METHOD AND DEVICE FOR CLASSIFYING MOTION BETWEEN TRAVELING OBJECTS, IMAGE RECOGNIZING DEVICE, AND METHOD AND DEVICE FOR RECOGNIZING TRAVELING OBJECT

摘要

PROBLEM TO BE SOLVED: To make classifiable various motions between traveling objects. SOLUTION: A risk vector V/D obtained by dividing a relative speed vector V between the traveling objects by D=d+ε(d is distance between traveling objects andεis a positive constant or zero) is calculated (26), the vector V/D is turned so that the speed vector of one traveling object always point the X direction (30A) and quantized (31), its time series are stored as observation series in a memory 27A, observation series (O1O2...OT) are cut out from the memory 27A after a prescribed time passed when|V/d|is larger than a set value C, the similarity of the observation series to a reference series is calculated (32) and whether the motion between the traveling objects is a collision or not is judged on the basis of the similarity (33). The similarity is calculated on the basis of an imbedded Markov model or decided by using a neural network.