Smart Cameras Target Simplicity in Operation, Communication
POSTED 01/10/2017 | By: Winn Hardin, Contributing Editor
In the first quarter of 2016, North American sales of machine vision components and systems fell 11% year over year to $546 million. That included a 2% contraction of smart camera sales to $75 million, representing nearly 14% of the machine vision market share.
“Capex budgets tightened up in late Q4 2015, Q1 2016 and into Q2, loosening slightly,” says Alex Shikany, AIA’s Director of Market Analysis. “Couple that with the slower manufacturing industry over that time, and you didn’t have as many customers ready to shell out cash on new smart cameras for their factories.”
Much like the overall market, smart camera downsizing was largely cyclical, according to Shikany. “Q3 we saw a normalizing of the data, so we know it’s not a systemic trend,” he says.
Indeed, by the end of Q3, the smart camera segment returned to its previous growth trajectory with a 10% increase to $243 million in the first nine months of 2016 and a 24% jump compared to Q3 2015.
Smart camera and vision sensor manufacturers are meeting this market demand by expanding the range and capabilities of ‘vision systems in a box’ – from vision sensors to networkable vision systems made from distributed smart cameras – that are affordable, simple, and reliable while addressing limitations surrounding processing power and networking protocols.
Accommodating a Range of Capabilities
A conversation about smart cameras typically includes vision sensors as the line between the two product types often is blurred. Vision sensors and smart cameras both comprise an imaging sensor, light, computational elements, software, and I/O, all in a single housing.
“The difference between the two is that vision sensors tend to be a lower entry price point because of lower development costs, but they have fewer capabilities than smart cameras,” says Steve Geraghty, general manager of Teledyne DALSA Industrial Products (Billerica, Massachusetts). “On the flipside, vision sensors are very easy to use. From what we see in the market, that’s a real benefit.”
Teledyne DALSA’s offerings in the space include the BOA Spot vision sensor and its newest smart camera release, the 5-megapixel BOA2. According to Geraghty, both products share the same fundamental automation requirements for industry with applications including assembly verification, 1D and 2D barcode inspection, and robotic guidance.
“BOA2 has more software tools and tends to be more horizontal in its capabilities,” says Geraghty. “It can do spot checking, edge finding, intensity checks, and matching, plus more elaborate tools for handling environmental variation.”
BOA vision systems are available with three different software applications, including two variants of INspect Express software, and Sherlock Embedded for the BOA PRO family of cameras. Equipped with a flexible programming interface, the more powerful Sherlock vision software allows users to define tools from the basic (edge, count) to the complex (calibration, pattern matching), and then define programming around those tools.
With its In-Sight product line, Cognex (Natick, Massachusetts) targets the full complement of smart cameras users, from sensor to networkable smart cameras. One of the most notable additions to this line is the In-Sight 2000. This vision sensor with built-in field-changeable lighting and optics is designed to solve presence/absence and basic pass/fail measurement applications, making In-Sight Explorer software available to the vision sensor line for the first time. The In-Sight 2000’s field-changeable components include lenses, filters, polarizers, and light rings, including white, red, IR and blue. Additionally, the company rolled out a high-performance In-Sight 2000 color vision sensor. “The color vision processing is extremely powerful, but unlike most color systems, this one is easy to set up,” notes John Keating, Director, Product Marketing, In-Sight.
For more complex or challenging applications, In-Sight Vision Systems, including the In-Sight 8000 “give complete flexibility for vision tools with an EasyBuilder interface, backed up with a flexible spreadsheet configuration,” Keating says. “There is a much more comprehensive vision tool set with full parameter control.”
Addressing Processor Performance
Despite advantages such as low cost and ease of use, smart cameras still face challenges, especially as customers’ expectations of the product grow more demanding. “At up to 5 MP you have a pretty beefy processor with 1.5 GHz speed, and that requires more memory to handle the higher resolution,” says Teledyne DALSA’s Geraghty. “Our standpoint is that you should be able to achieve a fairly rapid cycle whether you’re using 5 MP or VGA camera.”
