The $500 FLIR Thermal Camera Tested

I've been aware of this upcoming camera for a while and I can't figure out what the real point of it is.

These days, 720p is essentially entry-level resolution, 1080p is mainstream and 3MP and 5MP are fairly common for cases where you want a step up. And, obviously, there are options for even higher resolution way beyond that.

To call something with an 80x60 sensor, 4800 pixels, a "camera" in the modern cctv market seems almost devious. Even the camera boards for hobbyist microcontroller products like the Arduino have 640x480 resolution cameras.

As noted above, you'll NEVER get any usable detail from the camera at any range. Thermal cameras in general suffer from lack of good details, and this is almost purely a blob-vision unit.

I like that this review concentrated on the analytics comparisons, because that's what this unit is really about.

To me, this is an advanced outdoor motion sensor, not a "camera". It should have a more robust API (IMO) and 4+ configurable contact-closure outputs. Something that can make it useful both for ONVIF-style motion-activation recording, as well as useful and easy to install for burglar alarms, outdoor lights, and similar cases.

I have a feeling that the actual reliable range of this unit would be more like a 100' range and 75' HFOV though. In large part due to the fact that because of the very low resolution objects are very "blobby" and you lose a good amount of granularity if you want to set ROI's or Tripwires with any degree of precision, as is often the case when using products like this for perimeter protection purposes.

It's somewhat surprising and concerning to see innovation in thermal cameras moving towards lower-end devices with questionable applications. As camera resolutions have increased, along with much better low-light capabilities and "pro" cameras coming with decent onboard IR, it's harder and harder to find cases where thermal cameras are a logical approach. I see 3 main categories where thermal makes sense over optical cameras, and they are all fairly niche in the grand scheme of the market:

1) Solar deployments where night video is important. Overall, thermal will be much lower power use than a camera + IR of any kind.

2) Requirements to spot objects at very long range. Not just the ability to cover a large chunk of perimeter with a single camera, because it's often similar price and better overall system to use 2x or 3x the number of optical cameras (even counting in VMS licenses, labor, etc.), but cases where you need to spot an object a distance where there is no way to place the camera closer, such as boats off shore. And this too assumes you have a strong night-time requirement because an optical camera or PTZ would probably overall be better during the day.

3) Applications where you need to see at night and need an absolute passive approach. Prisons, etc. are concerned about the red glow of ir illuminators giving away the location of the ligths or cameras, so thermal can be useful here.

A lot of the marketing for thermal cameras used to concentrate on things like the ability to deal with strong backlighting or other scenarios where an optical camera wouldn't be able to provide a good image even in daylight. While those situations CAN occur, I think they are fairly rare and can often be solved by adjusting camera placement or other things. The current crop of optical cameras is also much better at dealing with challenging lighting than what we had even 3 years ago.

In day-to-day use, for just basic CCTV or applications needing some kind of analytics I don't think the current crop of thermal cameras, even at $500, can provide better all-around performance than a similarly priced optical camera option. You give up 100% of the detail for potentially a slight false alarm reduction?

The overall CCTV market seems to be trending away from thermal cameras. We haven't seen any of the mainstream manufacturers announce a thermal option in the last few years (after the wave of the offerings from Axis, etc. a few years back). DRS basically imploded and is not just OEM'ing stuff, and FLIR went the other direction and is selling thermal cameras plus low/mid-market Dahua gear with FLIR logos.

It would be great to see some real innovation with thermal cameras, but this isn't it, IMO.

Reactions:

(1)

John Honovich

IPVM

"I don't think the current crop of thermal cameras, even at $500, can provide better all-around performance than a similarly priced optical camera option. You give up 100% of the detail for potentially a slight false alarm reduction?"

This camera at $500 plus a 3MP FLIR at under $200. Total: $700. How's that for detail and false alarm reduction?

Kick around thermal camera innovation all you want but there's another segment with super weak innovation - video analytics. And now that Avigilon has made it harder for 3rd party VMSes to integrate with their analytics, what is the rest of the market supposed to do?

I'd like to see FLIR do a camera combining the Lepton thermal imager + CMOS, use the CMOS for high res and Lepton for analytics. Even without that, doing a combo play is still relatively inexpensive and a better alternative than most 'analytics' on the market.

Reactions:

(3)

(2)

Undisclosed #1

This camera at $500 plus a 3MP FLIR at under $200. Total: $700. How's that for detail and false alarm reduction?

Yes, that's what I was alluding to when I was talking about strong API's and ONVIF (I recognize that is has ONVIF motion support). Basically this isn't a "camera" so much as it is some kind of intelligence to make your actual camera smarter or more functional.

Kick around thermal camera innovation all you want but there's another segment with super weak innovation - video analytics.

