Vehicle tyre parameterisation using binary search thresholding and contour fitting

This paper describes a visual tyre location identification technique using a series of localised edge detection operations coupled with dynamically assigned binary search contrast adjustment to parameterise each of the tyres from a lateral viewpoint image of a vehicle undercarriage. The technique is applied to a data bank of 130 images with an 82% success rate in parameterising the near tyres and 80% success rate in parameterising the far tyres for a given image. The parameters derived from this technique can provide the basis for pose estimation algorithms to perform automatic parking and to aid mobile robots in path planning with respect to motor vehicles.