Good question. Here's some high level differentiators to start:
Breadth of Analytics: AgentVi has the broadest array of analytic 'flavors' - from various outdoor alerting to people counting and searching
All in One System: VideoIQ is the most turnkey of the options as it bundles camera, hard drive/recorder, VMS and analytics all in a single device.
Running inside Third Party Cameras: Only AgentVi supports that (primarily Axis and IQ cameras) though VideoIQ now has an appliance/encoder that can connect to third party IP video streams (note: AgentVI allows support server side ingesting of third party streams). By contrast, Sightlogix's analytics only work on their own cameras.
GIS Target Tracking: Sightlogix is the only one who offers the ability to track target's physical location overlayed on the camera / map views.
Most Expensive: Sightlogix's products are clearly the most expensive in terms of unit cost ($2,000 or less for an AgentVi or VideoIQ analytics + camera, vs $5,000 or more for Sightlogix). But Sightlogix claims that overall cost is lower because they can cover much longer ranges (we've never tested so we cannot confirm or deny).
- Sightlogix is the most military / defense oriented option
- VideoIQ is the most turnkey option
- AgentVi is the most flexible option for different cameras and analytic flavors
All 3 of them are reputable, which is saying something in a space like video analytics.
I can't give you a mathematical value comparison, but SightLogix maintains its effectiveness over MUCH longer ranges than the other two. IPVM should definitely test this and try to quantify it.
FLIR Security | 06/15/13 07:48pm
Which of the 3 are the most immune to patent challenges from other analytics entities (or other players claiming to own patents)? 1. Do all 3 of those develop the analytics they use internally, or do they use 3rd party analytics (OV-style). 2. Even if developed internally, can 3rd party patent holders claim infringement?
(1) I don't know but, even worse case scenario, they will have to pay OV $5, $10, $20 per channel, which is not a huge deal for enterprise analytics.
(2) They all develop analytics internally.
(3) See point (1)
Brian, I can only go by what I've seen. If I'm wrong, I am glad to admit it.
Brian, can you elaborate on what models/combination of FLIR/VideoIQ products you believe matches Sightlogix? It would help to be specific so we can better understand if it's 'apples to apples', etc.
Having sold analytics for years, both simple and advanced, I would say it's all about the application and expectations. I sell thermal imaging which is almost always combined with analytics for perimeter detection and while it increases the cost and changes the value proposition it truly becomes a 24 hour solution. The added benefit is the reduction of undesired information the analytics have to filter and providing containment to the detection area. First and foremost manage the expectation and design to the requirements.
IPVMU Certified | 06/17/13 12:34pm
These comparison are very subject to how the camera analytic configuration is tuned. You really need to do your own field comparisons starting with the manufacturers defaults as a starting place. the algorithms used by different manufacturers MAY be totally different in terms of how they sense and track.
Basically boils down to doing the grunt work in the field to get really true comparision. Its like picking a stereo system. It requires you hear it first hand.
Think also of the phrase "suitability of use". Does the camera / analytic do the job you expect? You may have two (or more) quite capable manufacturers equipment which both do the job as expected. So it comes down asking yourself is the additional $$ / camera really going to be a benefit?
With regards to Agent VI specifically – can someone explain (40,000 ft overview) how they analyze the video? Don’t they do something different from the competitors in such that they peel the analytic data off of the video at the camera and send it to a SQL database as metadata? This vs. sending the raw video and processing it at a central server?
Thanks in advance.
[IPVM Note: Poster is from Agent Vi]
60,000 ft. overview:
Agent Vi uses a patented distributed image processing architecture to do VCA. The solution can be deployed in an "Optimized" configuration or in a "Flexible" configuration. In the "Optimized" configuration, instead of running the algorithm entirely at the edge or at the server, it is distributed so that a part of the processing is done at the edge and a part at the server. Without getting into too much technical detail this architecture allows us to overcome the inherent processing resource limitations in an edge device while maintaining a very high camera to server ratio (one CPU can handle 200 cameras). The software at the edge (the Vi-Agent) analyzes the raw video before compression and sends back a "feature stream" to the server where the analysis is completed. The feature stream has a proprietary format, is low on bandwidth (50-100kbps) and does not contain video images. The server analyses the feature streams from one or more cameras and converts them into metadata.
The "Optimized" configuration requires using a supported camera from one of our camera partners (Axis and IQ are two of them). When using cameras from vendors that do not support the “Optimized” configuration you can deploy our "Flexible" configuration. In this case the video needs to be streamed to the server, decoded and then analyzed to extract the metadata. This configuration allows you to use any camera but has a lower camera to server ratio (typically 20-40 per CPU).
