The objective of this project is the development of embedded vision systems for intelligent transport. The aim is to capture the specificities of this field of application and incorporate them into a holistic design flow. In this way, we will develop embedded vision systems adapted for autonomous platforms and vehicles and to be incorporated to the traffic control and monitoring infrastructure. The main challenge will be the implementation of an important amount of computing power under a restricted power budget. The conventional approach, in which the different components are developed separately from specifications derived from a high-Ievel description, can be inefficient, leading to sub-optimal performance. Our approach consists of multi-parametric and multi-Ievel optimization.
We will develop a system description tool that will allow us to navigate the hierarchy of the vision system and propagate specifications and restrictions from the device- to the application-Ievel and vice versa.