Testing Time CompressorBy: Ethan Ace, Published on Jan 31, 2014
Our VMS Search Shootout showed that performance varies greatly and is an important differentiator in product selection.
One uncommon but interesting feature is the ability to overlay people and vehicles from different times on a single video to quickly show who and what passed by in a scene. Axxon's VMS has this, calling it Time Compressor. Here's their very short demo:
It's free and built into their VMS, so we decided to test it to better understand how well it performed.
Here are our key findings from this test:
- Time Compressor was simple to use, with little configuration required and minimal control over number of simultaneous objects shown on screen (maximum 6).
- Users have no control over exact time range compressed, only number of objects displayed on screen. A section of video (such as 12am-6am) cannot be selected and compressed explicitly. Objects simply appear in order of detection.
- In high traffic areas, objects overlaid on top of each other may be very confusing to view, especially for new users. Adjusting number of objects on screen (minimum of 2, maximum of 6) lower helps to reduce this visual clutter, but reduces time gains using Time Compressor.
- Since Time Compressor relies on Axxon Next's built-in analytics to operate, inaccuracies should be expected. For example, a single object may be detected as two objects, headlights or shadows detected as objects, etc. This results in inaccuracies and may make search difficult.
Future enhancement: Axxon reports that the next major version of their software (4.0), due in the second half of 2014 will increase maximum objects displayed from 6 to 20.
Time Compressor may save moderate time in searching for archived video, especially in scenes with infrequent motion. However, analytics must be properly configured (often by trial and error) for best performance, and object overlap and false detections may reduce benefits in many scenes.
Most importantly, since it cannot compress over large periods of time, it still requires scanning through video manually to find objects of interest, limiting is application significantly.
This video provides an overview of playback, with and without time compressor:
This video reviews issues with using Time Compressor, generally due to analytic detection performance:
Small Scene Performance
In this screencast, we review potential drawbacks using Time Compressor in small spaces, where motion may be higher and angles not ideal:
Finally, we briefly review the few configuration steps required for Time Compressor in this screencast:
There are few options which compress video similarly to Time Compressor. The standout comparison is BriefCam [link no longer available], whose Video Synopsis is similar in functionality, though with more configurable options, including specifying time range, object direction, and more. However, BriefCam is a third-party add-on requiring a separate license and integration to the user's VMS, not built-in as Time Compressor is.