Author Topic: Flir Boson  (Read 14182 times)

0 Members and 1 Guest are viewing this topic.

Offline cynfabTopic starter

  • Regular Contributor
  • *
  • Posts: 175
  • Country: us
Flir Boson
« on: October 19, 2016, 02:09:31 pm »
 
The following users thanked this post: -jeffB

Offline sam1275

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Re: Flir Boson
« Reply #1 on: October 19, 2016, 04:56:53 pm »
I watched that for a long time, it just become available :-+
 

Offline sam1275

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Re: Flir Boson
« Reply #2 on: October 20, 2016, 11:25:41 am »
I read the long manual...
It's a good core, it provide good image and have many settings can be changed by user, some processing settings really makes the picture looks amazing.

However, if anyone want to buy it, there's somethings you might want to know:
The current batch of Boson is more like a Beta version, some features are not available now, Flir manual says they will be added by the next few firmware version, the update can be done by user, but they cannot promise those new feature will work on old core version.
The difference between 3 Grades - Industrial/professional/commercial are more than NETD. According the manual, The lower the grade, the more bad pixels there will be. Further, the lowest commercial version even have the gain setting locked to high, this means you cannot see things above 150C degree.

There's many advertisement showing Boson can do many cool things such as detect motion, detect animals and vehicle and so on, they put those AI features as the main feature for the Boson. But I didn't see any of those been mentioned in the manual, will they be added by the next few releases, or they are archived by the core itself?
 

Offline cynfabTopic starter

  • Regular Contributor
  • *
  • Posts: 175
  • Country: us
Re: Flir Boson
« Reply #3 on: October 21, 2016, 12:08:42 am »
Totally agree, Boson is a very new core and as such will no doubt suffer from (or the user will suffer from) early adopter syndrome.
The on board processing capability could be a game changer or a stumbling block for those who want to (and have in the past) built their own post processing systems.
It is really too early to tell how good those algorithms will be.
I'm sure all the images in the videos were from the higher end devices... that's just normal marketing.

What I'd really like to see is Fraser get a hold of one and do a comparison to his "collection" in terms of image quality.
I'd also like to see Mike get one and do a teardown... just out of raw curiosity.
 

Offline Jeroen3

  • Super Contributor
  • ***
  • Posts: 4078
  • Country: nl
  • Embedded Engineer
    • jeroen3.nl
Re: Flir Boson
« Reply #4 on: October 21, 2016, 07:26:59 am »
Why can't the rest of the world enjoy 60 FPS thermal imaging?
 

Offline Bill W

  • Super Contributor
  • ***
  • Posts: 1102
  • Country: gb
    • Fire TICS
Re: Flir Boson
« Reply #5 on: October 21, 2016, 02:54:18 pm »
Why can't the rest of the world enjoy 60 FPS thermal imaging?

Because the US don't want you to, and FLIR detectors are out of the US.
<9Hz, no license required
>9Hz US Dept. of Commerce license required, and that means signing up to a US license requirement for every subsequent transfer and to telling the FBI if you lose it.

On the other hand this does provide some protection for non-US suppliers at least in the middle ground.  Low end put up with 9Hz, high end would get a license if they wanted.

Bill

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13745
  • Country: gb
    • Mike's Electric Stuff
Re: Flir Boson
« Reply #6 on: October 21, 2016, 03:10:40 pm »
Just noticed Bosons are now available via Groupgets - starting at $1129
https://groupgets.com/manufacturers/flir/products/boson
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline lukier

  • Supporter
  • ****
  • Posts: 634
  • Country: pl
    • Homepage
Re: Flir Boson
« Reply #7 on: October 21, 2016, 08:05:20 pm »
Because the US don't want you to, and FLIR detectors are out of the US.

So if I got a thermal camera from an EU company, let's say InfraTec GmbH, then I don't have to worry about this US Dept of Commerce nonsense and can buy 30 or 60 Hz straight away?

Some time ago I was trying to find some information how to get this BIS paperwork sorted out, but it looks extremely complicated (involving me, US exporter vouching for me, me contacting US Dept of Commerce? etc). Might ask a friend who once went through that when buying FLIR Tau some time ago (and in Switzerland). I could understand restrictions to the countries that US considers "axis of evil", but UK or Switzerland, NATO allies, come on.

