djiuser_SUilMk3Qjz42 Posted at 12-10 14:08
Both the 3T and FLIR calibration function was run before I took the pictures.  The 3T took the first picture and the FLIR took the second less than 20 seconds later.  The unit never started, so there is no way that hot spot was not there for the 3T but was there later for the FLIR.

I guess I should have kept my mouth shut about these issues and let other folks like me believe the advertising and specs and spend their hard-earned money.

About which hotspot are you talking about ? 1.5deg C ?

About 47 minutes, please READ. It is cleary explain on DJI website :

[4] Measured with Mavic 3 Enterprise Series flying at a constant speed of 32.4 kph in a windless environment at sea level until the battery reached 0%. Data is for reference only. Please pay attention to RTH reminders in the DJI Pilot 2 app during flight.

The hot spot I was referring to is the large bright white area on the AC unit immediately to the right of the FLIRs measurement point. There is way more heat (more than the advertised 2 degrees or 2 percent whichever is greater) in that area than the 3T reported.

djiuser_SUilMk3Qjz42 Posted at 12-10 14:51
The hot spot I was referring to is the large bright white area on the AC unit immediately to the right of the FLIRs measurement point. There is way more heat (more than the advertised 2 degrees or 2 percent whichever is greater) in that area than the 3T reported.

This is where we realize that you do not yet have sufficient skills to interpret these images.

Thermal imaging is the "value" returned by the infrared band. Emissivity is the ability of a material to reflect or absorb infrared. (for simplicity).

I suggest you do a very simple exercise with your drone and your FLIR, on the most impossible surface to measure, a mirror.
Try filming the mirror and change the angle. You'll see it's magic the temperature will change.
And even worse! stand in front of the mirror and move. You radiate so much that you instantly heat the glass.
No of course it's a joke, it's just your infrared waves that are reflected in the mirror and captured by the camera.

For painting it is the same thing, the more it is varnished, the more it reveals parasitic values by reflecting somthing arround. This is why it is imperative to choose your emissivity well and know what you are looking at.

The example of glass also works with galvanized sheet metal, stainless steel sheet... a whole bunch of materials unsuitable for thermal imaging.
Here you can see a cold area on a roller where the bearing is KIA :


Hey no! it's just the reflection of the ceiling because the metal is polished :


You understand that if I change the angle, I change the reflection, I change the temperature value. Test on unvarnished wood, your hotspot issue will disappear.

djiuser_SUilMk3Qjz42 Posted at 12-10 13:31
I had the color palette in black and white originally (the default) for the wetlands shot.  There was nothing there for the person.  I then changed to the gradient to see if the person would show up better in the gradient.  Nope.  The person was standing in a open area between the trees with a line of sight from the person to the drone.  The reason I chose that spot is it is very similar to the rest of the environment behind it where I'm looking for hot spots at night.  
The drone was purchased for two reasons : (1) to fly over wetlands looking for hotspots and then get a visual picture of the hotspot.  After fiddling with it for weeks, it can't do that.  It can't even pick up a person when I know where that person is standing. (2) to get temperature readings of structures.  Even at 6 feet away on a table, it can't accurately do that, forget about being 50-100 feet from the target and hovering.

Love this, because there's a couple of things you've said which I can answer with pleasure.

"As for not knowing anything about thermal, do you wonder why I own a FLIR?"

I own a baseball bat. Does that mean I can play to a professional standard?

Second point. About the Xray vision you want through trees; Here's a video I did for a client who wanted to see how well they could spot animals on the land. Perfectly capable of seeing through trees, so long as the vegetation wasn't fully obscuring the camera and a perfect side by side comparison for you. The trick is to knowing what colour palette to use and also not thinking you're flying an XRay device.

LV,

Here is where I realize you don't understand a simple comparison (or are simply defending DJI's product no matter what result it provides).  Two pictures of the same surface taken from the same point at the same time allow one to compare the output of the two devices taking the picture.  FLIR is pretty much the standard for thermal imaging (I noticed you had. a couple handy).  The 3T is the comparison picture.  A simple comparison between the FLIR and 3T shows the 3Ts thermal pictures are poor compared to the FLIR.  That is unless you think taking pictures of metal is beyond the design capabilities of the 3T.

