For many years cameras have been used in various locations to ensure the safety of people and assets. CCTV cameras can help with security in town centres while Automatic Number Plate Recognition (ANPR) can identify vehicles and other assets. We were one of the pioneers of ANPR, having long been involved in image processing research and contributing successfully to the area of intelligent surveillance.
Robust Image Processing
As the number of cameras increase, automatic monitoring systems become preferable as the workload becomes too great for individuals. Image processing systems only have value if they work, and continue to work effectively in difficult situations. We provide robust image processing solutions with several advantages. The vision system is quick to start up, learning the scene in 0.5-3 seconds. Also, illumination is unaffected by slow, rapid or even instantaneous lighting changes which means reliable location and tracking of people/objects and low false alarm rates. Finally, the imagine stabilisation process allows rapid correction of camera shake or vibration in situations where other systems might fail.
Original images courtesy of DFT-DEC and Bristol University
Image Fusion
Significant advantage can be gained by fusing images from a number of sources. Image fusion is the combination of thermal, visible and other camera sources into a composite image. This is performed such that the most important features from each camera’s view are combined. The resulting fused video leads to better surveillance task performance, by vision systems or humans, than using a single video source.
Advanced Statistical Techniques
We have novel research underway in the use of very complex computational methods for reliably identifying individual objects even in highly cluttered moving scenes. This promises a revolutionary capability, and a host of new applications, even in crowded scenes that have previously been impossible for machine vision systems to operate in effectively.