Vision Chip

BACKGROUND Conventional computer vision systems comprise a sensor/camera with a separate computer system for data processing. Computer vision is one of the most intensive data processing applications and until recently this has been the only feasible approach. However, this architecture suffers from data transmission (I/O) bottlenecks (between the sensor, processor and memory devices) when higher performance sensors are used. Conventional systems are also power-hungry, thus constraining the use of high-performance computer vision in portable, battery-powered applications. The University of Manchester has developed a software programmable vision chip, called SCAMP, in which a programmable image processor is integrated within each pixel of the image sensor thereby creating massive parallelism (thousands of processors working together) and providing the computational power required for real-time image processing. By performing image processing on the chip the I/O bottleneck between the sensor and the processor is eliminated and the power dissipation, size and cost of the system is dramatically reduced. A single chip can replace the combination of a camera and a powerful computer. THE TECHNOLOGY The architecture of the programmable vision chip is based on a fine-grain SIMD (Single Instruction Multiple Data) processor array. Processing elements simultaneously execute identical instructions on their local data. The processors can also exchange information with their nearest neighbours. Each processing element is a simple, complete processor, comprising; Arithmetic Logic Unit, Memory, Local Control and I/O ports and also includes a photodetector. The key idea behind the SCAMP approach is the introduction of an analogue processing element (APE). Although the APE executes software instructions in a manner akin to a digital processor, it operates on analogue samples of data. This novel combination of analogue circuitry and digital architecture results in unprecedented efficiency in terms of performance, cost and power dissipation. The SCAMP-3 chip, fabricated in a 0.35 μm CMOS technology, contains a 128x128 processor array and achieves cell density of 410 processors/mm2 (a single cell measures under 50μm x 50μm). Operating with a 1.25 MHz clock it executes over 20 GOPS (billion operations per second) while dissipating 240mW of power. TURNING INNOVATIVE TH Figure 1. The ‘SCAMP-3’ Vision Chip. FROM IDEAS TO BUSINESS SUCCESS lens vision chip photodetector pre-processed imageoptical input(real-life scene) processor circuit single pixel Figure 2. Principle of the vision chip operation. The image is focused on the surface of the chip through the lens. The chip integrates image sensing with high-performance parallel image processing and outputs pre-processed images. KEY BENEFITS
• Autonomous decision making from real-time analysis.
• Low power consumption.
• Small physical size.
• Low cost.
• Software programmable.
• Fabricated using industry standard CMOS technology. APPLICATIONS
• Automotive applications (e.g. airbag deployment, collision avoidance).
• Industrial machine vision (rapid inspection, robotics).
• Domestic robots and interactive toys.
• Autonomous vehicles (terrain recognition, navigation and search).
• Image recognition (facial, medical).
• Motion detection.
• Intelligent security and surveillance systems (e.g. high speed target tracking).
• Retinal prosthesis.
• Human / computer interaction. TURNING INNOVATIVE THINKING INTO COMMERCIAL REALITY FROM IDEAS TO BUSINESS SUCCESS THE OPPORTUNITY This technology will be of interest to organisations which are involved in the above applications and in particular:
• Vehicle manufacturers.
• Suppliers and users of computer vision and machine vision systems.
• Interactive toy manufacturers. A number of smart camera systems based on the vision chip are being built and will be available for evaluation. CONTACT David Eales, Business Manager, UMIP, The Fairbairn Building, PO Box No 88, Manchester M60 1QD [email protected] Tel: +44 (0) 161 306 3153

Type of Offer: Licensing



Next Patent »
« More Software Patents

Share on      


CrowdSell Your Patent