Description: In this paper, we propose a new adaptive visual object tracking method based on
online feature evaluation approach. First, a feature set is built by combining color
histogram (HC) with gradient orientation histogram (HOG), which emphasizes both
color and contour representation. Then a feature confidence evaluation approach is
proposed to make features with higher confidences play more important roles in the
instantaneous tracking ensuring that the tracking can adapt to the appearance change
of both the object and its background. The feature evaluation approach is fused with
filter frameworks, e.g. Kalman and Particle filter, to keep the temporal consistency of feature confidence evolution.
To Search:
File list (Check if you may need any files):
中科院研究生院.pdf
Download