TRP: A Foundational Platform for High-Performance Low-Power Embedded Image Processing

Abstract

Embedded image processing systems face stringent and conflicting constraints which commonly result in developers overly specialising systems to the problem-at-hand. In other words, they give priority to efficiency, which is an immediate concern, over the longer term development cost reduction benefits of building reusable components. In this paper, we present the foundational TULIPP Reference Platform (TRP) which enables making domain-specific generality versus specificity trade-offs through the definition of TRP instances. Each TRP instance includes the key software and hardware components for a given domain as well as productivity-enhancing components if these can be accommodated within the typical constraints of the domain. While TRP instances primarily enable intra-domain reuse, they also enable inter-domain reuse as collections of components used in one instance may be straightforwardly reused in other instances. At present, TRP instances are defined for the space, medical, automotive, robotics, and Unmanned Aerial Vehicle (UAV) domains.