Main Page
From EVA
About EVA Project
Embedded vision systems become indispensable measurement instruments in modern electromechanical equipment like pick and place units, industrial printing, inspection machines, industrial robots, photo-copiers, semiconductor production equipment, medical systems and future automotive equipment. Combined with traditional sensors they form an indispensable component and often the heart of the control systems and the reason of being of the machine. This role puts a high pressure on requirements, e.g. for accuracy and real-time, both for the embedded hardware architecture as well as for its embedded algorithms.
In contrast with commodity goods, the design of professional equipment is often a process in which the application sets the requirements for the algorithms and the algorithms set the requirements for the hardware architecture, and hence this architecture is an application specific one. Professional machines are often designed as product families, e.g. to serve various performances, sizes, volumes, throughputs and price ranges, meaning that solutions must be scalable. Additional customer requirements, version control and maintainability all make that the system must be flexible and (re)programmable. The high throughput and low latency requirements for its embedded machine vision, especially when used to control motion stages, most often leads to the massively data parallel processing of pixels, while task parallelism or pipelining is often used as a second method to lower the latency due to processing.
Consequently, Embedded Vision Architecture (EVA) features are: Embedded, High Accuracy, High-Throughput, Low-Latency, useable for Application Specific, Scalable, (Parallel) programmable, for Machine Vision domain.
EVA Partners
- Eindhoven University of Technology (Penvoerder)
- Faculty of Electrical Engineering, Embedded System Architectures
- Faculty of Mechanical Engineering Embedded Vision in Dynamics and Control
- Philips Applied Technologies (Apptech)
- CHESS
- Assembléon
- OTB
- BrainCenter
The partners in the EVA project will work together on the design of architectures and algorithms for embedded vision systems that are embedded in electromechanical equipment for industrial inspection and production. The three industrial partners will contribute with a key problem that needs to be solved, where Philips Applied Technologies and the TU/e will support the salvation of the problems with dedicated R&D. The heart of the problems is that the design of embedded vision systems for motion control is very application dependent, and that - due to the severe timing constraints - the hardware design should follow the algorithms that implement the application.
Objective
- The overall EVA project objective is:
- To develop a generic Embedded Vision Architecture (EVA) based on suitable applications
- To apply digital design tools to develop application specific versions of this generic Embedded Vision Architecture (EVA) template
- To further develop design tools to make flexible use of FPGAs in the designs possible
- To demonstrate this by the designing of offspring of the generic EVA for three applications:
- one Distributed vision for component placement machines
- two Vision in the loop for industrial ink-jet printing
- three Cooperating tracking cameras for Augmented Reality and Automated Vehicles
- To realize and test the offspring and successfully integrate it in existing products
- All applications have sufficiently in common (high accuracy, high speed geometric measurements with cameras) and yet are also enough different in their constraints to form interesting test cases. Applications that EVA is interested in:
- Application 1: Distributed Vision (the Assémbleon case)
- Application 2: Vision in the loop (the OTB-Engineering case)
- Application 3: Cooperating tracking cameras (the Chess-Engineering case)
