Moving object classification in a domestic environment using quadratic neural networks

We present a moving object recognition system. A description is given of the whole system from the image acquisition through the preprocessing and feature extraction stages to the classification of objects. We use quadratic neural networks (QNN) to model the input data and then extract features from the model which are translation and rotation invariant. We have applied the idea to a practical problem of classifying moving objects in a domestic environment such as moving heads, curtains blown by the wind and external events such as moving tree branches. Reasonable results are obtained using only the spatial information