Actual resolution of Flir One V2

[Fraser]

No reply from FLIR to my question on physical resolution yet but it is the weekend.

In truth, FLIR's response to my question will provide teh answer to this threads question. Answers will be one of the following:

1. Interpolation and image processing used to create a 160x120 image from a 80x60 microbolometer array

2. Microbolometer array is constructed in a 160x120 array of 19200 pixels

3. Nil response or decline to comment due to commercial sensitivities ..... read this as meaning interpolation of a 80x60 microbolmeter is being used in the design.

We shall see what FLIR say this week.

Aurora

[TopLoser]

In a visible light camera you use a resolution chart to see the true imager resolution. This gets below any interpolation that may be in play. Sadly I have no such chart for use in the thermal domain.

I'm curious... why is the finned heatsink not usable as a substitute for resolution chart?

[amyk]

In a visible light camera you use a resolution chart to see the true imager resolution. This gets below any interpolation that may be in play. Sadly I have no such chert for use in the thermal domain.

Would dark and light surfaces (i.e. ink and paper) have different enough emissivities to be distinguishable?

[jadew]

In a visible light camera you use a resolution chart to see the true imager resolution. This gets below any interpolation that may be in play. Sadly I have no such chert for use in the thermal domain.

Would dark and light surfaces (i.e. ink and paper) have different enough emissivities to be distinguishable?

No, however, a chart like that can be easily constructed by cutting rectangular holes in a piece of paper and having a higher temperature object behind it.

[Fraser]

@Toploser,

The following article explains how I use resolution charts with visible light cameras

http://blog.almalence.com/how-to-measure-the-resolution-of-a-digital-camera/

It is important to match the target 'grille' to the pixels and doing this with a fixed pitch heatsink is challenging and relies upon the distance of the camera to the lens to set the width of the 'fin' so that it covers only one pixel width and is 'in sync' the others that the microbolomers sees. A resolution chart provides 'grilles' or varying resolutions and is calibrated for such.

I am not saying a heatsink cannot be used, but it will not be easy.

Aurora

[TopLoser]

One last picture to ponder:



Landscape orientation 731 x 579 pixels 1.25:1 aspect ratio. Not quite the 1.33:1 ratio as would be expected so some horizontal pixels are being 'lost' offscreen.

Notice the jaggies on the right hand side sloping edge...

777 pixels total width after correcting for the missing pixels. I count 13 jaggies in a 63 pixel horixontal interval, one every 5.25 pixels, that rather conveniently gives what looks like a 160.3 pixel sensor resolution. Notice that Flir do some interpolation and processing between each jaggie edge to make it look better, but underlying resolution seems to be 160 pixels.

[TopLoser]

@Toploser,

The following article explains how I use resolution charts with visible light cameras

http://blog.almalence.com/how-to-measure-the-resolution-of-a-digital-camera/

It is important to match the target 'grille' to the pixels and doing this with a fixed pitch heatsink is challenging and relies upon the distance of the camera to the lens to set the width of the 'fin' so that it covers only one pixel width and is 'in sync' the others that the microbolomers sees. A resolution chart provides 'grilles' or varying resolutions and is calibrated for such.

I am not saying a heatsink cannot be used, but it will not be easy.

Aurora

And almost impossible with an unfocussed camera, but this is not a lab and I'm using stuff to get a VERY rough idea.

We're looking for a factor 2 here, not a few lines/mm difference between different lenses.

[marshallh]

That looks a lot better, but I'm not 100% convinced. I will try and get some comparison shots with my NEC.

[tomas123]

One last picture to ponder:

Landscape orientation 731 x 579 pixels 1.25:1 aspect ratio. Not quite the 1.33:1 ratio as would be expected so some horizontal pixels are being 'lost' offscreen.

Please post an original saved jpg from the flir app (this includes the raw values)

Then we can get the real resolution- see my samples (convert with exiftools):
https://www.eevblog.com/forum/testgear/flir-one-thermal-imaging-camera-teardown-and-hacks/msg551882/#msg551882

[tomas123]

Just took a pic of a finned heatsink. MSX turned off.

from this nice image we get a sensor resolution of 120x90 pixel like FLIR E5

Code: [Select]

$ convert crop772x579.png -resize x1200 res1.png
$ identify res1.png
res1.png PNG 900x1200 900x1200+0+0 8-bit sRGB 565KB 0.000u 0:00.000

$ convert -size 10x10 xc:none -stroke gray -strokewidth 0  -draw "line 0,0 0,9"  -draw "line 0,0 9,0"  1.png
$ convert -size 900x1200 tile:1.png -transparent white grid3.png

$ convert res1.png grid3.png -flatten overlay3.png
crop of a overlay with a 120x90 pixel grid (see steps on vertical lines)

[Fraser]

@tomas123,

120x90 

The E5 spec always interested me. I know of no microbolometer array that uses a 120x90 format.
I always suspected that such was just some weird idea from FLIR for an intermediate resolution step. As we know the Ex is 320x240 and they manipulate the data to create whatever resolution they want.

