Yet another cheap thermal imager incoming.. Seek Thermal

[miguelvp]

You probably need to grab a flat field reference first
something of uniform temperature

[frenky]

I forgot to mention that I did that. That is working ok. I can see that by toggling Use ref frame checkbox.

Sent from my LG-D855 using Tapatalk

[miguelvp]

I forgot to mention that I did that. That is working ok. I can see that by toggling Use ref frame checkbox.

Sent from my LG-D855 using Tapatalk

Looking at the code it writes Cal_OffsetMap.dat
in function Kernel_DoCalibration that is in MainForm.cs

Grab Reference triggers the Btn_CalibrateClick in MainForm.Designer.cs that causes the calibration to happen.

I don't have the drivers installed so I can't test it, but that's what the code shows.

Edit, looking at the code more it seems if it's checked the code earlies out at sub_PreProcessMedian  in MainForm.cs

[miguelvp]

Try this, select the checkbox and grab reference, then uncheck it.
Or viceversa perhaps.

Edit: or do grab reference while unchecked and leave it that way (unchecked).

[frenky]

I tried many combinations without success...

And in the code the part that checks for the state of this checkbox is commented out:

Code: [Select]

void Kernel_PreProcessImage()
        {
if (LastPreProcessed_Uint16==null) { return; }

            // Display last sensor min/max values.
if (thermal!=null) {
label6.Text = "Cal Hi: " + thermal.FrameID1.MaxValue;
            label7.Text = "Cal Lo: " + thermal.FrameID1.MinValue;
}
//            if (chk_useMedian.Checked) {
//            sub_PreProcessMedian();
////            sub_PreProcessMedian();
////            sub_PreProcessMedian();
//    }
...
//the only method that is using chk_cal_useDefPixMap.Checked is this one:
        void sub_PreProcessMedian()
        {
        long val=0; int cnt=0;
        //Frame without edges
        for (int y = 1; y < 155; y++) {
                for (int x = 1; x < 205; x++) {
        if (CalDefPixelMap[x,y]) {
        if (chk_cal_useDefPixMap.Checked) { continue; }
        }


[miguelvp]

WIP probably (Work In Progress)

[frenky]

I was interested to find out how come JadeW's app is producing such a clean and noiseless images with external calibration.
His solution is really smart and simple.

I did my image correction by simply subtracting external calibration image and current image.
But he is calculating gain of each pixel compared to avg values of external calibration.

https://github.com/rzva/ThermalView/blob/master/ThermalView/frame.cpp

Code: [Select]

//////////////////////////////////////////////////////////////////////////
/// getOffsetCalibration
//////////////////////////////////////////////////////////////////////////
std::vector<int> ThermalFrame::getOffsetCalibration() const
{
std::vector<int> calibration;

calibration.resize(m_pixels.size());

for (size_t i = 0; i < calibration.size(); ++i)
{
uint16_t val = m_pixels[i];

if (val != 0)
calibration[i] = m_avg_val - val;
else
calibration[i] = 0;
}

return calibration;
}


//////////////////////////////////////////////////////////////////////////
/// applyOffsetCalibration
//////////////////////////////////////////////////////////////////////////
void ThermalFrame::applyOffsetCalibration(const std::vector<int> & calibration)
{
for (size_t i = 0; i < calibration.size(); ++i)
m_pixels[i] += calibration[i];
}


//////////////////////////////////////////////////////////////////////////
/// getGainCalibration
//////////////////////////////////////////////////////////////////////////
std::vector<double> ThermalFrame::getGainCalibration() const
{
std::vector<double> calibration;

calibration.resize(m_pixels.size());

const uint16_t min_val = 0;

for (size_t i = 0; i < calibration.size(); ++i)
{
uint16_t val = m_pixels[i];

if (val != 0 && !is_pattern_pixel(i))
calibration[i] = (double)(m_avg_val - min_val) / (m_pixels[i] - min_val);
else
calibration[i] = 1;
}

return calibration;
}


//////////////////////////////////////////////////////////////////////////
/// applyGainCalibration
//////////////////////////////////////////////////////////////////////////
void ThermalFrame::applyGainCalibration(const std::vector<double> & calibration)
{
for (size_t i = 0; i < calibration.size(); ++i)
m_pixels[i] = static_cast<uint16_t>(m_pixels[i] * calibration[i]);
}


[frenky]

I made some excel calculations with simulated data.

