Is the TE-M1 a software downgraded TE-Q1 ?

[Vipitis]

Oh man, that is quite the offer they are giving me there. But it's too risky for me as it's highly unknown if it will even work with my phone (although an upgrade is looking likely). It's still a hefty price for a toy... That might never get unlocked. An upgrade also makes my current imager pointless in a sense.

I got 4 days to think about this.

[guazan]

Good you could put some pictures made by TE-M1. Is that I'm in doubt about the flir one pro and TE-M1 would use it to repair phone plate. Thank you.

[lolli-us]

Has there been any progress on hacking the TE-M1 to a Q1 ?

For a while this thread looked very promising on delivering a great low cost thermal camera.

[bap2703]

I guess the camera isn't popular enough.

[lolli-us]

Strange... reading this forum I got the impression that Thermal Expert Q1s are considered the best performance/price cameras for under 1000€.

[frenky]

They are. But M1 version has not been for sale that long and not many people (on the forum) has it...

[Ben321]

The US store does not let you buy directly. You need to request a quote and contact them, and then arrange payment by email once you have decided to make a purchase (which will be done with paypal). There's no button that actually lets you buy an item (clicking the buy button takes you to the request a quote form). Even worse, the US store doesn't even acknowledge the existence of the TE-M1 camera. It doesn't appear anywhere in the catalog.
However, the European store shows all prices (in euros, not dollars), and lets you make a purchase just like Amazon or any other online store you normally buy from. Seems like their company must be located in Europe, as that version of the store is the best version of the store. And yes, even if you live in the US, you can purchase from their European online store. In the order form, where you can select your country, United States is listed in the dropdown list. I added an item to the "shopping cart" and started the checkout procedure just so I could if it would be possible to buy from the US. I haven't completed the purchase though, so I don't know if it will actually let me complete the purchase from the US though (and I don't actually want to buy anything from them now, so I won't complete the checkout form to test it). It does seem like it should work though.

[Requiem for a Dreamcast]

I purchased a TE-M1 during the sale. I'm not much of a programmer, so I'm not of much use on that end, but if I can be of any help, I'm available.

[nikoum]

Does anybody know why the TE-M1 camera is not mentioned in the thermal expert site (http://www.i3-thermalexpert.com) but only in the thermalexpert.eu site. It is rather strange to me.

[joe-c]

Hello,

Does anybody know why the TE-M1 camera is not mentioned in the thermal expert site (http://www.i3-thermalexpert.com) but only in the thermalexpert.eu site. It is rather strange to me.

interesting, its strange to me too.

i have a M1 with 30Fps option, and its a really nice device.
i add 2 images, just interpolated x2, no filters like sharpen or smoothing...

[mahony]

It seems like the ThermalExpert website is not being update regularly to say the least. They also got some nice OEM cores with the older 17µm detectors and the newer ones with 12µm should also be ready by now. The most up-to-date website currently seems to be the Austria-based EU shop.

@Joe-C: very nice images! Did you use the SDK in your software to capture them? I assume the 30 Hz give some very nice smooth video too. Did you try to get the full resolution out of it yet?

By the way and slightly off-topic: I played a little bit with bilateral filtering for noise reduction which seems quite capable compared to gaussian blur as used by the TE SDK. Using appropriate parameters this might even be suitable for on-the-fly processing, at least for the relatively small M1 / Q1 frame rate and image sizes. (mostly unoptimized C# code takes ~32ms on 240x180 and ~86ms on 384x288 while 640x480 already requires ~240ms).

[joe-c]

Hello

@Joe-C: very nice images! Did you use the SDK in your software to capture them? I assume the 30 Hz give some very nice smooth video too. Did you try to get the full resolution out of it yet?

Yes i use the SDK and no, i don't worked on the full resolution yet, sadly no time for that.
But yes, the 30fps are nice and smooth.

By the way and slightly off-topic: I played a little bit with bilateral filtering for noise reduction which seems quite capable compared to gaussian blur as used by the TE SDK. Using appropriate parameters this might even be suitable for on-the-fly processing, at least for the relatively small M1 / Q1 frame rate and image sizes. (mostly unoptimized C# code takes ~32ms on 240x180 and ~86ms on 384x288 while 640x480 already requires ~240ms).

I have also played with it, the aforge framework has a filter for it, i use it in my seek Tcam Dll sample. But sadly this only works with 8bit grayscale or color images.
sady i am not really familiar with convertig cryptic formulas from some papers to a working algorithem in c#.
if you have something interesting, just send me a mail
best wishes

[Vipitis]

My last exam is on 15th May, if I don't find anything else by then. I will join the M1 Club. Your image convinced me.