Cloud,

If the $5600 bat you bought is a gray hollow thin plastic bat, no, you can't play professional baseball with it even though the bat manufacturer advertised it was a professional quality bat.  Maybe the manufacturer designed it for whiffle ball and not professional baseball?  And if you did tests and showed your $5600 grey plastic bat can't hit a baseball 325 feet, I'd agree with you vs telling you that you simply don't understand how to properly hold and swing your grey plastic bat.

I also don't know why you insist X Rays are needed to get a thermal picture (and by the way, in case you never took physics, those are different parts of the spectrum entirely).  I told you there was a clear line of sight from the drone to the person standing in the trees.  I guess you need a picture to prove that, so see the attached.  The drone simply could not resolve the person in the open area under the trees.

Could the 3T find a person in the middle of a mowed field in the winder that started a fire?  Yep.  Can it find a 90 some degree person in 70-degree trees?  Nope.

djiuser_SUilMk3Qjz42 Posted at 12-11 05:48
LV,

Here is where I realize you don't understand a simple comparison (or are simply defending DJI's product no matter what result it provides).  Two pictures of the same surface taken from the same point at the same time allow one to compare the output of the two devices taking the picture.  FLIR is pretty much the standard for thermal imaging (I noticed you had. a couple handy).  The 3T is the comparison picture.  A simple comparison between the FLIR and 3T shows the 3Ts thermal pictures are poor compared to the FLIR.  That is unless you think taking pictures of metal is beyond the design capabilities of the 3T.

I will stop the debate here.  You don't want to understand, it's your choice.

Good luck

djiuser_SUilMk3Qjz42 Posted at 12-11 05:48
LV,

Here is where I realize you don't understand a simple comparison (or are simply defending DJI's product no matter what result it provides).  Two pictures of the same surface taken from the same point at the same time allow one to compare the output of the two devices taking the picture.  FLIR is pretty much the standard for thermal imaging (I noticed you had. a couple handy).  The 3T is the comparison picture.  A simple comparison between the FLIR and 3T shows the 3Ts thermal pictures are poor compared to the FLIR.  That is unless you think taking pictures of metal is beyond the design capabilities of the 3T.

I didn't insist you need Xrays, but if there's a tree in front of the infrared heat source, then it's not going to reach the camera. If there's enough space for the IR to break past the foliage then you'll see what's hidden in the trees, much like the video I posted above,

Going full circle back to your original point, maybe, just maybe, there is an issue with your drone and its camera system. I've said over and over that this isn't the right support channel.

If you are trying to use it for applications which are outside of your remit, then perhaps it's not the IR/camera that's at fault. I can safely assure you that using the IR camera and looking down through the gap in those trees from the spot you've taken that photo, that you would very clearly see a human-shaped object. If you can't, then there is either something wrong with the camera or you're pointing it in the wrong direction.

I do still think you've bought this as a novelty and currently feeling buyer's remorse. Do you actually have a commercial application for this drone?

CloudVisual Posted at 12-11 06:41
I didn't insist you need Xrays, but if there's a tree in front of the infrared heat source, then it's not going to reach the camera. If there's enough space for the IR to break past the foliage then you'll see what's hidden in the trees, much like the video I posted above,

Going full circle back to your original point, maybe, just maybe, there is an issue with your drone and its camera system. I've said over and over that this isn't the right support channel.

I don't think we need to struggle. As you wrote, he has remorse because he probably thought he could see through walls like in action movies.
It's a joke huh! but most probably we are not far from it.

This reminds me of technicians who take thermal images of electrical installations without removing the plexiglass protection.

I reassure everyone, no electrical box door has been abused during this thread.

Cloud,

The gimble is at -90 and the drone is flying a slow mapping pattern via FlightHub 2. at 100 feet until I got a hotspot, then taking manual control from there.  I was hoping to get thermal images similar to the one below.  My expectations were not FLIR ZenMuse level, but at least accurately find a hot.spot.  The 3T can't do that.