And the ITAR system doesn't seem very effective anyway as Iran managed to get Siemens PLCs (that Stuxnet attacked), precision multi-axis CNCs to manufacture the centrifuges (or Mitutoyo CMM case). BTW fun fact, cameras with fps > 1000000 are also ITAR controlled.
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13745
  • Country: gb
    • Mike's Electric Stuff
Re: Flir Boson
« Reply #8 on: October 21, 2016, 08:40:33 pm »

So if I got a thermal camera from an EU company, let's say InfraTec GmbH, then I don't have to worry about this US Dept of Commerce nonsense and can buy 30 or 60 Hz straight away?
Yes, or a local distributor for a US company. e.g. Flir 60fps cameras are available from catalogue distributors like RS.
It will still be illegal to ship them to dodgy countries though.
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline Jeroen3

  • Super Contributor
  • ***
  • Posts: 4078
  • Country: nl
  • Embedded Engineer
    • jeroen3.nl
Re: Flir Boson
« Reply #9 on: October 21, 2016, 09:26:26 pm »
Because the US don't want you to, and FLIR detectors are out of the US.
So if I got a thermal camera from an EU company, let's say InfraTec GmbH, then I don't have to worry about this US Dept of Commerce nonsense and can buy 30 or 60 Hz straight away?
You should always prefer EU stuff above US stuff. Unless you are a US citizen.
I thought this was common sense already?

And the ITAR system doesn't seem very effective anyway as Iran managed to get Siemens PLCs (that Stuxnet attacked), precision multi-axis CNCs to manufacture the centrifuges (or Mitutoyo CMM case). BTW fun fact, cameras with fps > 1000000 are also ITAR controlled.
They most certainly didn't buy them. You can't get any money to europe from Iran. It even bounces when it goes through Turkey first.
« Last Edit: October 21, 2016, 09:28:15 pm by Jeroen3 »
 

Offline Bill W

  • Super Contributor
  • ***
  • Posts: 1102
  • Country: gb
    • Fire TICS
Re: Flir Boson
« Reply #10 on: October 21, 2016, 09:34:53 pm »
So if I got a thermal camera from an EU company, let's say InfraTec GmbH, then I don't have to worry about this US Dept of Commerce nonsense and can buy 30 or 60 Hz straight away?
Yes, as long as the detector is EU (eg ULIS), or from another country applying straight Wassenaar rules (eg i3 Korea), then movement within EU is OK.
Older US cores (in Argus, Cadillac etc) are also OK as the re-export rule only appeared in the last 10 years or so.

Yes, or a local distributor for a US company. e.g. Flir 60fps cameras are available from catalogue distributors like RS.
It will still be illegal to ship them to dodgy countries though.

Not just dodgy countries, movement to ANY countries requires DoC re-export clearance.  Slightly surprised RS(UK) are able to offer FLIR E60 (60Hz) as next day delivery though, assume the licensing is OK for any UK individual.

regards
Bill

Offline sam1275

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Re: Flir Boson
« Reply #11 on: October 22, 2016, 12:11:36 am »
Totally agree, Boson is a very new core and as such will no doubt suffer from (or the user will suffer from) early adopter syndrome.
The on board processing capability could be a game changer or a stumbling block for those who want to (and have in the past) built their own post processing systems.
It is really too early to tell how good those algorithms will be.
I'm sure all the images in the videos were from the higher end devices... that's just normal marketing.

What I'd really like to see is Fraser get a hold of one and do a comparison to his "collection" in terms of image quality.
I'd also like to see Mike get one and do a teardown... just out of raw curiosity.
The groupget site says Flir will release SDK for Boson in a few months, so I guess the AI processing will still be done in the core, but need user to develop by it's SDK.
By the way, it seems to use a 12-core 600mhz VLIW processor, with 1GB or 4GB memory, seems good.
 

Offline lukier

  • Supporter
  • ****
  • Posts: 634
  • Country: pl
    • Homepage
Re: Flir Boson
« Reply #12 on: October 22, 2016, 10:03:02 am »
You should always prefer EU stuff above US stuff. Unless you are a US citizen.
I thought this was common sense already?

I agree, but often it's easier said than done. I wish there could be an EU alternative to Google or Facebook for example. FLIR makes very good sensors, but I'll research EU manufacturers as well.

The groupget site says Flir will release SDK for Boson in a few months, so I guess the AI processing will still be done in the core, but need user to develop by it's SDK.
By the way, it seems to use a 12-core 600mhz VLIW processor, with 1GB or 4GB memory, seems good.

