superimposed on top of the learned background and thus the shadow is removed.

Selected publications.

• Automatic Make and Model Recognition from Frontal Images of Cars
  G. Pearce and N. E. Pears
  Proc. 8th IEEE Int. Conf. on Advanced Video and Signal Based Surveillance, Klagenfurt, Austria, Aug 30th - Sept 2nd 2011.
  [PDF] [DOI][BibTeX]
  
  
• Background Subtraction in Video using Recursive Mixture Models, Spatio-temporal Filtering and Shadow Removal
  Z. Chen. N. E. Pears, M. Freeman and J. Austin
  Advances in Visual Computing : Lecture Notes in Computer Science (LNCS), vol 5876
  5th International Symposium on Visual Computing November 30 - December 2, 2009
  [PDF] [DOI][BibTeX]
  
  
• Road Vehicle Classification using Support vector Machines
  Z. Chen. N. E. Pears, M. Freeman and J. Austin
  Proc. IEEE Int. Conf. on Intelligent Computing and Intelligent Systems, pages 214 - 218, Shanghai China, November 2009.
  [PDF] [DOI][BibTeX]
  
  

BACK to Nick Pears' Research Projects page.