Regardless of the "Optimized" or "Flexible" configuration, the created metadata can be used in one or more ways depending on the customer use case:
- Analyzed in real time to perform event detection
- Stored in a SQL database to enable intelligent forensic search
- Analyzed statistically for a variety of BI applications
Disclaimer, poster is Dir at VideoIQ.
Another key point is VideoIQ video analytics classify objects against a dbase of >500k images of people, vehicle and boats. You can set any of the 15 rules to respond to any movement, people, or vehicle. Or a combo, say people and vehicle. For outdoor deployments letting the wildlife cross your camera's FOV dramatically lowers the ongoing cost to run the system and lower lifetime TOC. The new Rialto line of IP, analog and Hybrid appliances are ideal for new or existing deployments. For remote, off grid, or network challenged deployments VideoIQ's ICVR-HD, all in one HD camera is an good answer. overall though, key to any system deployment's success is understanding what the goals of the customer / program.
You have a SightLogix NS340 (13 degree FOV) that you borrowed for your thermal camera testing. If you compare the SightLogix NS340 camera with a Flir SR-313 (also a 13 degree FOV), you will see that without nuisance alerts we detect on the order of two times the distance and four times the area of the other companies. We look forward to your testing.
Eric Heller, SightLogix
Best Analytics? Well it depends. :) Each seems to target a different market.
I've have evaluated an earlier SightLogix (visible spectrum) camera setup with a PTZ add-on. The ability to have the PTZ sync up to to the area of interest was pretty cool. The distance range was good, well within the specs SightLogix's posts. I'd think the product seems well suited for long relatively flat installations. An airport comes to mind. Hilly areas may mislead the GPS (guessing), thus the PTZ sync may not work as well as expected. Obstructed view may require more cameras, thus added expense (these cameras are expensive). The analytics was pretty good, headlights tended to trigger some false alarms at night. I suspect the IR version of the camera performs better with fewer false alarms. The resolution on the detecting camera is quite low, so you would likely want the PTZ to get a good view of the situation. My location couldn't take advantage of the long effective range of the camera, so I had to buy several more than anticipated. It just became too costly.
I have also evaluated a megapixel VideoIQ IP camera. The internal HD recorder was subpar (I think they have a newer model that improves the quality). Issue was with non-motion recordings defaulting to a super low resolution. Analytics was excellent, almost no false alarms (even with headlights). Nighttime image isn't as good as the Axis Lightfinder cameras, but the analytics still work! At the time VMS integration was weak. Sightlogix defiantly had a better effective range. I didn't test the IR versions of either camera. The analytics setup is a breeze (just let it sit for a few hours and you’re done - no measuring distances!), might be worth the price just for this feature alone. You still have to define zones and line crossing (just like all the others). The other systems take almost an hour to get the initial configuration, and several days before your done tweaking. This seems like the best general all-around analytics camera I've seen so far. You can make it work for just about any application, although the other might work better in specific applications. I really liked this camera, good price, almost bought one.
You did not mention ioimage (I think DVTEL bought them). They have an analytics system that includes a PTZ camera. It is able to detect an intruder and follow them around your facility as long as the camera has line of sight. During my evaluation, it had some issues operating near a busy street (some false alarms). I'd think it would work great in a low traffic industrial yard or dock area. The alarming and VMS integration was weak.
I've never tried Agent VI.
"The internal HD recorder was subpar (I think they have a newer model that improves the quality). Issue was with non-motion recordings defaulting to a super low resolution. "
FYI, this is configurable. The default setting is that continuous recording is generally much lower than HD, usually 960x540. Many systems do NO continuous recording, and record on "motion" only. We do continuous recording by default, but do it at this reduced resolution to maximize overall retention time, bumping up to full res only for the alarm events.
But, if you want, you can set it to record 1080p/30fps 24/7 if you want, it's just going to use about 5x as much storage (or, cut your total retention time down by ~80%).
Also, as it relates to range, I have to say that very rarely is maximum range a deciding factor unless you are doing very large outdoor perimeter areas. You are also going to run into more issues with potential atmospheric interference if you push optical cameras beyond about 500', and thermal cameras beyond about 2000'. Many times even with a very long perimeter other things in the area prevent you from getting long, flat straight shots of anything approaching 1000'+.
Regarding the specific SL camera mentioned above, their spec sheet says the NS340 has an automated detection range of 1115', if you want to cover that distance, I'd recommend the FLIR 612 core, not a FLIR 313 core based camera. Price-wise (camera, storage, VMS, etc.) I think you'd be in the same range.
IPVMU Certified | 06/27/13 12:46pm
I would be interested in seeing a more definitive and robust testing shoot out between the Sightlogix and their competitors.