Section: Application Domains

Embedded Computer Vision for low-power Bolometric Imaging

In cooperation with Schneider Electric, we have developed techniques for embedded real time image analysis and tracking algorithms using Bolometric Imaging Sensors. Such sensors capture light in the far infrared and return an image where each pixel is a measurement of temperature in degree centigrade. We have designed an integrated low-cost sensor that combines an 80x80 pixel Bolometric imager with a low power micro-processor. The device provides real-time, embedded image processing for target detection and activity analysis where all sensing and interpretation is local. No images are recorded and only relevant information is about activity is communicated. The design of this system is under consideration for a patent, and thus has not been published. The software system has been registered with French APP , and is to be licensed to Schneider Electric for use a line of sensors for detecting falls and other activities for assisted living for seniors, as well as monitoring customer activities in commercial environments.