Multi-Class Multi-Object Tracking using Changing Point Detection

摘要

This paper presents a robust multi-class multi-object tracking (MCMOT)formulated by a Bayesian filtering framework. Multi-object tracking forunlimited object classes is conducted by combining detection responses andchanging point detection (CPD) algorithm. The CPD model is used to observeabrupt or abnormal changes due to a drift and an occlusion based spatiotemporalcharacteristics of track states. The ensemble of convolutional neural network(CNN) based object detector and Lucas-Kanede Tracker (KLT) based motiondetector is employed to compute the likelihoods of foreground regions as thedetection responses of different object classes. Extensive experiments areperformed using lately introduced challenging benchmark videos; ImageNet VIDand MOT benchmark dataset. The comparison to state-of-the-art video trackingtechniques shows very encouraging results.