Avigilon Analytic Cameras Tested 2014By Ethan Ace, Published Nov 12, 2014, 12:00am EST (Research)
Analytics remains the 'next big thing'
But supply of high quality, ease to use analytics remain in short supply.
Now, building on VideoIQ, Avigilon has released their first cameras with built in analytics, the H3A bullet and dome series.
We bought these new analytic cameras and tested in two common long range outdoor scenes to find out how well Avigilon analytics work.
In this report we answer the following key questions:
- How do these analytics stack up against non-analytic Avigilon cameras and third party cameras, like Hikvision, connected to the Rialto appliance?
- Do they beat VideoIQ's previous iCVR-HD models?
- How well do these analytics integrate with Avigilon Control Center?
- How well do these analytics integrate with rival VMSes, like Exacq and Milestone?
Here are our key findings from this test:
- Human and vehicle detection distances and scene width for the new Avigilon analytic camera were greater than other cameras tested (older VideoIQ ICVR, the non analytic Avigilon bullet and the Hikvision camera attached to Rialto I4) by 10 to 50% depending on scene.
- Detection issues found in previous VIQ generations (iCVR and Rialto) such as greatly decreased range when human subjects approach the camera and headlights/taillights causing high missed detection rates are still present in new Avigilon Analytic cameras.
- Avigilon analytic events integrated only with Avigilon Control Center, unlike previous VideoIQ models. Avigilon claims analytics may be integrated with third party systems via ONVIF (Profile S, version 2.4),
- However, no analytic events were received in our testing of Exacq Professional and Milestone XProtect Enterprise via ONVIF. We were able to connect, streamed and configure for basic settings.
- Avigilon Control Center 5.4 moves analytic setup functions to ACC, simplifying setup of cameras. Configuration previously required Avigilon/VideoIQ View.
Rialto/ACC Integration Updates
- Rialto configuration process changed completely in ACC 5.4, with setup now performed via the ACC Client, and cameras associated with Rialto channels, not added separately through View.
- Many cameras which previously functioned in the Rialto IR are now unsupported when associated through ACC Client. Only Avigilon and Hikvision cameras added without issue in our tests.
- Cameras added to the Rialto via Avigilon/VideoIQ View are not automatically populated in ACC, meaning existing Rialto deployments must be defaulted and reconfigured using the new method.
Estimated street price of the Avigilon 3.0W-H3A-BO1 is ~$1,100 USD, far lower than the previous generation VideoIQ iCVR-HD, which had an estimated street price of about $1,900. Note that the H3A bullet does not include the on board storage of the iCVR-HD, and no iCVR models included built-in IR.
This pricing is also only about ~$100-$200 more than a non-analytic version of the 3MP bullet (3.0W-H3-BO1, $1,035 MSRP).
Finally, varifocal bullet 1080p/3MP IR cameras average ~$700 USD. The 3.0W-H3A-BO1 is notably more expensive than that, but built-in advanced analytics are extremely rare and would, otherwise, require, purchasing third party analytic software, etc.
Avigilon's H3A analytic cameras improve performance compared to past VideoIQ generations and third party cameras attached to the Rialto, at a lower price point, and with simplifed setup in ACC 5.4, making them attractive to those seeking video analytics.
However, unlike past models, these cameras' on board analytics are not integrated with third party VMS systems, making them usable only in end-to-end Avigilon systems. Those seeking a replacement for VideoIQ in their open platform VMS systems must look to other options. Given that Avigilon is supporting ONVIF Profile S including analytic events, if VMSes choose to add support for this optional ONVIF feature, integration is potentially possible.
The improved performance and lower price of the new Avigilon analytic camera over the previous 'Favorite' VideoIQ analytics, and their tight integration, increases Avigilon's overall competitive positioning.
If, and when, 3rd party VMSes, add support for Avigilon analytic cameras, it will become an even better choice than VideoIQ for analytics, especially in a weak overall analytic market. However, how soon and how wide that integration occurs remains to be seen. For example, Exacq broke basic VMD support for Avigilon cameras, so that bodes poorly for Exacq adding advanced analytic support for Avigilon.
