The global image sensor market is expected to pass US$10.75 billion by the end of 2018, and while overall growth is predicted at a relatively modest 3.84% CAGR (because of the high weighting of sensors in mobile phones), emerging applications, which include factory automation, UAV cameras and automotive are experiencing high growth rates1.
Over the same period the global machine vision components and systems market is forecast to grow at 8.2% CAGR, while the growth rate for smart cameras in this market segment is a predicted healthy 23.5%2.
This growth means that automation practitioners need to get to grips with this technology.
Key tasks in imaging systems may include:
* Data capture
* Data transfer
* Data analysis
* System response
Data capture
In its broadest sense, data capture refers to the capture of data that can be rendered such that it represents a 2-dimensional or 3-dimensional scene.
The first data capture device that comes into the minds of most people when considering vision systems is probably the camera capturing light in the visible light spectrum, but there are many other non-contact and contact technologies and physical phenomena that can be used for data capture. These include Gamma-ray, X-ray, radar, UV, lidar, IR, sonar, magnetic, capacitive and even multi contact devices that feel for surface defects3. Figure 1 shows a sampling of these technologies across the frequency spectrum.
The capital cost of a camera and the clarity of image returned by a camera depend on such factors as camera lens, sensor resolution, sensor size, sensor technology (CCD or CMOS), light level and variation in range of light level, trigger capabilities, shutter operation (rolling shutter or global shutter), requirement for monochrome or colour, suitability of application to line scan or area scan, frame rate, image processing and post processing / image analysis within the camera, 2D or 3D (single or stereo), and interface requirements. Many of these choices are determined by the nature of the target application (is subject static or moving? and if moving, how fast is it moving compared to frame rate?) and environmental conditions such as constant lighting or strobed lighting.
The correct matching of camera and lighting to application is essential for the success of an imaging project. While most automation engineers will rely heavily on vendors to help specify project solutions, they should also question vendors on equipment selection because when it comes to machine vision there is no one-size-fits-all.
CCD or CMOS?
With the rapid developments that have taken place in CMOS sensors, driven by the demand for camera sensors in consumer devices, some of the previous advantages (image quality, light sensitivity and inherent global shutter) of CCD technology for industrial vision systems have fallen away for most visible light applications. CMOS sensors for machine vision can match CCD in these areas and also offer higher frame rates and at lower cost. On either side of the visible spectrum choices may be swayed by sensor construction. For instance CCD may outperform CMOS for near infrared applications while specially treated CMOS sensors may be a better choice for UV imaging4.
Monochrome or colour?
Generally for dimensional inspection and outline recognition a monochrome camera will be adequate for the task. Cognex5 suggests colour for applications such as part sorting, colour recognition, and assembly verification and inspection – especially when parts are small.
16-bit or 24-bit colour?
The human eye can discriminate approximately 10 million colours. 16-bit cameras can discriminate 64 k colours. If the application requires discrimination to match the human eye then 24-bit colour cameras are necessary. Otherwise 16-bit colour cameras will meet most colour requirements.
In-camera or remote processing?
Cameras may act as little more than frame grabbers, passing the image to a remote processor for analysis, or they may be smart devices with on-board image processing capability. In-camera processing may mean lower latency and lower jitter for control applications. Remote processing may make greater demands on network bandwidth and result in higher latency and jitter.
In some applications, image processing and analysis may be optimally divided between in-camera and remote processors.
Data transfer
Where images must be analysed off-camera or need to be archived for inspection verification purposes, they need to be transferred from the data capture device to another work-station or server.
As network connected camera counts, image sizes and frame rates increase, networks and network devices need to be able to provide higher bandwidths. This has led to the development of high capacity digital image transfer protocols such as Camera Link, Camera Link HS, GigE, 10 GigE and CoaXPress and also to the provision of Ethernet, Thunderbolt, FireWire IEEE1394 and USB3 interfaces on cameras.
These interface standards are summarised in the glossary of this article.
Data analysis
Where analysis does not take place in-camera, or there is a need for additional off-camera analysis this typically happens on a PC. On the PC the heavy lifting may be done by the CPU, but there is a growing trend for image analysis tasks to be delegated to single or multiple PC graphics processing units (GPUs).
There are image processing libraries available from multiple vendors for use with development environments like Microsoft Visual Studio, enabling SIs or end-users to develop complex applications. Examples include ImagingLab Robotics Library for NI’s LabView, Matrox Imaging Library and LEADTOOLS Image Processing SDK.
Alternatively many of the suppliers of imaging hardware can provide integrated development environments (IDEs) for rapid prototyping and implementation of vision projects. Examples include the Cognex In-Sight family, Matrox Design Assistant and Banner Engineering’s PresencePlus vision software.
System response
In a machine vision application there is typically a system response required once an image has been captured and analysed. In a sorting application this response may be as simple as activating a digital output to energise a solenoid controlling a diverter. In a multi-axis articulated robotic application the response may be much more complicated – for instance telling the robot controller in real-time that there is a part number XYZ at a particular set of co-ordinates and lying in a particular orientation.
In the former scenario a smart camera fitted with some discrete I/O can be easily integrated with a PLC or PAC. In the second scenario a smart camera or off-board processor needs to communicate a much larger amount of data to the robot controller in a known protocol.
