Abstract

This paper presents an object recognition algorithm which extracts the moving objects out of its background region from a dynamic image. Shape feature and color feature are employed for objects characterising, and neuron network is used for their classifying. Initially, the neuron network is trained with a priori defined patterns for its recognition. Then, if the system finds any modification in object shape, the neural network will automatically learn these features. This will enable the system to adapt to modification of the object shapes while maitaining its recognising capability