Tracking of Multiple Objects Using MPEG-7 Visual Standards

Monitoring and surveillance systems play an important role in many aspects of present day applications. Multiple cameras are needed to completely cover an environment for monitoring activity and establishing correspondence between the views obtained by different cameras of the same person. A framework for a semi automatic system for visual surveillance activity is proposed in this paper. The minimum graph cut algorithm results in a qualitatively better segmentation. An approach for consistently/uniquely labeling a person depending on his color and textural features is discussed. Every foreground segmented image is sent into each of the MPEG-7 feature extracting modules and a label is attached to it depending on the feature data and correspondences of objects in different cameras are established by verifying the labels of the images.