Processing of a time sequence of digitized images, e.g. for interpretation of road traffic situations from a vehicle

摘要

The method is based around a time delay neural network that has an architecture with a lower level (NE1) that has space - time regions (RZB) for the part display fields. A second level (NE2) connects with the elements of the first. The training of the receptive fields is made using a series of adaptive training image sequences.