Abstract

The "Point Distribution Model", derived by analysing the modes of variation of a set of training examples, can be a useful tool in machine vision. One of the drawbacks of this approach to date is that the training data is acquired with human intervention where fixed points must be selected by eye from example images. This is a laborious process and may lead to a non-representative set of training examples being used. A method is described for generating a similar flexible shape model automatically from real image data. A cubic B-spline is used as the shape vector for training the model. Large training sets are used to generate a robust model of the human profile. The resulting modes of variation show the potential of the model for labelling and tracking of pedestrians in realworld scenes. Furthermore, an extended model is described which incorporates the direction of motion of the human, allowing the extrapolation of direction from shape. 1 Introduction We wish to generate a 2D flexib...

Citations