Glossary
Camera Link:
Camera Link is a robust communications link using a dedicated cable connection and a standardised communications protocol. The hardware specification standardises the connection between cameras and frame grabbers. The interface includes provisions for data transfer, camera timing, serial communications, and real time signalling to the camera6.
Available since: 2000
Speed: 2Gbps (1 cable) - 6Gbps (2 cables)
Cable length: 10 metre
Adapter: Frame Grabber
Connection type: Point-to-point
Camera Link HS
Camera Link HS is designed to meet the needs of vision and imaging applications. Its low latency, low jitter, real-time signals between a camera and a frame grabber carry image data and configuration data6.
Global approval: 2012
Speed: 3.125 Gbps on 2 cables – scalable up to 8 conductors
Cable length: 15 – 300+ m over fibre
CoaXPress
CoaXPress (CXP) is an asymmetric high speed point to point serial communication standard for the transmission of video and still images, scalable over single or multiple coaxial cables. It has a high speed downlink of up to 6.25 Gbps per cable for video, images and data, plus a lower speed, 20 Mbps uplink for communications and control. Power is also available over the cable ('Power-over-Coax')7.
Global approval: 2011
Speed: 6.25 Gbps per cable (up to 40 m)
Cable length: Up to 13 0m (but at reduced bandwidth)
Connection type: Point-to-point
FireWire IEEE1394
The IEEE 1394 interface is a serial bus interface standard for high-speed communications and isochronous real-time data transfer. It was developed in the late 1980s and early 1990s by Apple, who called it FireWire. The specification originally covered data rates of 100, 200 and 400 Mbps in half-duplex mode. Later enhancements to the standard define data rates of up to 3.2 Gbps10.
GigE Vision
GigE Vision is a global camera interface standard developed using the Gigabit Ethernet communication protocol. It allows for fast image transfer using low cost standard cables over very long lengths. With GigE Vision, hardware and software from different vendors can interoperate seamlessly over GigE connections. The standard is widely adopted, with many leading companies offering GigE Vision compliant products5.
Available since: 2006
Speed: 1 Gbps
Cable length: 100 meter
Adapter: On PC, NIC, Frame Grabber
Connection type: Network
10 GigE Vision
Version 2.0 of GigE introduces support for faster data transfer through 10 Gigabit Ethernet and link aggregation. It also enables transmission of compressed images (JPEG, JPEG 2000 and H.264), accurate synchronization of multi-camera systems and enhanced support for multi-tap sensors5.
Launched: 2006
Speed: 10 Gbps per cable – up to 40 Gbps on 4 cables
Cable length: Depends on physical layer used – many km if over fibre
Adapter: Frame Grabber
IR
Infrared (IR) light is electromagnetic radiation with longer wavelengths extending from the nominal red edge of the visible spectrum at 700 nm to 1 mm. This range of wavelengths includes most of the thermal radiation emitted by objects near room temperature10.
lidar
A remote sensing technology that measures distance by illuminating a target with a laser and analyzing the reflected light10.
Thunderbolt
Thunderbolt is an interface standard that defines bi-directional operation transfer at 10 Gbps speed, and offers daisy chaining to multiple devices. In April 2013, Intel announced plans for the Thunderbolt controller, doubling the bandwidth to run at 20 Gbps. Thunderbolt technology enables simultaneous 4K video file transfer and display8.
USB3 Vision
Based on USB3.0 aka SuperSpeed USB. USB3.0 is a USB technology capable of transferring data at up to 5Gbps. Completion of the specification for USB3.1 was announced at the end of July 2013. This will operate at up to 10 Gbps9.
Speed: > 350Mbps
Cable length: 5m passive. Longer for active cables.
UV
Ultraviolet (UV) light is electromagnetic radiation with a wavelength in the range between 400 nm and 10 nm. These frequencies are invisible to humans, but Near UV is visible to a number of insects and birds10.
References
1. Markets and Markets - Market report: Image Sensors Market Analysis and Forecast (2013 – 2018); http://tinyurl.com/n73c6f6
2. Markets and Markets - Market report: Machine Vision Systems & Components Market Forecast (2013 – 2018); http://tinyurl.com/lghs2d7
3. GE Energy - Oilfield Technology; http://tinyurl.com/n7gksll
4. LASER+PHOTONICS - Issue 2013/02 Article: CMOS vs. CCD imagers: Nixon O & Eric Fox; http://tinyurl.com/ledep3e
5. Cognex - Expert Guide: When to Choose a Color Vision System for Your Application; http://tinyurl.com/n3bu6cj
6. AIA - Web site http://www.visiononline.org
7. CoaXPress - Web site http://www.coaxpress.com
8. ntel - Web site http://www.intel.com
9. USB Implementers Forum - Web site http://www.usb.org
10. Wikipedia - Web site http://en.wikipedia.org
About the author
Andrew Ashton has electrical, mechanical and business qualifications and has been active in automation and process control since the early 1980s. Since 1991 he has headed up a company that has developed formulation management systems for the food, pharmaceutical and chemical manufacturing industries and manufacturing solutions involving the integration of various communication technologies and databases. Developed systems address issues around traceability, systems integration, manufacturing efficiency and effectiveness. Andrew is a contributing editor for SA Instrumentation and Control.
| Tel: | +27 11 543 5800 |
| Email: | [email protected] |
| www: | www.technews.co.za |
| Articles: | More information and articles about Technews Publishing (SA Instrumentation & Control) |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version