Smart camera vendors are leveraging embedded processors and on-board software to increase the ease of use of their products
Andrew Wilson, Editor
Leveraging the developments in digital signal processors (DSPs), RISC-based processors, and low-cost CPUs, smart camera vendors are now offering relatively low-cost products that can replace host-based machine-vision systems. Embedding image capture, processing capability, on-board memory, and standard camera interfaces, these autonomous networked smart camera systems dramatically reduce the final cost of machine-vision and image-processing systems.
In the development of smart cameras, vendors have recognized the need to provide system integrators with products that can be easily tailored for specific machine-vision applications. Encompassing a number of different image sensors, processors, and I/O interfaces, many of today’s smart cameras are bundled with software packages that range from basic compiler support to sophisticated machine-vision software capable of performing functions such as geometric pattern matching and color analysis.
Although currently available smart cameras can be differentiated by their hardware functions, the choice of which camera to purchase may be more dependent on the software that is offered by each manufacturer. Cameras can be classified as those that offer basic software development tools and low-level development libraries, those that can be configured to perform machine-vision tasks such as gauging and image analysis, and those that are offered with full software development packages. The sophistication of the developer, time to market, and cost of deploying the systems is of paramount importance.
Programmers who wish to implement relatively low-cost smart camera systems may only require products such as the leanXcam smart camera from Supercomputing Systems, a Blackfin DSP-based camera that runs under μClinux. The camera is supplied with the freely available Open-Source Camera (OSCar) Software Framework that provides a library of programs together with source code.
Although low-cost hardware may be more difficult to program, it does allow system developers to build more sophisticated machine-vision cameras. FiberVision, for example, offers a number of smart cameras based around camera modules from Vision Components. In its latest Caminax smart camera, FiberVision’s software allows developers to graphically edit machine-vision programs without programming or using script code (see Fig. 1). In addition to allowing up to 250 regions of interest to be monitored, the camera’s on-board software allows programs to be configured as images are acquired.
To rapidly configure machine-vision applications more easily, companies such as PPT Vision and Soliton Technologies offer software developers graphical development tools with their smart cameras. At present these software tools can only be used with each company’s smart camera, forcing system integrators into adopting a single development environment when developing a machine-vision system.
Despite this limitation, the software tools do supply developers with an easy method to deploy machine-vision functions. Included in PPT Vision’s IMPACT Software Suite, for instance, the company’s IMPACT Vision Program Manager (VPM) features approximately 120 tools including OCR, blob analysis, circle gauge, circular pattern find, line find, and subpixel gauging. A preconfigured operator panel is built within VPM so inspection data and pass/fail results can be displayed within the same software application.
For its Digital Spot-It smart camera, Soliton supplies software that enables the developer to develop custom image-processing routines without programming. The company’s Vision Artist (SVA) package allows system integrators to develop an image-processing script that includes image acquisition, image-processing functions, and I/O control (see Fig. 2). Using SVA, various image-processing algorithms can be applied to the images acquired from Digital Spot-It and tested to ensure that the script produces the correct results for each of the test images.
Time to market
The trend toward faster time to market has also been recognized by those companies that have traditionally been associated with offering sophisticated machine-vision and image-processing software. Both Matrox Imaging and National Instruments, for example, have tailored their software as easy to configure GUI-based development packages that can be used with their smart camera offerings.
With the introduction of its Iris GT smart camera, Matrox offers an integrated development environment (IDE) called the Matrox Design Assistant, in which machine-vision applications are created by constructing a flowchart instead of writing code. This IDE also enables users to directly design a graphical operator interface to the application (see Fig. 3).
FIGURE 3. For its Iris GT smart camera, Matrox offers an IDE called the Matrox Design Assistant in which machine-vision applications are created by constructing a flowchart instead of writing code.
Similarly, for its range of smart cameras, NI offers its Vision Builder for Automated Inspection (AI) software, a configurable machine-vision development environment that requires no programming. With Vision Builder AI, developers can build machine-vision applications without the use of a programming language. Interestingly, both the Matrox Image Library (MIL) and NI LabView are also available on other smart cameras, most notably those from Sony Electronics.
Companies that in the past only offered software products for use with their own smart cameras are now beginning to recognize the importance of establishing a broad base for their products. Acknowledging that time to market is important, companies such as Cognex have begun strategic alliances with camera manufacturers.
Started in January this year, the Cognex alliance now has 10 members including Allied Vision Technologies, DALSA, Point Grey Research, and Sony. While many of these companies do not (yet) offer smart cameras, Sony now offers VisionPro from Cognex as an option on its XCI series.
Capitalizing on the growing trend toward embedding software in smart camera products, companies that presently only develop machine-vision software have also ported their software to smart cameras. Halcon 9.0 software from MVTec Software also runs on Sony’s XCI-V100/C and XCI-SX100/C and, by using Windows XP Embedded in the camera, application development can be performed either in the smart camera or on a PC. This can be done using MVTec’s Integrated Development Environment (IDE) that allows developers to build image-processing solutions fast while acquiring images (see Fig. 4).
In addition to running on Sony smart cameras, Halcon is supported on the eXcite camera from Basler Vision Technologies, the SM2-D1024-80/VisionCam PS intelligent camera from Photonfocus, and the eneo SC Series from Videor. MVTec has also performed initial performance measurements of Halcon 9.0 running on smart cameras from Vision Components.
Despite the large number of cameras available, the number of companies currently offering smart camera products is still relatively small. Perhaps the main reason for this is the cost of embedding sensors, FPGAs, on-board processors, memory, and interfaces into standalone systems. However, with miniaturization on the upswing, it will become increasingly easy for camera manufacturers to develop ever smaller smart cameras at lower cost.
To compete, companies that currently support limited software capability with their camera offerings may be forced to develop more sophisticated programs. Those companies that only offer machine-vision software on their own camera products may also move toward deploying this software across cameras from multiple vendors. Likely as not, a meeting of both of these approaches will occur, providing a boon for software developers and system integrators alike.
Allied Vision Technologies Stadtroda, Germany
Basler Vision Technologies, Ahrensburg, Germany
Cognex, Natick, MA, USA
DALSA, Waterloo, ON, Canada
FiberVision, Würselen, Germany www.fibervision.de
Dorval, QC, Canada
Austin, TX, USA
Photonfocus, Lachen, Switzerland www.photonfocus.com
Point Grey Research
Richmond, BC, Canada
PPT Vision, Bloomington, MN, USA www.pptvision.com
Park Ridge, NJ, USA
Videor, Rödermark, Germany