Active learning for transferrable object classification in cross-view traffic scene surveillance

Abstract

We discuss the problem of object classification in cross-view traffic scene surveillance videos in this paper. To classify moving objects in traffic scene videos into pedestrian, bicycle and variety of vehicles, an effective intelligent classification framework has been proposed which takes advantage of a transfer machine learning method to bridge the gap between source scene data and target scene data. The transfer learning algorithm makes one classifier adaptive to perspective changes instead of training two different classifiers for corresponding perspectives. The samples transferred from source scene database have saved much manual labeling work on target scene database. In this paper, we propose an active transfer learning method to decrease manual labeling work further for target scene traffic object classification. Redundant experiments are conducted and experimental results demonstrate the effectiveness and convenience of our approach. © 2012 Springer-Verlag.