The explosion in the availability of low-cost, highly functional video cameras has provided many engineers, marketers, researchers, and others a whole new way to collect data. Capturing an event, person of interest, or machine performance on video allows one not only to collect much more information than with a small number of "simpler" sensors; but it also enables easy visualization of what has been collected. While the camera may seem like an ideal data collection tool, the problem faced by many is: how do I get the measurements or useful data out of the moving video images? That is, how do I translate what I see onscreen into numerical graphs that I can quantitatively analyze and understand? In this manner, for example, comparisons between computer simulations of and actual dynamic events can be made and the fidelity of the models improved.
Quantum Signal has been developing technology to address these needs. We have developed a host of algorithms and software tools to allows one to specify, detect, and track objects in video and extract motion and related data. These techniques have been applied to extract the motion of objects in controlled and uncontrolled environments, such as crash-test videos and street-corner surveillance, and adapted to a wide variety of individual customers' needs. The capabilities break down into two key categories:
- Tracking with fiducials - technology to automatically detect, track, and extract data from moving objects by leveraging specifically placed fiducial markers in areas of interest. Such markers are common in automotive engineering (for crash tests) and in the film and gaming industries (for facial animation capture). Examples of such markers are those placed on vehicles for crash tests, a graphic or warning sign on the side of a moving truck, or a UPC on the side of a cereal box. As long as the marker is not subject to drastic appearance changes in the video, it can be tracked with these technologies.
- "Free form" Tracking - technology that allows the user to graphically select, via point-and-click, unmarked areas of objects for tracking. For instance, the window area or tire of a vehicle can be tracked in this manner. Based on user input, the software intelligently creates an "area" and tracks the motion of that area over the time course of the video. This is an extremely powerful capability that allows even the novice user to become a video tracking expert!
The combination of the two tracking technology bases yields a powerful tool to extract useful numerical data from your video database. The tracking technology is highly robust, and works with videos and films of various types and qualities.
If you have video or are thinking of video as a data collection method, contact us today! We know video, and are here to help!