Smart cameras get smarter, easier to use
By Ann R. Thryft, Contributing Technical Editor -- Test & Measurement World, 8/1/2009 2:00:00 AM
Smart cameras are becoming smarter as the processors upon which their intelligence is based gain computing power and as more powerful sensors provide better image quality. Smart cameras are also becoming smaller, more integrated, and easier to use.
 Basler’s eXcite smart cameras, shown with a heat sink (left) and a contact plane (right), are equipped with 64-bit, 1-GHz MIPS processors. Courtesy of Basler Vision Technologies. |
In contrast to standard machine-vision cameras, a smart camera includes some kind of computing module with local memory, a sensor, and often a frame grabber, as well as a communication module with interfaces such as Ethernet, digital I/O, and perhaps a serial port, said Matrox Imaging’s Fabio Perelli, product manager for smart cameras. Although smart cameras often use FPGAs (field-programmable gate arrays) or DSPs (digital signal processors) for their intelligence, general-purpose processors such as the Intel Atom are becoming more common because they increase processing power and make cameras simpler to use.
As DSP chip performance has increased, image sensors have also improved. Many sensors can now run complex, performance-intensive algorithms that previously required a PC-based vision system, said Narayan Subramaniam, product manager for Cognex’ vision systems business unit. For example, the company’s pattern-recognition algorithm used to be available only on a PC, but now it’s available on even entry-level smart cameras.
| Read more from our August 2009 Machine-Vision & Inspection Test Report. |
Aside from components, the other major difference between smart cameras and standard cameras is the fact that smart cameras require software, said Patrick Gailer, business development manager for Stemmer Imaging. “[Smart cameras] can range from one that is extremely easy to use, such as a low-end camera with point-and-click software, to a Linux-based, open-source, 'naked’ smart camera that requires you to build your own software.”
 A typical smart camera acquires, converts, and digitizes an image; places it into memory; processes the image and extracts information from it; and then outputs the result. Courtesy of Vision Components. |
A smart camera’s basic software is a shell or operating system, and most are sold with preloaded application software for system integrators or end users, said Endre Toth, business development manager for Vision Components. “Since we are mainly an OEM provider of board-level and packaged smart cameras, most of our smart cameras come only with basic software,” he said. Customers can also use an optional, open-source machine-vision software package that runs on all of the company’s cameras and is used in semiconductor and electronics inspection.
Although plain machine-vision cameras can also perform image processing, the processing is usually done with an FPGA, said Michael Schwaer, product manager at Basler Vision Technologies. Machine-vision cameras need a PC to evaluate the captured images, but a smart camera does not. Compared to PC-based systems, smart cameras are cheaper, easier to use, take up less space, and need less system integration. A PC-based vision system usually requires more effort from the system engineer, since it needs more complex wiring between its components. “A smart camera is easy to integrate because the PC is already inside,” he said. “Only a digital I/O cable has to be installed; no further cable for monitor or keyboard is needed.”
 Linescan smart cameras, such as the In-Sight 5604 from Cognex, are still relatively new, and are especially useful in applications that call for lower resolutions. Courtesy of Cognex. |
In contrast to smart cameras, PC-based vision systems involve a frame grabber and software running on a PC, said Subramaniam. They target very different types of applications. “For example, in solar-cell inspection, if all you want to do is detect edge cracks or perhaps discontinuities in the bus bars, you don’t need very high resolutions, so a smart camera fits in well,” he said. “The maximum resolution in our smart cameras is 1600x1200 pixels. However, if you are trying to detect pinholes or the grain pattern on a solar wafer, you need the higher resolution of a PC-based system.”
Schwaer said that smart cameras are a good fit for applications that require only a moderate amount of CPU performance; examples include bar-code reading and number-plate detection, which use algorithms like edge detection or blob analysis merely to compute a result. In contrast, a PC-based system is a better fit for applications that need high resolution and high frame rates, such as web inspection.