Movidius Myriad 2 that is used by FLIR Boson is a very cool chip. A year ago or so I asked David Moloney (Movidius CTO, very knowledgeable guy btw) about devkits that they were planning to manufacture and I could possibly get one, but only if I signed the NDA, so I backed down (it would probably require a lot of paperwork and possibly uni's legal team involvement). Not very good way to get widespread adoption. Maybe things have changed, as back then AFAIR Myriad 2 was still in the "engineering samples" stage, hence the secrecy.

I know some people that got the kits and according to these sources programming of this chip is not bread and butter. nVidia CUDA development is a walk in the park apparently.

Maybe the FLIR SDK will be a wrapper around the low level Myriad API, so it will be easier to use (but possibly limited in some respects). Personally I don't care much about Myriad enabled FLIR, I'd rather buy a high speed sensor with simple LVDS output and I can do my algorithms on the GPU (R&D work, not product).

BTW For those in the imaging/computer vision world, recently FLIR acquired PointGrey. I wonder why and what this means for the future. PointGrey makes very good cameras, but it's not rocket science, they buy various sensors in bulk, slap an FPGA and an interface chip (USB 3.0/GbE/IEEE1394) and that's pretty much it, they don't even bother adding a few dollar cheap IMU there (would be beneficial for some inertial-vision research). They don't publish any internal details, and the FPGAs there are not huge, probably doing basic preprocessing (dead pixels) and buffering, so it is rather impossible to turn these into "smart-cameras". Will FLIR start integrating visible spectrum cameras or PointGrey selling thermal-enabled ones? No idea.
 

Offline sam1275

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Re: Flir Boson
« Reply #13 on: October 22, 2016, 08:37:07 pm »
You should always prefer EU stuff above US stuff. Unless you are a US citizen.
I thought this was common sense already?

I agree, but often it's easier said than done. I wish there could be an EU alternative to Google or Facebook for example. FLIR makes very good sensors, but I'll research EU manufacturers as well.

The groupget site says Flir will release SDK for Boson in a few months, so I guess the AI processing will still be done in the core, but need user to develop by it's SDK.
By the way, it seems to use a 12-core 600mhz VLIW processor, with 1GB or 4GB memory, seems good.

Movidius Myriad 2 that is used by FLIR Boson is a very cool chip. A year ago or so I asked David Moloney (Movidius CTO, very knowledgeable guy btw) about devkits that they were planning to manufacture and I could possibly get one, but only if I signed the NDA, so I backed down (it would probably require a lot of paperwork and possibly uni's legal team involvement). Not very good way to get widespread adoption. Maybe things have changed, as back then AFAIR Myriad 2 was still in the "engineering samples" stage, hence the secrecy.

I know some people that got the kits and according to these sources programming of this chip is not bread and butter. nVidia CUDA development is a walk in the park apparently.

Maybe the FLIR SDK will be a wrapper around the low level Myriad API, so it will be easier to use (but possibly limited in some respects). Personally I don't care much about Myriad enabled FLIR, I'd rather buy a high speed sensor with simple LVDS output and I can do my algorithms on the GPU (R&D work, not product).

BTW For those in the imaging/computer vision world, recently FLIR acquired PointGrey. I wonder why and what this means for the future. PointGrey makes very good cameras, but it's not rocket science, they buy various sensors in bulk, slap an FPGA and an interface chip (USB 3.0/GbE/IEEE1394) and that's pretty much it, they don't even bother adding a few dollar cheap IMU there (would be beneficial for some inertial-vision research). They don't publish any internal details, and the FPGAs there are not huge, probably doing basic preprocessing (dead pixels) and buffering, so it is rather impossible to turn these into "smart-cameras". Will FLIR start integrating visible spectrum cameras or PointGrey selling thermal-enabled ones? No idea.
Are there any AI machine vision software already available and free? I think it's interesting to play with.
 

Offline dpenev

  • Regular Contributor
  • *
  • Posts: 189
Re: Flir Boson
« Reply #14 on: October 25, 2016, 08:55:03 am »
Hello all,

I am trying to understand what camera will serve me best for PCB close up inspection.
However I don't have experience in optics and I need some help.

From the boson datasheet https://groupgets-files.s3.amazonaws.com/boson/Boson%209Hz%20datasheet%2C%20102-2013-43%2C%20Rev%20100.pdf I see the following options available

f number =1 for all lenses

FocalLength = 2.3mm; 4.3mm; 6.3mm; 9.1mm 14mm; 18mm; 36mm; 55mm
 
I found an equation which relates the distances (relative to the lens) between the object on focus and the distance between the lens and the sensor

1/DistanceToObject + 1/DistanceToSensor = 1/FocalLength

As per my understandings the camera lens adjust the focus by adjusting the DistanceToSensor (I mean here from lens to Sensor). Is this so?
 
