Bookmark and Share

Other Semiconductor Devices for Computer Vision

Click HERE for additional resources


Everything that isn't a processor or a sensor.

Embedded vision applications involve more than just programmable devices and image sensors, since these devices require other components for creating a complete system.  Most applications require data communications of pixels and/or metadata, and many designs interface directly to the user.  Some embedded vision systems connect to mechanical devices, such as robots or industrial control systems.


Devices for memory, storage, networking, bus interfaces, etc.

The list of devices in this “other” category includes a wide range of standard products for embedded systems.  In addition, some system designers may incorporate programmable logic devices or ASICs. In many embedded vision systems, power, space and cost constraints require high levels of integration with the programmable device—often into a system-on-a-chip (SOC) device.  An example of an SOC is the Freescale i.MX family of Applications Processors that integrate ARM CPUs, application accelerators, and a wide variety of peripherals.


Processors can integrate megabytes of SRAM or even DRAM, so many designs will not require off-chip memory.  However, computer vision algorithms often require multiple frames of sensor data to track objects.  Off-chip memory devices can store gigabytes of memory, though accessing external memory can add hundreds of cycles of latency.  For systems with a 3D graphics subsystem, the system will usually already require external memory (as much as a gigabyte on a high-end system) to store the frame buffer, textures, z-buffer, etc.  Sometimes this graphics memory is stored in a dedicated, fast memory bank that uses specialized DRAMs (ex: GDDR5) from companies such as Qimonda and Hynix Semiconductor.


Some embedded vision implementations store video data locally to reduce the amount of data that needs to be sent to a centralized system. For a solid-state, non-volatile memory storage system, the storage density is driven by the size of Flash memory chips.  Sandisk has announced 8 gigabyte devices in their 19nm NAND chip fabrication technology, allowing extremely large, fast and low-power storage in an embedded vision system.

Networking and bus interfaces:

Mainstream computer networking and bus technology has finally started to catch up to the needs of computer vision to support simultaneous digital video streams. With economies of scale, more embedded vision systems will use standard buses like PCI and PCIe.  For networking, Gigabit Ethernet (GbE) and 10GbE interfaces offer sufficient bandwidth, even for multiple high-definition video streams.  However, the trade association for Machine Vision (AIA) continues to promote Camera Link, and many camera and frame-grabber manufacturers use this interface.  National Semiconductor (now part of Texas Instruments) manufactures discrete Camera Link transmitter/receiver chips, but most devices will integrate Camera Link on-chip.


Additional Resources: