Technologies


Systems and Methods for Remote Tagging and Tracking of Objects Using Hyperspectral Video Sensors

Reference#: P02622


There is a need for systems and methods that provide for the acquisition and tracking of an object where tracking can be maintained even if its appearance changes (due to variations in illumination and object orientation) or if tracking is momentarily lost.

Appearance-based methods are challenged by the changes in an object's appearance. The changes may be due to several factors. These may include a) different illumination sources and conditions, b) variations of the object's orientation with respect to the illumination source and camera, and c) different reflectance properties of objects in the scene. Such variability compromises the performance of many vision systems, including background subtraction methods for motion detection, appearance-based trackers, face recognition algorithms, and change detection. Furthermore, conventional systems are unable to track objects through gaps in coverage, such as the object going out of a hyperspectral video camera's field of vision or the object going behind other objects and reappearing minutes or hours later.

APL researchers have developed a hyperspectral video surveillance (HSV) system which utilizes a hyperspectral video camera and illumination-invariant vision algorithms to simultaneously capture images with high temporal, spatial, and spectral resolution, thus combining the advantages of both video and hyperspectral imagery. The APL HSV system utilizes a sensor which captures hyperspectral imagery at near-video rates. While standard HSV systems capture only three wide-bandwidth color images, the APL HSV system utilizes many narrow-bandwidth images of the scene, thus providing several advantages. Specifically, the APL HSV system provides high spatial and temporal resolution to detect and track moving objects, allowing for the distinction between objects with similar color. In addition, the APL HSV system provides the ability to incorporate radiative transfer theory models to mitigate the effects of illumination variations. This is achieved by estimating the object's reflectance spectra from the observed radiance spectra in the image. The object's reflectance spectrum is insensitive to a wide range of illumination effects and allows for illumination-invariant video surveillance.The invariant reflectance spectrum also allows the APL HSV system to track through gaps in coverage by exploiting spectral matching algorithms to determine when the object has re-entered the field of view of the camera.

Patent Status: U.S. patent(s) 8,295,548 issued.

CONTACT:
Mr. J. E. Dietz
Phone: (443) 778-8782
ott-techmanager5@jhuapl.edu

Additional References:

Link to U.S. Patent and Trademark Office