View Invariant Object Recognition Using Coloured Receptive Fields

This paper describes a technique for the recognition and tracking of every day objects. The goal is to build a system in which ordinary desktop objects serve as physical icons in a vision based system for man-machine interaction. In such a system, the manipulation of objects replaces user commands. This method is based on sampling a local appearance function at discrete viewpoints by projecting it onto a vector of receptive elds which have been normalised to local scale and orientation. This paper reports on the experimental validation of the approach, and of its extension to the use of receptive elds based on colour. The experimental results indicate that the technique does indeed provide a method for building a fast and robust recognition technique. Furthermore, the extension to coloured receptive elds provides a greater degree of local discrimination. The coloured receptive eld approach is applied to the recognition of objects under changing view points. Appearance of objects depends strongly on the view point and the lighting. In the experiments we show that the developed technique based on coloured receptive elds allows the recognition of objects invariant from the view point of the camera. This is obtained by training images from view points that sample the view sphere. This experiment shows that the approach is suitable for the recognition of general objects as physical icons in an augmented reality.