Motion Analysis Shifts Into Digital Domain


by Jack Vaughan, Automotive Test Report
June 1999

Motion analysis, like most other aspects of automotive test, has undergone a significant shift and moved into the digital domain. This has significantly altered the nature of many automotive test methods, and test engineers find that they must grapple with a host of new issues even as they say goodbye to old ones.

In a variety of tests, ranging from crash sled and airbag deployment tests to engine compartment flame propagation studies, digital video cameras are replacing 16-mm film cameras. Moreover, miniature digital cameras make it possible to view events in almost any automotive nook or cranny.

There are many advantages to using digital video. Sometimes the advantage is lower cost; sometimes it is fast test turnaround. Even so, even digital proponents caution you to temper your enthusiasm and be aware of some of the drawbacks.

For example, one of the advantages of using video methods is that you do not need trained photographic personnel at the test site. Digital cameras are in many ways quite similar to the familiar 8-mm home-video cameras, and most people can operate them. The problem is that most people don’t have the experience necessary to properly frame a scene. Digital photography is not as complex as 16-mm photography, but you still must adhere to basic photographic principles to capture usable images.

“Back in the 1980s, high-speed photography was used extensively,” says Chris Balch, engineer at Kodak (San Diego, CA). “I look at that as a black art [in that] you have to have an experienced photographer who knows light settings, field views, and so forth, for the camera. The film can take days to develop. That made the cycle of test and review much slower.”

Balch adds that the immediate feedback that just-acquired images provide becomes a very powerful new means to help you understand the problem you are looking at. Almost any type of test can be enhanced by the ability to immediately test out the soundness of test setup itself.

Balch continues: “The value people are finding with high-speed video is that it is immediate.” This may be the case even in crash sled or other tests that rely significantly on film as well as video.

“Production engineers who want to resolve something on a seatbelt test, for example, can review the video and modify the next firing of the sled so that they can get better quality data and converge faster to solve the problem they are working on.”

“You don’t have to have special skills in setting the camera up,” he says. “You see through the monitor what you’d see you see through the lens.”

Balch points out a number of things that engineers should consider when preparing tests. “You have to look first at what type of motion you want to capture and how fast that motion is. If you don’t have an idea of this ahead of time, than it really is case of trial and error.”

Issues to consider are frame rate, sensor dimensions, depth-of-field, resolution, lighting techniques, and color, among others. “If the image has too much blur, you have to increase the frame rate or increase the shutter speed of the system, and there are depth-of-field issues to consider,” he said.

The test team must treat lighting as a design element, and relate lighting to the event it is studying, points out Jeff Yorsz, assistant general manager at Adaptive Optics Associates (Cambridge, MA). Lighting is often the most important test element in visual motion analysis. Sums up Yorsz: “If you can’t see it, you can’t analyze it.”

John Capellie, manager of sales and marketing at Sensors Applications (Utica, NY), emphasizes the advantage of rapid turnaround. “If you are doing multiple, repetitive tests, it’s nice to be able to fix problems right away, not have to wait two days to find out there was a mistake in the test setup.”

Visual motion analysis is often successful as a supplement rather than as a replacement. The ability to relate visual data to other data (for example, sensor data) is a plus, says Corey Miller, president, Microsys Technologies (Concord, Ontario, Canada). “You measure something visually and correlate it to events [measured using other means]. You can see what is happening in the visual world and the real world.”

Storage Needs

Knowing ahead of time how much memory to allot for image capture is "a difficult thing for customers to grasp," says Kodak’s Balch. The problem naturally grows as testers opt to deploy multiple cameras, the output of which must be correlated correctly.

"Each digital imager, if it’s recording at 1,000 frames per second, and the image is sized at 512x384 pixels, is going to record 200 Mbytes in 1 second. If you’re doing this over and over all day long, it can fill a hard drive pretty fast," says Balch.

"People are not sure how to deal with so many digital images and how to archive them. Some are using CD-ROMs, some are using digital video tape, and some are looking at DVD to act as a storage medium."

Of course, the high-speed films of crash data are so important that an infrastructure is already built up to properly archive them. The films must be retained for many years to meet legal requirements. In crash tests and related tests, flat file CD-ROM storage appears to be the most prevalent method of motion analysis data storage used by individuals we talked with.

Storage, says John Capellie, is just now becoming an issue for many companies. "If you store the results of several tests a day, this can turn into terabytes of data over time."

"For example, if a facility has two test areas and ten cameras in each area running five tests per day for an average of 250 days per year, and they are getting 100 frames per second out of each of those cameras, that works out to about 4.47 Tbytes of data per year," says Capellie. The image data take up the most space, but when you add in sensor data of the same event, your archive problem grows. If you end up burning 1,500 CDs a year, that’s a lot of data to manage, he notes.

As testers using digital vision systems move to higher resolutions there is going to be an ever more enormous amount of data to move around and track. Filing systems must be redundant and structured so that all pertinent test elements are linked. Trying, for example, to correlate far-flung sensor and visual data months after a test can be time-consuming if plans weren’t carefully made prior to testing. Files and non-relational database storage methods are giving way in some cases to relational database storage means. Even advanced object databases are being tried in some special instances. All these methods require development of new skill sets among testers.

Jack Vaughan is a technical journalist in Boston, MA. You can reach him at jvaughan@ix.netcom.com.


  Printer friendly version

For further information contact:

Marketing Manager
Microsys Technologies Inc
3710 Nashua Dr, Mississauga, ON, L4V 1M5 Canada

tel:  +1 (905) 678-3288
fax: +1 (905) 678-3319
email: marketing@micro-sys.com


Head Office: 3710 Nashua Drive, Mississauga ON L4V 1M5, Canada
Telephone: +1 (905) 678-3288 | Fax: +1 (905) 678-3319

© Microsys Technologies Inc. All rights reserved.