Image processing system for pedestrian monitoring using neural classification of normal motion patterns

The concept of automated incident detection is based on the idea of finding suitable image cues that can represent the specific event of interest with minimum overlapping with other classes. Motion is adopted as the main cue for abnormal incident detection. Firstly scene motion information is extracted by a hardware-implemented exhaustive block-matching motion detector, followed by a storage of motion segmentation and filtering. The extracted motion information from sequences with normal events is fed to a neural network as training data. A complex neural network architecture is used to model the motion patterns in the scene, where the network states and weights are represented using complex numbers.