Let say we have only 3 different pixels with values (5500,6000,6500) when pointing to uniform thermal plane.
Then we replace this thermal plane with one with higher temp and we observe that values of pixels have increased by approx 10%.

We can easily compensate for that with formula:

Corrected pixel val = (Pixel val of new thermal img * Average pixel val of calibration img / pixel val of calibation img)  - Average pixel temp of calibration img

Corrected pixel val= (6050*6000/5500)-6000=600
You can then add any number to that value to push image pixel values into desired range...

Last column shows the correct pixel values of the higher temp thermal plane:

[joe-c]

Hey Guys,

sorry for confusing you, but I am very busy the last time.

the "Use defect pixel map" was only implant in the Median filter.
in a earlier version, the Median only uses "non defect pixel" but the results are not as good.
a second version "Median-diff" that replaces only pixel if the calculated value of the pixel is differs more than a threshold
works also not as good as wanted.

Also there 2 DefPixelMaps in this program. One is inside the Seek Class and calculated automatically with reference frames.
the second one (that is used for the Median) was just self calculated with a reference frame.
if a pixel differs more than the threshold, it was marked as defect (see image).
not really useful too.

its version is still a construction area. I forget to remove some of the (now) useless functions.
Joe-c

[eneuro]

Found this on youtube

Another review of Seek Thermal Reveal there:



Looks much better than first Seek Themal's design with crappy "smart" phones USB interface

Anybody knows, is it possible turn off Seek Thermal Reveal screen and only capture IR frames directly to PC/tablet using its USB interface, in similar way we did it with previous Seek products, I guess?



Who will make teardown of this Seek Reveal worth $399 first? 

[Trax]

Who will make teardown of this Seek Reveal worth $399 first? 

If some one pays me 300$ I'll do that.

how about a crowd founded tear down?

[eneuro]

how about a crowd founded tear down?

Probably, someone already did it?
We'll not have wait too much, I think... Easy $es for YT video proucers, but this site is rather promoting Flir$ products, so I do not expect Dave will ever make any Seek Thermal teardown, instead of Flir  due to some kind of site agreements with Flir, I guess, becaue of I see no other reason make Seek Thermal teardown for EEVblog fans, but some kind of marketing promotion of other thermal stuff 

[Fraser]

Not wanting to be too mean, but what would a teardown of a Reveal reveal 😆 it's a SEEK core bolted into a case with processing electronics and a display.  No new core, no visible light camera, no better image processing. Adding a daft LED flashlight was a joke, but I am not laughing.

I am waiting for SEEK to produce a decent core coupled to becent imaging software. Until then it is just a SEEK Mk1 core dressed up, but sill flawed.

As to Dave being pro FLIR only...... No idea but he doesn't appear to be a thermal fan boy, or heavily engaged with FLIR. People tend to only review stuff that they find interesting. Maybe no one finds the Reveal interesting enough to buy one for teardown? That is my view of the product anyway.

Fraser

[eneuro]

I am waiting for SEEK to produce a decent core coupled to becent imaging software. Until then it is just a SEEK Mk1 core dressed up, but sill flawed.