Both the visual camera images and thermal camera images on the 3T are poor.  A mini 2, air 2, mini 3, mini 3 pro, mavic 3 series (all but the 3T) can shoot RAW visuals.  This can't.  Multiple tests on multiple surfaces shorw the thermal sensor is nowhere near close to 2c absolute or 2% in a real world scenario.  Maybe in the same controlled lab where it flys for 47 minutes.....
I have no remorse as the drone is going back for a refund.  I'm just disappointed in the product not living up to the advertising hype.  I'll be trying a competitor's drone that lists for $2K less.

djiuser_SUilMk3Qjz42 Posted at 12-11 07:54
Cloud,

The gimble is at -90 and the drone is flying a slow mapping pattern via FlightHub 2. at 100 feet until I got a hotspot, then taking manual control from there.  I was hoping to get thermal images similar to the one below.  My expectations were not FLIR ZenMuse level, but at least accurately find a hot.spot.  The 3T can't do that.

Please stop here, and watch that, especially around 4:15  :


If your M3T is not able to do this, you do as CloudVisual told you, you open a ticket on DJI support to send it as warranty.

djiuser_SUilMk3Qjz42 Posted at 12-11 07:54
Cloud,

The gimble is at -90 and the drone is flying a slow mapping pattern via FlightHub 2. at 100 feet until I got a hotspot, then taking manual control from there.  I was hoping to get thermal images similar to the one below.  My expectations were not FLIR ZenMuse level, but at least accurately find a hot.spot.  The 3T can't do that.

That video you've posted is yet another example of someone not understanding how to use a thermal camera for a specific job. Why on earth would they use a colour palette which shows that the trees are slightly warmer than the ground?

Surely you can see the issue here. You need a palette which separates out cold stuff from warm stuff, not a gradient. Whitehot is the correct one to use for searching for hot objects in areas of foliage because it's easier to see a speck of white light in a sea of dark and black.

You're very keen to show us images of the location, so please can you take a dual image of the trees with you stood below. From the photo you've shown, I should be able to see you clearly through the trees and the thermal image will have a big white blob of heat surrounded by darkness.

djiuser_SUilMk3Qjz42 Posted at 12-10 14:08
Both the 3T and FLIR calibration function was run before I took the pictures.  The 3T took the first picture and the FLIR took the second less than 20 seconds later.  The unit never started, so there is no way that hot spot was not there for the 3T but was there later for the FLIR.

I guess I should have kept my mouth shut about these issues and let other folks like me believe the advertising and specs and spend their hard-earned money.

Here is me in the same spot (clear picture flying right over the branches) and my neighbor in his backyard (taken from higher as I didn't want to buzz him in his own yard).

djiuser_SUilMk3Qjz42 Posted at 12-11 11:05
Here is me in the same spot (clear picture flying right over the branches) and my neighbor in his backyard (taken from higher as I didn't want to buzz him in his own yard).[view_image][view_image]

You didn't need to keep your mouth shut about this, it's just that we've been trying to help you and it seems you've got a chip on your shoulder about this camera.

I need to see the RGB photo because I can't work out what's going on here, as there are big hot and cold changes on the trees themselves - was it sunny on wet trees by any chance?

The IR image doesn't look how I would expect it to, so I'm thinking that there is something wrong here. You can see my video above and they're not alike at all, but using the same IR sensor.

Normal 73-degree day, no rain for days. I can't get the neighbors small enough to upload.

djiuser_SUilMk3Qjz42 Posted at 12-11 11:56
Normal 73-degree day, no rain for days. I can't get the neighbors small enough to upload.

[view_image]

I cut the neighbor picture in half horizontally to get it to fit.  This is the bottom half.

djiuser_SUilMk3Qjz42 Posted at 12-11 12:00
I cut the picture in half horizontally to get it to fit.  This is the bottom half.

[view_image]

Try "Tint" instead of "White Hot"

LV_Forestry Posted at 12-11 12:18
Try "Tint" instead of "White Hot"

I'll do that later.  I did white hot as that is what Cloud requested.

djiuser_SUilMk3Qjz42 Posted at 12-11 12:23
I'll do that later.  I did white hot as that is what Cloud requested.