There's LOTS of segments in this industry with weak innovation. We haven't seen much groundbreaking from access control, either. I'm not sure what you're getting at with that statement. At least with video analytics we're seeing trends UP in resolution, with multiple companies doing 720p/1080p analytics cameras for instance.

I'd like to see FLIR do a camera combining the Lepton thermal imager + CMOS, use the CMOS for high res and Lepton for analytics.

I know they were kicking around a similar idea at some point, and I agree that would be something closer to innovative and have more uses (IMO) than this $500 unit. But they'd need to use a CCD sensor, since those are better than CMOS ;)

Reactions:

John Honovich

IPVM

"At least with video analytics we're seeing trends UP in resolution, with multiple companies doing 720p/1080p analytics cameras for instance."

What manufacturers are delivering this with high analytics accuracy?

That's the challenge I see.

Reactions:

Undisclosed #1

I'm confused.

Are you trying to branch this discussion into the current state of video analytics, are you essentially agreeing with me that this unit is more about the "analytics" than the "camera", or something else?

Reactions:

John Honovich

IPVM

You claim this camera is not interesting / useful / valuable compared to existing offerings.

I am asking you to be specific about who those 'multiple companies' are that are better. Please list them.

Otherwise, how can we evaluate your vague line of argument. At least if you give specific counterpoints, we can either agree or refute.

Reactions:

Undisclosed #1

My claim is that as a camera this is not very interesting, yes.

However I believe that it is potentially valuable as a kind of advanced motion sensor, coupling the intelligence of this unit with the video stream of an optical camera to create an alternative to the current analytics offerings.

I also stated that it could have potential uses beyond the standard CCTV-stuff, but that it would probably need a more robust API and/or DIO's to be fully utilized in that aspect.

Further I stated the CCTV market seems to be trending away from thermal cameras due to optical camera offerings having better overall value propositions, or being closer to customer expectations, for a given price point.

I listed what I see as the 3 main practical uses for thermal cameras these days, and stated they are all fairly niche.

Otherwise, how can we evaluate your vague line of argument.

If you honestly cannot find something more constructive to comment on or expand on from my original comment, then me listing some companies that I think might have better performance doesn't seem like that is the missing piece to really opening this discussion up to detailed analysis.

Reactions:

John Honovich

IPVM

"I listed what I see as the 3 main practical uses for thermal cameras these days, and stated they are all fairly niche."

And all of that is irrelevant for this test.

The rest of your comment literally repeats your previous comment.

I was hoping that you would back up your analytics claim with specific manufacturers but evidently you cannot.

Reactions:

Undisclosed #1

I was hoping that you would back up your analytics claim with specific manufacturers but evidently you cannot.

I was hoping you might be able to reference this to other IPVM analytic camera tests. You posted seemingly decent reviews of the Bosch cameras less than 2 weeks ago.

You tested the updated DvTel cameras with analytics less than a year ago.

In one of your earlier replies you write:

What manufacturers are delivering this with high analytics accuracy?

That's the challenge I see.

So forgive me if I'm missing something obvious, but are you saying you felt these other products you tested don't have high accuracy? Do you think this FLIR unit is competitive with those in an overall price/performance ratio?

I'm not sure why you can't offer a response to my comment without me listing specific companies. It would seem like you have at least 2 recent-ish test cases that you could use to support or refute my comment already.

Reactions:

John Honovich

IPVM

"You posted seemingly decent reviews of the Bosch cameras less than 2 weeks ago."

If you read the Bosch test report, you would not make such an uninformed statement. The Bosch test was about object classification (people, cars, bikes, trucks, etc.), different than this test. Also, their classification had issues.

DVTel is one example of a solid analytic camera (our test report here) but that's just one company beyond Avigilon.

By contrast, how many companies offer solid integrated IR or WDR? Dozens Easily.

So my point is, after 5 comments, you have yet to established more than 2 companies with strong analytic performance. I wish there are more.

This is part of the reason why the FLIR results are useful. And, it is very low cost. So you can harp on the fact that the FLIR camera has poor video quality or you can acknowledge that you can put a conventional camera next to it and still be less expensive than most analytic cameras in the market.

Reactions:

Undisclosed #1

So you can harp on the fact that the FLIR camera has poor video quality or you can acknowledge that you can put a conventional camera next to it and still be less expensive than most analytic cameras in the market.

It might be less expensive, but it won't be nearly as effective.

Try setting up some of the following scenarios and show us how it performs:

1) A tripwire with precision of +/- 10'. Tripwires aren't my favorite analytic, but they are at least something everyone does.

2) A region of interest with precision around the perimeters of +/- 10'

3) A loitering rule for 30 seconds within a specified region.

