Pixel Motion Detection Vs. Video Analytics Outdoors

Is anyone aware of a study comparing the results for the outdoor night environment with IR lighting between pixel based motion detection and lets say Videoiq video analytics, same view, same size of ROI, over an extended time period say >10 days?

I am thinking of comparison for different weather conditions (eg. clear/cloudy, raining, snow flurries)

Pixel based motion has usually several sensitivity settings (eg. 1~6)

Videoiq analytics has settings for sensitivity (1~10) and duration (0~5 sec)

Comparison of Results between detection methods could be frequency of false positives, false negatives, true positives. Lots to compare I know, but I haven't found anything.


Robert, There are so many variables that a realistic study could not be completed. Let's take the obvious of indoor versus outdoor. What lighting and at what distances? Will there be lightening like in Tucson, snow flurries like Colorado, horizontal rain like Hawaii? Are we talking just detection? How many pixels and what sensitivity? Johnson criteria? Are we looking for which would filter more FAR/NAR versus PD? Pixel detection works for enhancing recording and such but not alerting without complex algorithms. Essentially all Analytics is pixel detection with additional intelligence.

Robert, we are working up to that test in the near future. We have already done VMD shootout and camera analytic shootout, with upcoming ones to compare vs 'real' video analytics.

I would focus less on the sensitivity settings as they are not directly comparable across manufacturers - an '8' on Manufacturer A may be much different than '4' on Manufacturer B.

Robert, one important question is how long / far / wide do you need to see? One thing our tests so far already show is that range if a big differentiator.

Hi John:

What prompted me to ask is that I set up 4 cameras with videoiq analytics and they are set to view 20~150' with a bosch flexidome 3~10mm and 150'~350' with 6~50mm lens. I happened to also have the same video outputs fed into a Hikvision DVR and I was curious how pixel detection actually behaves in an outdoor environment so I compared them overnight 6pm~6am last week (raining/overcast) and the pixel detection was going off every 3 seconds, while the videoiq was maybe every 15 min.

I guess no one has actually done any test comparisons between pixel detection and analytics because pixel detection is unsuitable for the outdoors?

Rob

Robert, what device was doing the pixel detection - the Hikvision DVR? If so, it might be an issue with the Hikvision's motion detection (it could be bad or possibly need to be reconfigured).

Tomorrow, we are releasing a test report on 'pixel' / crossline detection in the rain and none of the three cameras tested (Axis, Bosch, Sony) had a false positive problem. I'll add a link to it when it's released.

I was using the DVR, it breaks the scene into a grid (~15 x 15). Frankly I don't know much about how the DVR uses the grid and how it interacts with the choice of sensitivity.