DESCRIPTION:
Novel algorithm automates qualitative surveillance video detection of independently moving objects in two modes: (1) real-time (in combination with sensors) and (2) very fast scanning (faster than real-time play speed) of surveillance video database and retrieval of shots containing detected movement for further review by image analysts. The approach overcomes technical hurdles associated with detecting independently moving objects with a camera that is itself in motion, i.e., not stationary.
POTENTIAL APPLICATIONS:
Technology can be integrated into systems and programs fielded by military / homeland security / intelligence / law enforcement and risk management, for example, unmanned aerial vehicles (UAVs) / drones, and smart acquisition and tracking of targets.
ADVANTAGES:
-
Time-saving: fast scanning speeds are accomplished by the application of computationally efficient algorithms which can scan compressed MPEG video at a speed much faster than a real time play speed.
-
Simplicity: requires as few as two frames and only one uncalibrated video camera to detect any independently moving target.
- Automated: reduces manpower requirements and probability of operator error.
PATENT STATUS:
Patent pending (U.S. Patent Application # 10/364,011, Filing Date: 2/12/03)
KEYWORDS:
Independent motion detection; linear system consistency analysis; unmanned aerial vehicle (UAV); video data mining; image analysis; surveillance
ADDITIONAL REFERENCE INFORMATION:
Zhongfei (Mark) Zhang, Mining Surveillance Video for Independent Motion Detection, Proc. IEEE International Conf. Data Mining (ICDM) 2002, Maebashi City, Japan, December 2002, available at Professor Zhang’s homepage:http://www.cs.binghamton.edu/~zhongfei/ with additional information accessible via the link to the Multimedia Research Lab which he directs http://www.fortune.binghamton.edu/
See also RB-176, Hierarchical static Shadow Detection for Color Aerial Images. |