Do the above at a target range of at least 75' (this is not a random number, I'm basing it on the common deployment of cameras mounted on a building looking out across a parking lot at a perimeter line).

Next, show us how an event from this camera can send an alert to an operator with a usable video clip of the event (the typical 3ish seconds of pre-alarm, 5 seconds of past alarm video) from the camera it's paired with.

When you're comparing costs, also factor in the extra costs for the additional licenses for these cameras.

This camera is potentially a good start, but as it is, it's not very interesting and its not a very practical alternative to what most people would use analytics for.

All it's really offering is a better grade of motion detection, but there are no options to go beyond that, and there does not appear to be a lot of granularity in setting the motion area. This is going to limit the practical working area, and the overall applications.

If you read the Bosch test report, you would not make such an uninformed statement.

I did read the report. It said that the sub-classifications didn't work very well, but it did a decent job overall with person vs. vehicle in general (the FLIR camera does not appear to have any classifications). The report also said:

With little activity in the scene, cars, people, and trucks are classified correctly >90% of the time.

And I would wager that for most applications that want analytics the Bosch camera would overall be a better option than this FLIR unit, or this FLIR unit coupled with another camera.

Reactions:

John Honovich

IPVM

So now you have abandoned any attempt at listing those 720p / 1080p analytic offerings you posited 4 comments ago.

And no, we are not doing another test of loitering for 30 seconds at 75 feet away from a camera. I do not see the point.

I asked you to clarify your comment about other manufacturers and instead you have responded by asking us to do another 40 hours of testing.

Reactions:

Undisclosed #1

So now you have abandoned any attempt at listing those 720p / 1080p analytic offerings you posited 4 comments ago.

The following companies have all moved to doing analytics on higher-resolution camera streams over the last several years. Some of them used the higher resolution to increase range, others analyze a downsampled stream but at least provide a video clip of higher resolution:

-AgentVi

-Avigilon

-Bosch

-DvTel

-UDP/VCA

I find it hard to believe that you were not already aware of these companies, but since you require a list, there it is. Now please tell me how naming those materially changes the conversation around this low-res FLIR camera with motion detection?

Reactions:

John Honovich

IPVM

Thank you for the list. I knew your list would prove my point. There's very few analytic options. On that list, from our testing, I am not convinced about Bosch being on the same level of false alert minimization as Avigilon, DVTel nor this FLIR. As for UDP/VCA, I have yet to hear enough positive things from third parties to gain confidence. Even their supporters admit that they still get a lot of false alarms but that it is less than old VMD.

Net / net, there just aren't a lot of options that have sufficiently low false alert levels to justify using.

Reactions:

Undisclosed #1

And no, we are not doing another test of loitering for 30 seconds at 75 feet away from a camera. I do not see the point.

I asked you to clarify your comment about other manufacturers and instead you have responded by asking us to do another 40 hours of testing.

You're making a big deal in this thread about other analytic companies. This FLIR camera does not have "analytics" in what I think most people would define as the mainstream interpretation. It has motion detection.

Since you insist on positioning this camera alongside analytics, I asked you to do some "analytics" tests to see how it works. *That* was the point, if you feel this unit is a solid alternative to common analytics cameras show something more than just motion detection.

I wanted to see not only how this would handle something like a trip wire, but also how you "plumb" a usable event clip all the way through to a guard or central station or something. I think this might be possible on some VMS's, but I also think it could be very cumbersome most of the time.

I mentioned doing the tests at some distance from the camera because that is, in my experience, a common range for those kinds of applications.

Reactions:

(1)

Trisha (Chris' wife) Dearing

IPVMU Certified

I'd like to see FLIR do a camera combining the Lepton thermal imager + CMOS, use the CMOS for high res and Lepton for analytics.

That would be nice. It also might be of use to send the combined thermal/visible image to analytics, to put some definition around the blobs. They already are merging them in the FLIR ONE, which merges a 80x60 lepton sensor with a VGA visible sensor, together in one picture. For $199. Of course it still needs a part or two (encoder) to make it a IP camera, but seems do-able.

As for higher resolution cheap leptons, FLIR showed off the new version of the FLIR ONE at CES in January. They have increased the horizontal pixel count from 80 to 180!

Although, it should be noted they are only claiming 4x greater resolution, not the 5x one might expect from the increase of horizontal pixels. ;)

Reactions:

(1)

Michael Miller

In reply to Trisha (Chris' wife) Dearing

Flir's MSX® Thermal Image Enhancement is very intresting and I would imagine could be very usefull for analytics and general CCTV use.

MSX Off (Thermal only)

MXS On (Thermal and visible light camera)

Reactions:

(2)

Michael Miller