Scalable Visual Analytics of Video Data

Applicant:

Professor Dr. Gunther Heidemann, Stuttgart
Universität Stuttgart
Institut für Visualisierung und Interaktive Systeme (VIS)

Professor Dr. Daniel Weiskopf, Stuttgart
Universität Stuttgart
Institut für Visualisierung und Interaktive Systeme (VIS)

Project:

Scalable Visual Analytics of Video Data (Publications)

Summary:

A substantial amount of electronic data that is acquired on a regular basis is in the form of video data. A typical example is the continuous data input from closed circuit television (CCTV) cameras. The current practice of directly viewing video footage does not scale well with the amount of video because it permanently requires user attention even for, in most parts, uninteresting video material. The main goal of this project is to support users in the interactive analysis of video data in order to efficiently identify and understand regular and irregular behavior in video recordings. One challenge is that irregular behavior cannot be completely defined beforehand and, therefore, fully automatic computer vision techniques cannot provide a complete analysis. Another challenge is the difficulty of interpreting image data of complex scenarios that is often affected by noise and that contains only partial image information due to occlusion. Our strategy is to combine partially automatic image analysis with visualization and interaction: in this way, ambiguities and uncertainties in computer-based video analysis can be resolved by the human users with their excellent capabilities of interpreting image data and identifying structure. By integrating the applicants’ expertise in computer vision, visualization, perceptionoriented graphics, and interactive systems, this project has the goal of a scalable visual analytics tool for video which allows for video analysis on various levels of abstraction and which supports single camera and multiple camera CCTV setups.

Presentations: