Low Bandwidth System for Adaptive Video Acquisition and Object Tracking

NU 2020-022

INVENTORS
Srutarshi Banerjee
Henry Chopp
Juan Gabriel Serra Pérez
Zihao Wang
Olliver Cossairt
Aggelos Katsaggelos*

SHORT DESCRIPTION
Method for high-resolution imaging in bandwidth constrained-applications

BACKGROUND
In a world which is increasingly being dominated by cloud-based computer architectures, internet bandwidth capabilities between the remote users and the cloud are often the limiting factors in performance. This is particularly relevant for remote imaging applications, which are quite bandwidth intensive. Imaging under low bandwidth connections involves video compression-decompression, which adversely affects transmitted video resolution. One solution is to eliminate the transfer of redundant portions of videos or reduce the image quality, so that bandwidth is reserved for particular ROIs (regions of interest). However, in a dynamic activity as video streaming, correctly identifying redundancies remains a challenge for such adaptive video acquisition techniques.

ABSTRACT
Northwestern researchers have developed a new technology that utilizes an adaptive host (high computational power) and a chip (high-resolution focal plane array with low computational power) for video transmission via a low-bandwidth communication channel.  The chip, specifically, segments video frames into quadtree (QT) segments and then transmits QT-segmented video frames to the host. The limited channel bandwidth between the host and chip introduces a distortion in the video frames due to QT compression. Subsequently, the host computes the optimal QT reconstruction by minimizing a weighted rate distortion equation using Viterbi algorithm with the user-defined weights, thereby allowing for adaptive acquisition.Acquisition performance is measured using MOTA (multiple object tracking accuracy) as a metric.  This work, which is one of the first in its kind by virtue of its end-to-end manner, has significant applications in surveillance, object tracking and defense applications.

APPLICATIONS  

  • Defense and military imaging applications in remote areas with manned/unmanned missions
  • License plate imaging for traffic ticketing
  • Facial recognition in low-bandwidth CCTV
  • Underwater imaging applications with constrained/leaky communication channel
  • Imaging in deep space missions with high transmission delay over communication channel

ADVANTAGES  

  • Increased capability to detect and follow objects/anomalies within image frames
  • Greater ability to provide high resolution image/video on particular regions of interest
  • Unimpeded ability to operate even under low bandwidth situations

PUBLICATIONS
Banerjee S, Serra J, Chopp H, Cossairt O and Katsaggelos A (2019) An Adaptive Video Acquisition Scheme for Object Tracking. 2019 27th European Signal Processing Conference.

IP STATUS
A provisional patent application has been filed.

Patent Information: