DocumentCode :
80900
Title :
Fully Unsupervised Learning of Camera Link Models for Tracking Humans Across Nonoverlapping Cameras
Author :
Chun-Te Chu ; Jenq-Neng Hwang
Author_Institution :
Dept. of Electr. Eng., Univ. of Washington, Seattle, WA, USA
Volume :
24
Issue :
6
fYear :
2014
fDate :
Jun-14
Firstpage :
979
Lastpage :
994
Abstract :
A multiple-camera tracking system that tracks humans across cameras with nonoverlapping views is proposed in this paper. The systematically estimated camera link model, including transition time distribution, brightness transfer function, region mapping matrix, region matching weights, and feature fusion weights, is utilized to facilitate consistently labeling the tracked humans. The system is divided into two stages: in the training stage, based on an unsupervised scheme, we formulate the estimation of the camera link model as an optimization problem, in which temporal features, holistic color features, region color features, and region texture features are jointly considered. The deterministic annealing is applied to effectively search the optimal model solutions. The unsupervised learning scheme tolerates the presence of outliers in the training data well. In the testing stage, the systematic integration of multiple cues from the above features enables us to perform an effective reidentification. The camera link model can be continuously updated during tracking in the testing stage to adapt the changes of the environment. Several simulations and comparative studies demonstrate the superiority of our proposed estimation method to the others. Moreover, the complete system has been tested in a small-scale real-world camera network scenario.
Keywords :
brightness; cameras; estimation theory; image colour analysis; image texture; object tracking; optimisation; transfer function matrices; unsupervised learning; brightness transfer function; camera link models; deterministic annealing; estimation method; feature fusion weights; fully unsupervised learning; holistic color features; multiple cues; multiple-camera tracking system; nonoverlapping cameras; nonoverlapping views; optimal model solutions; optimization problem; region color features; region mapping matrix; region matching weights; region texture features; small-scale real-world camera network; systematic integration; temporal features; tracked humans; training data; training stage; transition time distribution; unsupervised scheme; Cameras; Color; Estimation; Histograms; Image color analysis; Training; Unsupervised learning; Camera link model; camera network; deterministic annealing; multiple-camera tracking; nonoverlapping view; unsupervised learning;
fLanguage :
English
Journal_Title :
Circuits and Systems for Video Technology, IEEE Transactions on
Publisher :
ieee
ISSN :
1051-8215
Type :
jour
DOI :
10.1109/TCSVT.2014.2302516
Filename :
6727542
Link To Document :
بازگشت