Vision applications highlight NIWeek
More than 140 authors from nine countries submitted 77 papers to NIWeek’s Virtual Instrumentation technical-paper contest. During the conference, 29 category winners and finalists presented their papers at user sessions. Addressing industries as diverse as aerospace, automotive, communications, control design, and semiconductor inspection, many of the papers showed how to combine electronics test, motion control, and machine vision using hardware from National Instruments (Austin, TX, USA; www.ni.com).
Taking the prize for the best overall paper, Ganesh Devaraj, managing director of Soliton Technologies (Bangalore, India; www.solitontech.com), and his colleagues discussed an automated optical inspection system for detecting, quantifying, and classifying surface defects on x-ray ceramic scintillator elements. Designed to identify and classify spots, cracks, and voids, the inspection system uses a pneumatic slide to move the scintillator into an imaging chamber. After the pneumatic slide positions the device in the chamber, a 1280 * 960-pixel FireWire camera from Sony (Park Ridge, NJ, USA; www.sony.com/videocameras) images the device.
Soliton Technologies optical inspection system for classifying surface defects on x- ray ceramic scintillator elements uses a FireWire camera interfaced directly to a PC. A PCI-6503 add-in board with 24 channels of digital I/O interfaced to two SC-2062 electromechanical relays controls both lighting and the pneumatic slide used to position the scintillator elements in a test fixture.
Because the application required different defect types to be identified and classified, Devaraj and his colleagues used NI’s Vision Assistant in the system’s prototype to identify image-processing parameters that best differentiated each kind of defect. To control both the lighting and the pneumatic slide, Soliton used a PCI-6503 add-in board with 24 channels of digital I/O interfaced to two SC-2062 electromechanical relays. “During the prototype stage,” says Devaraj, “we developed image-processing algorithms that were translated into production-ready application software, and we developed an HMI that allows the user to interact with the digital controls for sequencing the inspection process and generates reports and e-mails. According to Devaraj, all of this was completed within 16 weeks.
In addition to winning the overall best application award at NIWeek, Soliton’s system for automotive instrument cluster inspection was judged the best application in the automotive category. “Our customer’s Japanese collaborator was already producing and testing these clusters, and hence, our customer had an option of purchasing the already-developed test system,” says Devaraj. This used a single camera with high-speed motion-control hardware to inspect four gauges. “Our design used multiple industrial FireWire cameras with no motion-control elements in which each gauge could be tested in parallel to reduce the cycle time,” he says.
FireWire takes the lead
Interestingly, of the numerous vision-related systems described at the conference, most used FireWire-based cameras to perform image acquisition. These included the development of a remote telepathology microscope from Carinthia Technical Institute (Spittal, Austria; www.fh-kaernten.at) and a system for measuring the dimensions and flatness of steel plate from Joanneum Research (Graz, Austria; www.joanneum.ac.at).
“Clinics without a pathology department often require systems that allow the distance control of computer-based microscopes,” says Thomas Klinger of Carinthia. “But microscope images captured by digital cameras are usually transferred over low-bandwidth networks resulting in image-transfer times of one minute or more.” To overcome these limitations, Klinger and his colleagues used LabVIEW to simulate a remotely controlled microscope that transmits image data over a 4-Gbit/s dense wavelength-division multiplying connection between two clinics in Austria.
In the client-server-based system, LabVIEW transfers the entire screen content of the microscope-based server to the client. Virtual-instrument functions then allow microscope functions to be controlled through a serial link from the server using the remote client. “Although 4-Gbit/s speeds allow real-time remote microscope control, the system currently uses four ISDN lines and is, therefore, 16, 000 times slower,” says Klinger.
Developing a system that automatically measures the edge dimensions and flatness of steel plates using six FireWire cameras, Helmut Urban of Joanneum Research presented a system that measured plates of different curvatures and shape. Because the distance from the cameras to the plate edge surface can change as the plates move along the conveyor, it was necessary to compute the plate edge position for every measurement taken. To accomplish this, the cameras’ axes are oriented at right angles so that one lower camera supplies edge distance information to the upper camera. Computed distance information is then used to compensate for the change in scale factor as the plate moves closer or further away from the camera lens.
To document and store measured data, the system must also read barcode data. The variety in plate surface finish and design poses coupled with random field locations was solved with a brute-force solution, says Urban. National Instrumets IMAQ vision barcode-reading VI was used to search for the barcode rectangle by scanning the image line by line, column by column, until a valid code could be read. Although this simplistic approach consumes computing power, a 4-GHz Pentium processor coupled with LabVIEW built-in hyper-threading support was more than enough to accomplish the task.