How did you know this, since you haven't got this thing in your hands?
Redesign of previous Seek Thermal lens holder used before in "smartphone" toys could create different heat disipation patterns, co thermal core can work in different way.
Since, I'll never buy Flir$ thermal imaging camera for reasons I wrote many posts ago, considering buying this Reveal-no need to mess with crappy phone to see something-but I'm suspicious, that after successfull hacks they (Seek) could change internal design and make difficult to read easy raw sensor data like in its first design   

[Fraser]

@Eneuro

I have my sources 😉

[efahrenholz]

The reveal isn't really worth the tear down. Maybe it could uncover some hints? I dont know, i highly doubt they changed the firmware much. It's still the same crappy noisy sensor. You can tell just from the on screen images. It looks EXACTLY the same as the cell phone app image. The only thing that *might* be of interest is the update process. The only way to deliver updates to the firmware is through the USB connection. The reveal acts as a removable drive and you copy over a file called "upgrade.str", so if someone wants to go to the website and grab that file and tear it open, maybe they could figure out something that might be useful to both the reveal and seek app crowd.

[Bill W]

Fraser,

>>Adding a daft LED flashlight was a joke, but I am not laughing. <<

As someone who put an LED lamp into a (different, torch-like) thermal camera design, I convinced myself that they can have their uses for night / security applications. 
Try putting a key in a door with a thermal camera (no close focus).
Stairs are a challenge.
A nice bright light shows 'I have found you' and also will render the 'quarry' immobile as it destroys night vision.

regards
Bill

[Fraser]

Bill,

Very fair comment. I agree in your scenario it does make sense. I was thinking more about them omitting a visible light camera..... yet bolting on a flipping great LED flashlight which would have made perfect sense had there been a visible light camera that needed it. Not such a daft idea after all then. I retract that criticism

In security setting it is advisable to use a 5D cell Maglite as the light source though....... It doubles up as a perfectly legal truncheon

Fraser

[Fraser]

Picking up on a good point made by Eneuro......

We have seen some interesting development of affordable thermal imaging equipment over the last couple of years. I hope for even more development work in the coming months and years. BUT I do hope that designers do not get too fixated on marrying thermal cameras to mobile phones as the only viable solution. SEEK have obviously seen that there is merit in bolting their core into a self contained unit with its own display. Both approaches have advantages and disadvantages and both approaches should be offered.

There is little doubt that modern mobile phones are powerful computers with easy interfacing and APP updating via Google or Apple. BUT attaching a small thermal camera module to a mobile phone in a manner that is both ergonomic and secure can be challenging. IMHO the original FLIR ONE was a very well designed solution to this challenge but it was flawed. It could only fit one size and shape of phone.

As a regular user of thermal imaging cameras I have the following choices when selecting my tool for a job.....

1. FLIR PM 5xx or 6xx series industrial camera

Gorgeous build and image quality but heavy at 2kg and not the most friendly format (camcorder style)

2 NEC AVIO TH series industrial cameras

Same comments as above.

3.  FLIR E4

Robust and ergonomic with its pistol grip format. Light weight compared to the PM series. Decent image quality when upgraded.

4. FLIR E2

Torch format and relatively lightweight. Not as comfortable to use as the pistol grip format but it gets the job done.

5. FLIR F1G2 Android

A direct competitor to the SEEK mobile phone cameras.
Small form factor. Somewhat compromised image quality due to lens and deliberate image degradation by FLIR. Not the most ergonomic camera when attached to a mobile phone by just a micro USB connector. Fiddly to hold and presstouch screen buttons. Dependant upon mobile phones processing power for frame rate. Not the most professional looking tool. At a recent demonstration that I delivered, the audience thought it clever, but little more than a yuppy toy.
You have to carry the F1G2 in its silly little pod as well. Extract it and then attach it to your phone. This provides ample opportunity to drop it or lose it in your travels.

I have several other thermal camera types but these are a good representation of the current market.

So which format is my favourite to use ? Without a doubt I find the pistol grip with trigger image capture the most ergonomic. I love my industrial thermal cameras and they have advantages that I have not detailed here, but they are more cumbersome to use in my lab applications. For wildlife work, attached to a tripod, they are amazing performers.