Your picture looks like taken from far away with a high zoom-in. But it is not normal to have such poor view of warm bodies. If tint is unsuccessful get ready to open a ticket.
If you use it with the zoom, it can have an impact in the middle of the trees. It's a digital zoom, to summarize it's a coarse raster transformation. There is a zoom level to find so as not to dilute the data too much.

LV_Forestry Posted at 12-11 12:25
Your picture looks like taken from far away with a high zoom-in. But it is not normal to have such poor view of warm bodies. If tint is unsuccessful get ready to open a ticket.

In the picture of me, I'm flying aright above 60-foot tall trees.  Any lower and I would have had a nice drone crash video to share

I'm around 200 feet over the neighbors...only went higher due to the noise the Mavic 3 makes and I don't want to bug him working in his own yard.  He's been out there cleaning up in the wetlands for over an hour so he should be quite warm.

djiuser_SUilMk3Qjz42 Posted at 12-11 12:34
In the picture of me, I'm flying aright above 60-foot tall trees.  Any lower and I would have had a nice drone crash video to share  

I'm around 200 feet over the neighbors...only went higher due to the noise the Mavic 3 makes and I don't want to bug him working in his own yard.  He's been out there cleaning up in the wetlands for over an hour so he should be quite warm.

"I'm flying aright above 60-foot tall trees"

Fly higher! the autofocus must be latching on to the treetops, which would explain the ...ty resolution on the ground.

Aperture: f/1.0
Focus: 5 m to ∞

LV_Forestry Posted at 12-11 12:39
"I'm flying aright above 60-foot tall trees"

Fly higher! the autofocus must be latching on to the treetops, which would explain the ...ty resolution on the ground.

The neighbor was shot from around 200ft and the mapping from FlightHub had it flying around 360 feet.

djiuser_SUilMk3Qjz42 Posted at 12-11 12:51
The neighbor was shot from around 200ft and the mapping from FlightHub had it flying around 360 feet.

You are very nice to me my friend, I would like to be able to help you.
Try to do dynamic detection (live) by flying in Tint and flying at a reasonable distance from the trees.

If you want you can send me via google drive for example, a set of thermal photos taken at Nadir. Preferably photos with a connection as if you wanted to do mapping. I will be able to compare with sets that I have and potentially detect a problem.
Then if you need help opening a support ticket, well, we'll help you. But don't expect any help from DJI on the forum, maybe compassion.

LV_Forestry Posted at 12-11 13:15
You are very nice to me my friend, I would like to be able to help you.
Try to do dynamic detection (live) by flying in Tint and flying at a reasonable distance from the trees.

LV,

I  will do that later.  The trees are 60 in that area.  How high do you want the drone?

djiuser_SUilMk3Qjz42 Posted at 12-11 13:22
LV,

I  will do that later.  The trees are 60 in that area.  How high do you want the drone?

200ft AGL seems to me to be a good compromise. Overlap not below 80-70%

Cloud,,

Earlier you said :"you have a chip on your shoulder with this camera".  I don't know if I would have used the word chip, but I get where you are going. I think I can respond more calmly to that now sp I will.

I bought a RC Pro for my Mini 3 Pro and Mavic 3.  It's an outstanding controller.  And the 3T basically comes with the sam,e controller.

I love the way my Mavic 3 flies (after tweaking the stick responses).  The 3T is basically a Mavic 3 with a different camera payload.  The Mavic 3 shoots great stills and videos.

So what it comes down to is the camera payload on the 3T and how it is performing.  And I still don't get why we can have RAW on the mini 2, air 2, mini 3, mini 3 pro, and every Mavic 3 except 3T.  And there again, I take that as a camera thing. And that isn;'t for you or LV, that's a DJI thing, and DJI)Mindy already said 'because we decided it wasn't needed.  I disagree with not having RAW on an enterprise-level drone, but it is what it is.

I need to look into your IR images a bit closer, but I think that there's something obvious here when you were saying that your JPG images were low quality.

