The paper is organized as follows. In Sec. II we survey
state-of-the-art clustering methods along with clustering performance
evaluation measures. We describe how the motioncluster boundaries are computed and the incremental learning
model in Sec. III and Sec. IV, respectively. In Sec. V we
discuss the difference among our method and state-of-the-art
alternatives. In Sec. V-A we analyze the computational complexity
and in Sec. VI we evaluate the clustering performance
of the proposed method. In Sec. VII we draw the conclusions
and future research directions.