Low-pixel Object Recognition

  • A new resolution-enhanced automatic target detection and recognition system able to handle the challenges of detecting and recognising targets of extremely small numbers of pixels, such as surveillance from a long distance.
  • Machine learning based advanced algorithms for detection and recognition of various objects such as humans, vehicles, animals.
  • Capabilities of recognising different objects at extremely low resolution, e.g., 11×11 pixels for humans and 9×9 pixels for vehicles. The type of the vehicle can even be recognised for vehicles of 16×16 pixels.
  • Applicable to smart city applications, smart building and utility infrastructure management, intruder detection, animal tracking and so on, based on object detection and/or recognition in low-lighting conditions.
  • Co-funded by CENSIS and Thales UK, and won a Sottish Knowledge Exchange Award in 2018.

References:

H. Zhang, C. Luo, Q. Wang, M. Kitchin, A. Parmley, J. Monge-Alvarez and P. Casasea-de-la-Higuera, “A Novel Infrared Video Surveillance System Using Deep Learning Based Techniques”, (Springer)Multimedia Tools and Applications, in press (accepted), 2018. (doi: 10.1007/s11042-018-5883-y)(Free Full Text Downloading)