Compared to VideoIQ iCVR HD
In most instances, the Avigilon 3.0W-H3A-BO1 performed moderately better than the VideoIQ iCVR-HD which we previously tested in the same scenes. Nighttime performance increases are at least in part due to the H3A camera's integrated IR, as low light performance in the iCVR-HD was poor.
The iCVR-HD outperformed the H3A bullet in only one test: daytime detection of a human subject in the parking lot. In all others its detection range was lower.
Unlike past VideoIQ analytic cameras, the Avigilon H3A line's analytic rules are configured completely through Avigilon Control Center Client (starting in version 5.4). Rule setup is also simplified, with only the most common options shown by default (intrusion, loitering, and line crossing). Other rules, timeouts, and further adjustments are found in the Advanced menu.
We tested the cameras and Rialto in two test scenes representative of common applications, an open field with vegetation on multiple sides, ~500' long, and a parking lot and vehicle entrance, ~450' long.
All cameras were set to the same horizontal AoV of 60°.
Open Field Range
This overhead view shows approximate ranges from camera to target in the field.
We first tested in daylight with some clouds, as seen in this sample image.
In this scene, the Avigilon H3A analytic camera outperformed both cameras attached to the Rialto both when the subject was crossing the scene, and as he approached the cameras. The analytic camera had by far the best approaching performance of cameras tested.
Note that this crossing range was equal to the top cameras in our previous Rialto test, the Bosch NBN-932V, and slightly further than the old generation VideoIQ iCVR HD (~350' range) which were tested in approximately the same FOV.
At night, performance dropped significantly due to the low light levels in the field, ~5 lux in the near field of view, down to about 0.05 lux at the end of the field.
IR illumination in all cameras compensated somewhat it short to medium ranges, but was unable to illuminate far enough to maintain daytime ranges. The H3A bullet still outperformed other cameras in this scene, though crossing range was less than 50% of its daytime measurement.
Next, we moved to the parking lot/entrance scene. Approximate distances are shown in this overhead view:
This field of view shows the scene, with human and vehicle subjects in the parking lot, and vehicles passing by on roadways adjacent to the building.
Again, the Avigilon analytic camera outperformed other cameras in all tests, reliably detecting crossing humans out to about 315', and approaching at about 190'. The H3A model also reliably triggered on vehicles on the highway beyond the scene, about 600' range. Approaching vehicle distance was at least 450', though we could not test beyond this due to the limited test area size.
Finally, we tested in this scene at night. Light levels were about 1-5 lux due to the building's outdoor lighting.
Again, crossing human detection range decreased by 50% or more in this scene. The same was true of crossing vehicles. Approaching human range in all cameras was approximately similar.
Vehicles moving toward and away from the cameras were very difficult to detect due to the severe blooming effects of headlights and taillights, an issue we first noticed in our test of the Rialto. No camera reliably detected vehicles at any range when moving toward or away, noted as a star in the chart below.
The physical construction of Avigilon bullet cameras with on board analytics is the same as non-analytic versions. However, note that on current models of all form factors, an SD card slot has been added next to the service port (seen below), which was not found in the 2.0W-H3-BO models we previously tested.
Note that this storage does not integrate with ACC, though may be set up to record full time, on event, or when the camera loses connection to the server. Recordings must be retrieved through the web interface or by removing the SD card.
All devices were tested at default settings except for shutter speed which was standardized to 1/30s.
All devices tested with the same horizontal AoV at 60°, such that the scene width was always the same.
All devices were tested using current firmware:
- Avigilon 3.0W-H3A-BO1: 126.96.36.199
- Avigilon 2.0W-H3-BO1: 188.8.131.52
- Avigilon Rialto I4: 184.108.40.206392
- Hikvision DS-2CD4132FWD-E: 5.2.0 build 140721
Avigilon Control Center 220.127.116.11 was used for testing.
13 reports cite this report:
Back to Top