But why claim 160x120 and then provide a software created 120x90 resolution instead.

All very weird I must say 

Aurora

[TopLoser]

Please post an original saved jpg from the flir app (this includes the raw values)

Obviously out of focus range because I couldn't adjust the MSX distance value to align the IR and normal images.

EDIT: just added a diagonal view, probs from too far away to be useful but here it is anyway

[NathanFowler]

Very interesting comments all around, thank you TopLoser for sharing the images.

[jadew]

If you have a better thermal camera around, you could take an image with the Flir One, that covers a known width and height and then take another one with the higher resolution camera, of the same object but covering (width, height) * ratio. Even if the interpolation might fool you a bit, it should still be visible in the amount of detail.

[TopLoser]

If you have a better thermal camera around, you could take an image with the Flir One, that covers a known width and height and then take another one with the higher resolution camera, of the same object but covering (width, height) * ratio. Even if the interpolation might fool you a bit, it should still be visible in the amount of detail.

Not going to happen, sorry! This Flir One is already sold to an impatient group member who didn't want to wait, they picked it up a few hours ago.

I was impressed with it, I've already ordered another one for myself.

[NathanFowler]

I was impressed with it, I've already ordered another one for myself.

By chance did you have Gen 1 or the fine person who picked up yours, I'd be curious to see it compared and contrasted even with interpolation.

[frenky]

I did some playing with the images...

I measured the right side of this photo with 13 visible ridges which are actual thermal pixel and is 73.4px wide.
https://dl.dropboxusercontent.com/u/2063383/ir%20stuff/jaggies.PNG

So 1 thermal pixel is 5.6 pixels on image.

So I made a crop 504x504 which translates to 90x90 thermal pixels:


I resized it to 90x90:


And then up-scaled it back to 504x504:


If you look at the original and processed image you will see that new image has exactly the same ridges as the original:


So thermal resolution is not 120x160. It's closer to 103x131...

If we had a photo with hot wire going diagonally we could determine resolution quite precise.

[TopLoser]

I was impressed with it, I've already ordered another one for myself.

By chance did you have Gen 1 or the fine person who picked up yours, I'd be curious to see it compared and contrasted even with interpolation.

No I don't have a Gen 1 sorry. I asked if anybody with a Gen 1 could do the same kind of test as I did but got no response.

Over to you lot now, I posted enough pics to allow better minds than mine come to some conclusion.

[NathanFowler]

I have a Gen 1 on iOS, how can I help, I am very willing to take images/pictures?  I'm fixing to (American Southern Colloquialism for "happening in the next few seconds") to just order a Gen-2 and do a side-by-side comparison on-list for those that it could help.

[TopLoser]

I have a Gen 1 on iOS, how can I help, I am very willing to take images/pictures?  I'm fixing to (American Southern Colloquialism for "happening in the next few seconds") to just order a Gen-2 and do a side-by-side comparison on-list for those that it could help.

I just added a raw pic taken at a diagonal to a previous post, hope that helps thomas123 delve a little deeper.

Obviously I'd suggest an image of a heatsink with fins 1.5mm apart for direct comparison, but I guess that's not possible.

Maybe the group should suggest what they regard as a suitable comparison image.

[NathanFowler]

Obviously I'd suggest an image of a heatsink with fins 1.5mm apart for direct comparison, but I guess that's not possible.

Maybe the group should suggest what they regard as a suitable comparison image.

Would be great, I do have access to various heatsinks, boards, etc.  Selection of a common object with enough detail to discern differences between the two would be nice but I defer to the expertise of those here with the intent to help in any way I can.  Thanks all.

I do have some lenses but it's hit or miss.  I'm an amateur radio (ham)/computer guy so I just bang rocks together and call it science 

[frenky]

Maybe the group should suggest what they regard as a suitable comparison image.

Good image would be one without MSX and with straight (under angle) lines going from one side of the image to another. (Without any gaps on the sides)

Like this:

[TopLoser]

Maybe the group should suggest what they regard as a suitable comparison image.

Good image would be one without MSX and with straight (under angle) lines going from one side of the image to another. (Without any gaps on the sides)

Like this:

Bit of a pain without closeup lenses because of the fixed focus (min 1m I think, based on the MSX adjustment slider range ) of the standard lens.

Plus I really don't think single line images are any use due to the amount of interpolation and processing that can be applied to them. You really need multiple parallel lines.

[NathanFowler]

flir one gen one 1.0.11 firmware sent from iPhone 5s so sorry for rubbish typing

[frenky]

It's 80x60.