Fine-Grain Parallel Cellular Processor Arrays in 3D Silicon Technologies
Lead Research Organisation:
University of Manchester
Department Name: Electrical and Electronic Engineering
Abstract
The topic of this proposal is the design of silicon microelectronic circuits that integrate thousands of simple processing cores on a single chip. In particular we will investigate circuits that integrate image sensing and image processing capabilities in a single device. These devices, sometimes called 'vision chips', perform processing using massively parallel cellular processor array architectures. High computational performance and low power consumption of such devices make them uniquely suitable for applications in embedded real-time computer vision applications such as autonomous robots, industrial machine vision, automotive security and driver assistance, interactive toys, visual prosthesis, etc. During the course of the proposed research we will develop novel vision chip circuits and architectures suitable for implementation in emerging microelectronic fabrication technologies, in particular, using 3D integrated (multi-layer stacked die) devices. We will verify our ideas via fabrication of integrated circuit prototypes and by demonstrating their computational capabilities in realistic applications.
Publications
Carey S
(2014)
Vision chip with high accuracy analog S<sup>2</sup>I cells
Carey S
(2013)
Low power high-performance smart camera system based on SCAMP vision sensor
in Journal of Systems Architecture
Carey S
(2013)
Mixed signal SIMD processor array vision chip for real-time image processing
in Analog Integrated Circuits and Signal Processing
Carey S.J.
(2013)
A 100,000 fps vision sensor with embedded 535GOPS/W 256×256 SIMD processor array
in IEEE Symposium on VLSI Circuits, Digest of Technical Papers
Mroszczyk P
(2015)
Trigger-Wave Asynchronous Cellular Logic Array for Fast Binary Image Processing
in IEEE Transactions on Circuits and Systems I: Regular Papers
Mroszczyk P
(2014)
Tunable CMOS Delay Gate With Improved Matching Properties
in IEEE Transactions on Circuits and Systems I: Regular Papers
Wang B
(2014)
A Fast Self-Tuning Background Subtraction Algorithm
Wang B
(2013)
AMBER: Adapting multi-resolution background extractor
| Description | We have developed new processor architectures, particularly suitable to applications in machine vision. We have tightly integrated arrays of simple processors and image sensors, for compact, low-power, high-performance systems. We have demonstrated working hardware prototypes of these devices, in VLSI technologies, and developed software tools that enable users to effectively program these devices. This provides foundations for using the vision chip devices in various applications. |
| Exploitation Route | Our work has advanced the design and implementation of fine-grain processor arrays, in particular 'vision chips', based on tight integration of processing and sensing. There are opportunities for applying this technology in challenging application scenarios, especially where, small-size, low-power is important (e.g. wearable systems, autonomous robots) or high-speed (e.g. machine vision in manufacturing). We are making the current hardware prototypes and software available to interested collaborators, to be used in further research projects. |
| Sectors | Aerospace, Defence and Marine,Agriculture, Food and Drink,Digital/Communication/Information Technologies (including Software),Electronics,Healthcare,Manufacturing, including Industrial Biotechology,Transport |
| Description | ETH Zurich |
| Organisation | ETH Zurich |
| Country | Switzerland |
| Sector | Academic/University |
| PI Contribution | We provide expertise and vision sensors hardware |
| Collaborator Contribution | Expertise on vision algorithms, development of software |
| Impact | Research papers on applications of vision sensors |
| Start Year | 2015 |
| Title | SCAMP-5 vision chip |
| Description | A vision sensor integrating CMOS imager with a 256x256 processor array for ultra-fast and/or low-power machine vision applications. |
| Type Of Technology | Physical Model/Kit |
| Year Produced | 2010 |
| Impact | Used by collaborators |