[Ben321]

Is this i3 company a Chinese company?

[frenky]

They are based in Korea:
http://www.i3system.com/eng/

[TooQik]

@mahony

I've captured some raw frames using your TETestApp code which I've made some modifications to in an attempt to capture a higher resolution from my M1. Are you able to take a look at them and see if there are any images please? If everything is as I'm hoping it will be, you should have some captures of me sitting at my PC.

I'm trying to work out how to convert the raw files to png myself, but this might take a while as I'm unfamiliar with raw image processing and have some learning to do first.

[frenky]

Your raw files are 227.328 bytes in size.
384 x 288 = 110592
2 bytes per pixel: 110592 x 2 = 221.184

I would say it looks really promising.

[frenky]

First few byte pairs and their values:
1f91 8081
1e3c 7740
1e0c 7692
1fe3 8163
1d4f 7503
1fe7 8167
1ee7 7911
1eba 7866

Edit:
byte pairs are probably switched so this are correct values...
0x65 0x1f 8037
0x91 0x1e 7825
0x3c 0x1e 7740
0x0c 0x1f 7948
0xe3 0x1d 7651
0x4f 0x1f 8015
0xe7 0x1e 7911

[Vipitis]

hm, I am not far into image processing yet, but I looked at some frames in there and rendered them into a 388x296 array because that fit with the size. But it looks like a good two thirds of the image is null values. So either I am not doing it correctly(likely), or there is indeed no data for the rest of the pixels and this needs further tinkering.

[Vipitis]

what I got from two frames

[frenky]

I didn't notice empty values before...
Sadly there is only 86.400 bytes of "non empty" data in a .bin file.
That is exactly 240x180 resolution with 2bytes per pixel.

So there is no way to get 384x288 images out of this frames...

[Vipitis]

If they don't spit out the full sensor reading just by asking them to....

then we need help from a really great android developer who can understand how they were able to change the window location via app.

Given that the European store is currently having a sale (the prices you would onyl get on eBay-Kleinanzeigen before) as well as new products(a desk stand). They will continue their support for all models for some time and there is no TE-2Q/-2M on the horizon.

I still wonder where there is not 30hz option for the TE-1Q (Plus). While there is a enhanced range option for the Q1 called Pro(and uses at least a different housing). It might be due to bandwidth, but that would be a awful design choice.

[mahony]

Hi,
I just check the last of the raw frames - and using a bit of phantasy I assume this might be you in front of your PC. The problem here is: this is raw sensor data without the so called "shutterless correction" that is done in the TE software. I do know how to apply the shutterless correction for the Q1 and V1 and I assume it is quite similar for the M1 (if not identical) but some of the required constants may be in different locations in the flash data compared to the Q1 and V1 (they already differ slightly too).

I think the fact, the camera just sends 240x180 pixel data is expected behavior. From the the rough look I had at the reverse engineered java code there is a small change the actual region of interest or (very small chance) also the resolution may be set by some initialization byte send to the cam…

@Vipitis: I don't think that bandwidth is a limiting factor. The V1 does 30Hz easly too over USB 2.0. I assume they just did not include the option in the initial design of the Q1. But I would hope they included support for M1/Q1 in 9Hz and 30Hz in the modified M1 hardware? But it seems like this is also not the case.

[TooQik]

Thanks everyone who took a look at the frames I captured.

I've already taken a look at the initialisation data being sent to the M1 when the Read Flash command is performed and trialed a couple of changes. The raw frames I posted earlier were from the test of the changes I made.

My thoughts are either one of two things are controlling the output from the M1:

• The initialisation data configures the frame size for output
• The camera is hardcoded in firmware to only output 240x180

If the output is actually controlled via option one, then finding the correct bytes to change should trigger an increased output resolution.

When capturing USB data, I can see initialisation data bytes values for 384, 288, 240 and 180, which obviously correspond to some resolution settings. My test above changed the byte values 240 and 180 to 384 and 288 respectively to see if the output resolution increased - it obviously didn't as evident by the image mahony has converted and posted which is only 240 pixels wide, so more tinkering is definitely required to see if I can find the correct values to change. 

I don't think the camera is hardcoded to output 240x180 due to the fact that there exists initialisation data that clearly defines resolutions, but I could be wrong.

[frenky]

If you tell me how to capture this initialisation data I can try it with my TE-Q1 to compare that with M1 init data.