Bio-inspired camera for surveillance in IoT

Abstract

Using an event-based camera sensor instead of a frame-based camera sensor has many benefits such as reduced power consumption and reduced file output memory size. Adding cognitive abilities onto an event-based camera sensor would further reduce the information that this sensor needs to transmit. This paper seeks to combine the advances in deep neural networks for frame-based videos with the efficiency of event-based sensors. This will be done by comparing existing algorithms used in frame-based videos to identify objects in event-based camera output. Additionally, this paper seeks to reduce the complexity of these algorithms to ensure the practical implementation of the algorithm in resource constrained settings.