With faster, low-power processors such as the Intel Atom, the types of applications engineers can address with smart cameras increases, said Stemmer’s Gailer. “But whenever you need more processing power, you will fall back on the PC side,” he said. “Both will coexist for a long time.”
Weighing costs
Entry-level systems based on smart cameras, not PCs, may be cheaper to purchase, but as the vision system’s resolution and processing power increase, PC-based systems usually become less expensive. If a vision system needs several cameras, a PC is more cost effective because it can handle multiple cameras in a single system, said Perelli of Matrox Imaging. Implementing multiple smart cameras makes the cost go up faster, because each of them costs more than a traditional camera.
 Smart cameras such as Matrox Imaging’s Iris GT are beginning to incorporate general-purpose processors like the Intel Atom that improve the cameras’ ease of use. Courtesy of Matrox Imaging. |
“Let’s say you need to accommodate three different levels of inspection: an entry-level system with two cameras looking at a PCB [printed-circuit board], an intermediate solution that adds another camera for some additional inspection steps, and a deluxe level with many cameras,” Perelli said. “If you have a PC-based solution, you need the largest, highest-performance PC to handle the deluxe level.” But you will also use that same powerful PC on the simplest level, so as an entry-level solution, it would cost you more and have higher overhead than if you just used two smart cameras for your entry-level system. “With a system based on smart cameras, you only need to add one camera to increase your level of inspection, so your cost increment is more linear,” explained Perelli.
Smart cameras have gotten easier to network, as many of them now feature Ethernet or IEEE 1394 interfaces. In addition, the proliferation of open standards has helped foster more acceptance of smart cameras on the factory floor, said Subramaniam of Cognex. “Most of our smart cameras can take advantage of common protocols such as DeviceNet, Profibus, OPC [OLE for Process Control], and ActiveX.”
A recent trend is the development of open-source smart cameras, such as Supercomputing Systems’ smart camera based on open-source hardware and software (Ref. 1). An open-source architecture gives users the flexibility to extend and modify the base behavior of the software package, said Subramaniam. “Open-source systems that leverage customization of the software package are usually most appropriate for applications where hundreds of systems can be programmed to do the same task, so the software development support and change management costs of customization can be amortized effectively. But for the general smart-camera marketplace, the open-source model is still a small niche.”
 Developers can transfer their own algorithms to Linux-based, open-source intelligent cameras such as this one from VRmagic. Courtesy of Stemmer Imaging and VRmagic. |
Open-source smart cameras have definite uses, mainly for big OEMs building their own inspection systems that will be used many times for the same high-volume applications, said Gailer. Engineers do need to know how to program them. “But once you’ve done all that—the programming and all the development work—then you have a relatively inexpensive system,” he said.
The smaller real estate of a smart camera is one benefit for system integrators, said Vision Components’ Toth. Others include reduced cabling and the concomitant cost savings, less hardware complexity, less cooling, and only one unit to change during upgrading or troubleshooting. Smart cameras are definitely getting smaller and integrating more functions. “An even stronger trend is the same size camera containing even higher-resolution sensors, more memory, and increased computing power,” he said.
Because of heat-dissipation issues, however, smaller cameras are not always desirable, said Gailer of Stemmer. In semiconductor inspection applications, many customers are still using old-fashioned analog cameras because they do the job and there’s no need to change the technology. On the other hand, Gailer said, “Windows-based PC systems are becoming so small, such as the handheld ultra-mobile PCs, that you could connect a dumb camera to one of them with a cable, and you would have exactly the same solution as a smart camera: It would be the same size, it would cost less because of the high volumes, and the only additional cost would be the cable.”
Read more from ourAugust 2009 Machine-Vision & Inspection Test Report.
| Reference |
| Thryft, Ann R., “Smart cameras gain open-source platform,” Test & Measurement World, January 26, 2009. www.tmworld.com/article/CA6632153.html. |
No related content found.
- 0 rated items found.
Datasheets.com Electronic Parts & Inventory Search
185 million searchable parts
- Part Number
- Description
- Inventory
- Products
- Manufacturers
View All Webcasts
