84. background subtraction for neuromorphic image sensors

Department: Computer Science & Engineering
Faculty Advisor(s): Ryan Kastner

Primary Student
Name: Alireza Khodamoradi
Email: akhodamo@ucsd.edu
Phone: 916-458-2611
Grad Year: 2019

Abstract
High event rate neuromorphic sensors are becoming more popular. Recently Samsung announced the employment of event based sensors for gesture recognition in mobile applications. The interface of event based sensors can handle limited number of events per second, limited bandwidth causes poor performance at scenes with high event rates. In this work we try to address this issue by using a fast background subtraction at sensor head to filter unnecessary events to reduce the interface load. Therefore a consumer machine connected to the image sensor will receive more useful data over a simple interface. Our approach is based on simple rate recognition and implementation by employing already existed FPGAs in sensor heads.

Industry Application Area(s)
Electronics/Photonics | Energy/Clean technology | Life Sciences/Medical Devices & Instruments

« Back to Posters or Search Results