Based on this I have calculated that the smaller the lens aperture (2.3mm is the smallest) the smaller adjustment of the DistanceToSensor in percentage required.
So I guess 2.3mm objective will get focus the easiest. From the other hand I can not find in the datasheet what range of DistanceToSensor variations the objectives can do?

The other (probably related) consideration is the FOV. I consider this as a con? camera can view. Am I correct?
I think my PCB should span as much as possible from the cone at given distance. 
4.3mm objectives has FOV=50deg
6.3mm objectives has FOV=34deg
 
My PCBs are typically 200mm, and I want to take picture at about 300mm to get my PCB in the FOV I need the 4.3mm objective
With the 6.3mm I will need to increase the distance between the camera and the PCB to 350mm.

In addition intuitively I think that bigger lens the better as more light will enter.

Can someone shed some light on this subject. Is there rational of what I am thinking?
Any additional important things?

Thank you :)
Dimitar 
 

Offline lukier

  • Supporter
  • ****
  • Posts: 634
  • Country: pl
    • Homepage
Re: Flir Boson
« Reply #15 on: October 25, 2016, 09:25:02 am »
Are there any AI machine vision software already available and free? I think it's interesting to play with.

No idea about AI, but what I wanted to do is something along these lines:
http://www.ndt.net/article/qirt2014/papers/QIRT-2014-035.pdf

3D reconstruction with fused thermal data. Thermal camera alone seems rather poor for SLAM (no texture often, edges), but it can be coupled with RGBD (Kinect) or a passive camera. But to make it work I'd prefer something like Tau, with LVDS output, sync pulse (to synchronize/timestamp) and framerates similar to the RGBD/passive cameras. Then this could be coupled into something like ElasticFusion:



Also having multiple views with precise tracking allows super-resolution (FLIR calls that UltraMax markets that as rocket science  :palm:).
 

Offline Bill W

  • Super Contributor
  • ***
  • Posts: 1102
  • Country: gb
    • Fire TICS
Re: Flir Boson
« Reply #16 on: October 25, 2016, 05:04:55 pm »

As per my understandings the camera lens adjust the focus by adjusting the DistanceToSensor (I mean here from lens to Sensor). Is this so?
Yes that will be how it is done.

Based on this I have calculated that the smaller the lens aperture (2.3mm is the smallest) the smaller adjustment of the DistanceToSensor in percentage required.
So I guess 2.3mm objective will get focus the easiest. From the other hand I can not find in the datasheet what range of DistanceToSensor variations the objectives can do?
The 2.3mm will change focus more for a given movement or rotation.  You may or may not find this easier.

The other (probably related) consideration is the FOV. I consider this as a cone camera can view. Am I correct?
I think my PCB should span as much as possible from the cone at given distance. 
4.3mm objectives has FOV=50deg
6.3mm objectives has FOV=34deg
My PCBs are typically 200mm, and I want to take picture at about 300mm to get my PCB in the FOV I need the 4.3mm objective
With the 6.3mm I will need to increase the distance between the camera and the PCB to 350mm.
This is correct, although the 'cone' will be a rectangle because of the sensor shape.  Hence you are concerned with horizontal and vertical fields of view (HFoV / VFoV) and some people even refer to the diagonal field of view if they want a big number.
The other concern for you looking at PCBs is the size of one pixel projected onto the board.  If your board is 160mm wide then each image pixel of the 320 wide sensor is sampling a 0.5mm square on the board.


In addition intuitively I think that bigger lens the better as more light will enter.
No, if they are all f/1 the 'brightness' and resulting signal to noise is the same.
The narrow field of view lenses need to capture more 'light' as it is still spread out across the whole sensor area.

regards
Bill

Offline dpenev

  • Regular Contributor
  • *
  • Posts: 189
Re: Flir Boson
« Reply #17 on: November 09, 2016, 03:46:56 pm »
Hi Bill,

Thank you for your explanation.

So if you have to select which of the following two Boson lenses you would get for PCB inspection?
4.3mm objectives - FOV=50deg
6.3mm objectives - FOV=34deg 
f number =1 and price is the same for both.

Thanks
Dimitar
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf