CAREER: Elements in Solving the Multiple Object Tracking Problem

University Of California-Santa Cruz, Santa Cruz CA

Investigators

Abstract

The proposed project addresses several challenging issues in multi-object tracking such as failure modes, occlusion, background clutter, varying object appearance, multi-object confusion, and high computational complexity. This project will investigate five key elements in multi-object tracking including (1) a trajectory tracking framework, (2) a joint multi-object representation, (3) image-based likelihood functions, (4) the integration of object detection in tracking, and (5) the compact and efficient discriminative feature tracking. Based on recent preliminary results, it is expected that significant improvements in multi-object tracking will be achieved by integrating these elements into the proposed probabilistic estimation framework. An experimental system for tracking multiple pedestrians and vehicles will be implemented for testing the developed algorithms. The proposed educational program will focus on developing graduate and undergraduate courses in the area of computer vision, establishing an active research laboratory in visual computing and human-computer interaction, and promoting collaboration with the industry to offer students the opportunities to work on real-world research problems. Success of this project will provide a novel solution to the multi-object tracking problem and significantly advance the state-of-the-art. The proposed project is expected to have broader impacts in the areas of video surveillance, human-computer interaction, and intelligent teleconferencing. The proposed research will also have an impact on establishing an educational program in computer vision at the University of California, Santa Cr