In the past, more powerful processors have meant more heat generation, which is anathema to small-form electronics, including smart cameras. That’s changed thanks to a number of factors, according to Cognex’ Keating. “Advances in imaging technology and processing technology have broken through a big barrier here. For example the 8405 is a 5MP extremely high performance engine, but in the size of a “dumb camera”. There are advances of course in our own hardware design to dissipate the heat, but much of this is done by taking advantage of the processors and using them efficiently.”
Accommodating a customer’s ideal processing performance, however, means the smart camera may not be the ideal choice for some customers. “It’s the point of diminishing returns,” Geraghty says. “You want a smart camera that’s smaller, lower power, and has adequate performance. As soon you cross that threshold where you need to do 60 parts per second at 5 MP, then that’s beyond the capability of any embedded processor. At that point you should be looking at vision systems because the memory requirements are so demanding.”
While Cognex’ higher resolution cameras tend to incorporate the highest-performance processors, “it’s also true that lower resolution cameras — i.e., VGA, SVGA — also sometimes need the highest performance processors,” Keating says. “For example, our highest-performance ‘dumb camera size’ cameras are the 8402 2 MP and 8405 5 MP cameras. The processing power in those products tends to be needed to process vision tools that run on the entire image.” Meanwhile, the same processing power in Cognex’ 8400 VGA is used to run vision tools on the smaller image size in the fastest time to keep up with much faster frame rates. “The processing power needed is more a function of the application requirement — parts per minute and complexity of the vision task than it is of resolution,” Keating explains.
Cognex has another solution for smart camera lovers wanting to tackle high resolution, high frame rate applications, the In-Sight VC200 multi-smart camera system. “The VC200 vision platform is a very different approach with a small vision controller for managing the vision task, creating web-based HMI, and communicating with the factory network,” said Keating. “The smart cameras process the vision task locally and the results are managed by the controller. This also introduces a very flexible configuration interface to have more control over camera acquisition and to share results from multiple cameras.
Smart camera makers must contend with concerns over form factor as well when developing products. “The one thing that smart cameras have demanded from design standpoint is the need to minimize resources to keep the cost and footprint down,” Geraghty says. “You have to think about designing software in a different way, and you can leverage things like DSP [digital signal processing] to offload some of the repetitious processing that an Intel processor would otherwise do. You can also leverage FPGA to do preprocessing.”
Such considerations force manufacturers “to think more innovatively about how we can reduce the footprint for the system,” Geraghty continues. “It’s never going to be acceptable to produce smart cameras that are four times bigger because they have an Atom-based processor.”
Adapting to Varying Industrial Networking Protocols
Just as smart camera manufacturers must address challenges related to processing power, they also should accommodate a range of industrial networking standards in automation. The overarching problem is that each PLC manufacturer has its own variant of industrial Ethernet. Per Teledyne DALSA’s Geraghty, newer protocols like Ethernet/IP and PROFINET are gradually replacing older standards including EtherCAT and Profibus.
Because different customers and markets around the world have different communication needs, “our cameras must be able to communicate across all of the industrial protocols and with standard discrete I/O as well,” says Cognex' Keating. “Wherever possible, we integrate the factory protocol directly on the cameras. Some emerging protocols do require a hardware component. This requires supporting third-party converters, which we do as well.”
Cognex also has standard I/O modules it manufactures to expand the discrete I/O for smart cameras. In addition, the company’s VC200 supports factory protocols on board, but it has an expanded discrete I/O capability built in to the controller itself.
Heading into 2017, smart camera makers will continue their focus on simplicity of operation, communication, and integration. “The future of smart cameras will be defined by software, not hardware,” says Teledyne DALSA’s Geraghty. “Our role is to make the customer’s job easier by making the products more capable.”