Hikvision Video Analytics Tested 2016By Ethan Ace, Published Feb 04, 2016, 12:00am EST (Research)
We tested Hikvision's video analytics on six Hikvision models, ranging from entry level to high end Hikvision cameras, in this outdoor test scene:
Hikvision's detection performance and false alarm rejection were solid when configured for the scene, though poor using default settings. Performance of mid range and high end cameras was similar to or better than Axis VMD 3, though behind Avigilon's integrated analytic cameras.
Low end cameras and TVI DVRs offered the smallest coverage ranges of devices tested, though provided low false alarm rates similar to higher end models.
Those selecting Hikvision's video analytics should be sure to adjust time threshold and region of interest properly for their scene, as failing to do so will likely lead to numerous nuisance alerts. Users should be careful when adjusting sensitivity as well, as even slight adjustments may significantly increase false alerts or decrease range.
The inclusion of these on-board analytics at no charge on even low cost models (though limited to intrusion/crossline) helps Hikvision's positioning as more than simply a cheap camera brand. This brings them in line with Bosch, Sony, Samsung, and others who have offered basic analytics on their cameras for some time, and puts them ahead of Axis, who offers only VMD 3/2.1 (advanced motion/intrusion) and crossline analytics (though at an additional ~$30 fee per camera).
This should also help Hikvision in spec situations where these analytics are included as a checkbox item.
Here are our key findings from this test of Hikvision's on board analytics:
False Alarm Issues
- On all cameras, Hikvision's analytics were prone to numerous false activations due to moving branches and vegetation, but not shadows or moving lights (headlights, flashlights, flashes, etc.)
- Excluding vegetation from the region of interest eliminated nearly all false activations from these sources.
- No false activations were triggered in the 6026 and 4024 cameras overnight, similar to Avigilon's H3A analytic bullet, while Axis VMD 3 triggered three times.
- ROIs are less configurable than competitive analytics due to only four port inclusion zones instead of polygons, and the lack of exclude zones.
Detection coverage varied by camera. This chart outlines coverage for Hikvision and competitors:
Of Hikvision's cameras tested, the 4132 integrated IR dome and DarkFighter 1/2" super low light model performed best, with the largest coverage area, with the 4132 one of the top performing daytime cameras.
However, the 6026 outperformed all other cameras tested at night, reliably detecting subjects across a 215' HFOV/185' range, versus 140'/110' in Avigilon, the next best camera.
- Analytic events integrated with Genetec Security Center, but not Avigilon Control Center, ExacqVision, or Milestone XProtect. No bounding boxes are embedded in video using third-party VMSes.
- Events displayed properly in Hikvision's iVMS-4200 client, with ROI and bounding boxes overlaid on video streams.
The majority of Hikvision's current camera and recorder line includes intrusion detection and tripwire analytics at no cost. Higher end series include additional analytics, as well, shown in this chart:
There are two limitations to configuration which make detection zone setup more difficult than in some competitive models:
- Four sided regions only: Include regions may have only four sides/points, making odd-shaped regions difficult to draw or requiring multiple regions.
- No exclusion zones: The Hikvision UI does not include exclusion zones, only include, requiring users to use multiple inclusion zones to cover a full screen FOV with objects they wish to exclude.
Based on our tests, we recommend the following settings be used:
- Sensitivity 50: Cameras default to a sensitivity of 50 on all analytic rules, and we found this to be the best combination of range and false alarm rejection. Lowering sensitivity quickly reduces range, while raising it increases false activations.
- Threshold >1 sec: In the vast majority of applications, objects will be present in the region of interest for greater than 1-2 seconds. Raising threshold to 1 or higher eliminates many false activations (typically short-lived due to foliage movements) while not impacting accuracy. Threshold defaults to 0, meaning rules are triggered instantaneously.
We review these issues and more in this video:
False Alarm Issues
Using default settings, Hikvision's analytics suffered from high amounts of false activations from two sources:
The biggest source of nuisance alerts by far was moving branches/foliage. With even a slight breeze, branches triggered intrusion detection rules near constantly. Most of these alerts could be filtered by increasing threshold to 1 or higher, though for best results, these areas should be removed from the region of interest.
During the day, shadows cast by branches, overhead wires, or sudden changes in cloud cover caused some false alerts, though at a much lower rate than vegetation.
Analytic coverage area varied depending on the Hikvision camera model. The 4000 series (4024 and 4132 in our test) performed best in daytime scenes, with coverage range similar to Avigilon's H3A analytic camera, about a 290' HFOV. The DarkFighter 6026's daytime coverage was smaller, ~240' HFOV, similar to the Axis Q1615 with VMD 3.
However, the 6026's nighttime range was the best of all cameras tested, covering a ~215' horizontal FOV, only moderately smaller than its daytime range. Other cameras experienced significant drops in coverage in low light.
Low End Model Coverage
Coverage areas of the 2000 Value series and TVI cameras were much smaller compared to the 4000 and 6000 series. During the day,
At night, coverage dropped significantly,
In our tests, Hikvision's analytics integrated with Genetec Security Center, but not Milestone XProtect, Avigilon Control Center, or ExacqVision Professional. Bounding boxes around objects are not embedded in video in any of these VMSes.
Integration with Genetec requires no configuration. Events are simply passed from the camera to the VMS, and display in the event list.
Hikvision's analytics integrate with their own iVMS-4200 CMS, as well. Event data is displayed, as well as regions of interest and object bounding boxes, not found in third party VMSes.
2 reports cite this report:
Back to Top