The E2 torch format is similar to the Reveal and does work quite well. The issue with the format is that you are looking down on a screen that is showing the output of a camera lens pointing ahead. I personally find this less friendly than the pistol grip format with almost vertical screen as you can choose to look past the cameras vertical screen and at the area that is being viewed

I personally find that thermal cameras attached to a mobile phone are a faff that I can happily do without. On several occasions my phone battery has been dead and using the touch screen whilst holding the unfriendly mobile phone shape aggravates me. Yes these are cheap, yes they are very compact, but once attached to a phone they are just plain awkward in my opinion. Some OEM's recognise this and supply a camera bracket and pistol grip. It ends up like the Pistol from the Bond film, The man with the Golden Gun....... Having to assemble the parts before use. Not a tool format that I will ever love me thinks. Good for the very occasional job or party though

So SEEK have offered a torch format camera in their new Reveal range. I am pleased to see such offered. It is compact and easy to use. In most respects it is an ergonomic hand tool but how much better it could be if the image processing is improved. I could happily use such a compact camera. Look at the latest FLIR offerings that use the LEPTON 2 core though..... They have kept the upright screen format so you can look PST the screen if desired. I do prefer that format to the Reveal.

Let us hope that we see new and novel stand alone cameras that do not need a flippin mobile phone to function.

The ability to wirelessly link to a mobile phone/tablet in order to transfer or view images is a good facility to have though.

Just my ramblings on this topic. Other users will have different needs and opinions on format's.

Fraser

[encryptededdy]

You can download the Seek Reveal firmware updates off the Seek website: http://support.thermal.com/hc/en-us/articles/206535443-Updating-Seek-Reveal-Software

Perhaps useful if someone wants to poke around with it? They're really small <1MB.

[Trax]

Currently in the EU it seams that the Flir one G2 is selling for the same price as the normal Seek thermal...
Do you think one of this will soon drop in price? Such that waiting would be advisable?

[frenky]

If you are deciding between latest Flir One and First SeeK thermal I woud suggest go with Flir.

Flir has shutter in front of lens so the the produced thermal image is accurate. Flir also has ability to lock the thermal range (to the current range). And the MSX is cool to (except for macro).
Seek has shutter behind lens so any dust, "glue" or vapour on lens (behind or front) will show on thermal images. And you can't lock the thermal range.

Bad shutter design on Seek is clearly visible here:


The left image has a big dark spot on the top right. That is false data, due to dirty lens. The right image was taken with my app with external calibration...

The price of Flir One is prettly low right now:
http://de.camelcamelcamel.com/FLIR-ONE-W%C3%A4rmebildkamera-f%C3%BCr-Android-Ger%C3%A4te/product/B01478QCHU

[Fraser]

I agree with Frenky on this. The SEEK has potential but I still consider its core sub optimal. FLIR are old masters of thermal camera design and image processing. Love them or hate them, they still make decent thermal cameras, even considering the deliberate 'de-tuning' that they apply to limit performance in order to fit the market.

I have said it before but I consider the SEEK core to be little more than a prototype. Sadly we have yet to see them take the next step to produce a better image with less noise. Members of this forum have proven that it can be done using different software. The FFC shutter and optical block needs a redesign as well.

Be warned though, compared to my other thermal cameras, the F1G2 is the poor relation. Better phone mounted cameras are available if you can save a bit more for your purchase. Look at the Opgal ThermApp and i3 ThermalExpert detailed on this forum. Far more impressive imaging from these two offerings due to quality microbolometers, decent lenses and good software.

Fraser

[eneuro]

Seek has shutter behind lens so any dust, "glue" or vapour on lens (behind or front) will show on thermal images. And you can't lock the thermal range.

Maybe in original software, but who uses Seek with this useless app, while a few lines of OpenCV code opens new ways to see the dark 

[Trax]

I agree with Frenky on this. The SEEK has potential but I still consider its core sub optimal. FLIR are old masters of thermal camera design and image processing. Love them or hate them, they still make decent thermal cameras, even considering the deliberate 'de-tuning' that they apply to limit performance in order to fit the market.

could you please elaborate on that deliberate 'de-tuning' ?