On the classification of moving objects in image sequences using 3D adaptive recursive tracking filters and neural networks

Abstract

It is shown that 3D recursive filters may be used to classify the motion of objects in discrete-time spatiotemporal 3D image sequences. The 3D filter has a time-varying 3D frequency-planar passband that is adapted in a feedback system to automatically track a moving object on the basis of its smoothly changing trajectory, thereby rejecting noise and stopband objects that are not of interest. The adaptive spacetime velocity vector of the passband object is available within this feedback system and is used as the input to a multi-layer perceptron neural network which classifies the motion of the passband object according to a number of motion characteristics, such as its direction of travel, velocity, acceleration as a function of time and position and its stopping time. It is shown that such a system may be used to classify the motion of vehicles at an intersection of roads.