When you took the photos, were you zoomed in? The only explanation for the JPG artefacts is that you took a photo using the digital zoom and you're saying that the image quality is bad.

As for the IR data, if I've understood this correctly, you've also zoomed in on this IR image too? I want to see the whole image, not a crop.

Cloud,

The images of me at 60 ft were not zoomed in.  I'm 95% sure it was on wide / IR.

The image of my neighbor was zoomed as I was at ~200 ft and I needed to show where he was visually.

I'm happy to retest tomorrow as I was out trying to get LV his tint shot and the drone went critical battery and returned before I could get the actual, photo.  I only got a screenshot from the controller at 200 ft per his request.  

M3T has now got a new firmware update. The logs specify this:

"Fixed issue for Mavic 3T: blurred infrared images during a Waypoint Flight." This is what you've been doing, so it'll only serve to fix your issue. As I've said over and over, you need to raise your issues with DJI because they will fix them. Your JPG images shouldn't look like that and your IR images seem to be calibrated incorrectly for whitehot. Your neighbour should have been clearly visible from that shot.

djiuser_SUilMk3Qjz42, from all that I have read above, my comments below:

The RAW v JPG argument doesn't really stack up, and the image quality differences you are seeing are simple to explain with the sensor sizes below:
Mavic 3E = 20MP 4/3inch sensor (1.333inch sensor size and TRUE 20MP)
Mavic 3T = 12MP 1/2inch sensor (0.5inch sensor size and claimed to be 48MP only because it has a Quad Bayer filter, but NOT a true 48MP)
Mini 3 Pro = 12MP 1/1.3inch sensor (0.769inch sensor size and claimed to be 48MP only because it has a Quad Bayer filter, but also NOT a true 48MP)
As you can see, of the 3 x drone sensors above, the M3T is the smallest in actual physical sensor size... as such, worse pixel resolution that the M3E or even the Mini 3 Pro.

I suggest sensor size be prioritised for your next purchase to meet your needs. Matrice 300 with H20T would be a good bet. (Forget the Matrice 30T... its got no better sensors than the M3T and it's only real advantage is it''s IP rating.)

As for the M3T SAR function... it's a miniature power house, but certainly will NOT keep up with a FLIR... just try strapping the weight of your FLIR camera under a Mavic 3 and see how well it lasts...

I like the analogy used by CloudVisual about battery drills... I would like a sand dune capable 4x4 that can go around a race track at the pace of a Ferrari, but, alas... there is no such machine... it's a constant trade off, and... How do you expect DJI to squeeze a 4/3 or even 1 inch sensor into a housing that has a zoom and thermal camera in there too...

I'd suggest you get happy with the amazing versatility of the M3T or drop $35k+ on an M300.

I feel your pain. I just bought the Mavic 3T as well, assuming it had everything the Mavic 3E has but with the additional $2000 thermal camera. Really fumming about it, but I need the thermal capability. DJI could at least kept the spotlight and tracking features. Its just software and they purposely removed it.

MQNine Posted at 2-20 07:06
I feel your pain. I just bought the Mavic 3T as well, assuming it had everything the Mavic 3E has but with the additional $2000 thermal camera. Really fumming about it, but I need the thermal capability. DJI could at least kept the spotlight and tracking features. Its just software and they purposely removed it.

i also just bought the m3t for solar farm mapping, and the thermal images are out of whack, heavily distorted fisheyed,  quality is garbage.  cant use terra due to the cuda requirement,  agisoft, webodm, dronedeploy and a few others struggle to stitch these images.  i feel totally broken at a premium enterprize product produces imaging worse than a gopro 3.

Thinking about buying the Mavic 3 T
but after reading this maybe not.

Has there been any improvement with the last few firmware updates?

Dave

DWA Posted at 3-6 11:04
Thinking about buying the Mavic 3 T
but after reading this maybe not.

Your best option is to go to a physical DJI product reseller who has a demo unit.  

Try it before you buy.  

Only you are able to judge whether the quality level of the sensor meets your requirements/needs.