作者: S. Kamijo , Y. Matsushita , K. Ikeuchi , M. Sakauchi
DOI: 10.1109/6979.880968
关键词:
摘要: We have developed an algorithm, referred to as spatio-temporal Markov random field, for traffic images at intersections. This algorithm models a tracking problem by determining the state of each pixel in image and its transit, how such states transit along both x-y axes well time axes. Our is sufficiently robust segment track occluded vehicles high success rate 93%-96%. has led development extendable event recognition system based on hidden model (HMM). The learns various behavior patterns vehicle HMM chains then, using output from system, identifies current chains. can recognize bumping, passing, jamming. However, including other training set, be extended those events, e.g., illegal U-turns or reckless driving. implemented this evaluated it results, demonstrated effectiveness.