Hello Fellow Yuneec Pilot!
Join our free Yuneec community and remove this annoying banner!
Sign up

Project: Building an alternative to CGO-ET

Joined
Sep 28, 2020
Messages
1,287
Reaction score
507
Age
55
Location
Kozloduy, Bulgaria
Don't worry. The competitors are in the same situation.


As I see, the microbolometer for this EXPENSIVE camera looks the same as in the high-end HTi cameras. Due to the nature of the VOx sensors, I'm not sure radiometric with constants is possible at all. Differential measurement will be done on the fly, but real data not. More than. there is confusion in the mean of "radiometric". FLIR uses a trick to say "Temperature Measurement can be defined as a collection of techniques designed to measure electromagnetic radiation, including the spectrum of visible light. The terms “thermography” and “Temperature Measurement” are sometimes used interchangeably, since radiometric measurements are essentially reading the intensity of thermal radiation."
from here: The Benefits and Challenges of Radiometric Thermal Technology

Further, in the same text, they say about the importance of all spectrum to see if you want to protect your wires. ;) Good is Yuneec forget to plagiarise this too.
 
Forget to say. To prepare a good picture, the bolometers data is deeply transformed with sophisticated algorithms to produce a shiny image. The data BEFORE this task is important to be stored untouched. Unfortunately, this is not made as you can see.

Last few weeks I'm on the IR/Thermal wave. As much as I read and see, the VOx with a germanium lens is more than enough for drone operators. Which I'm trying to prepare for us.

Bit more from the DJI
 
Last edited:
  • Like
Reactions: 7cyclops
Because I have had some success, I should ask for some help. ;)

I can stream via V4L2 the following data:
- Detector original value output;
- 16-byte parallel image data output;
- 16-byte parallel image data and temperature data output.

There are more modes, but the most promised is the last one. So, the question. How to separate, obtain, or something like this one, the temperature data?

Linux library or source is preferred, but just explanation is welcome too.

The camera is Xtherm II if this does matter.

me.png
 
Last edited:
  • Like
Reactions: Steve Carr
MK58 emulated by Pi. YUYV to H.264 convertor is a tragedy, still have no idea how to use temperature metadata but the toolchain works itself.

DSC_0279.JPG
 
  • Like
Reactions: Steve Carr
ScreMe.png

The camera itself is a great piece. Step by step the production will be great too.

The lag to almost whole process from the object to the ST10+ screen is around 0.7 seconds. Image preparing with metadata is today's topic.
 
Because the things further are tied with a lot of work and are time-consuming, I'd like to ask for a few files with R-data from the best on the market cameras. Of course, the camera itself will not be the same, but the data presentation structure can be. Some user manuals, some pictures of what is the best to watch on the screen, and so on...
 
ST10+ native frame format is YUV420p. With them, the lag is lower than the original MK58. Works with packed YUYV too, but is not so happy, and do things slower. On the transmitter's side is faster a lot.

Should be prepared the same chain for CGO3+.

With the R data still get nothing. Not so big deal because the calculation is better to be done on the ground. The camera proposes all necessary for this in a very convenient format.

I found the digits for the IRONBOW colour scale. White hot and black hot is known, isotherm is possible to prepare. If someone has digits (array) for another used scale and want to share them...
 
  • Like
Reactions: Steve Carr
This is a very interesting project. Sadly I have no idea how any of this works but I'm glad that you are making progress.
 
I can explain the basic steps if someone is interesting to play with these things. A thermal camera is only one of the possible sources. About to share something ready-made I can't. All is complicated, will not work on a foreign desk from the first time, I'm not ready and don't plan to be ready to give explanations and so on.

In a very simplified explanation:
Two sources of signal UVC camera and Pi camera. About the Pi camera the progress is zero, just didn't attach it jet. The next is the FFmpeg from YUYV422 to YUV420p packed as H.264 via v4l2 to the loopback device. The next is transmitting as RTSP to 192.168.110.1:554/cam1/h264. For AP I'm using the cheapest 5G router on the market connected to the ETH0 to not mess the primary WLAN on the Pi. This will be, for now, one of the steps will be to add an RTL8720DN module to a Banana Pi M2 zero.
 
Last edited:
They have a software called "IR_PCTool" downloadable for radiometric post-measurements on their IR images (Autel IR PC Tool (EVO II Dual 640T Only)):

Where to ask for a few jpegs like these, to be compatible with the concurrency? Comments under the video are turned off, so no way to ask there. To ask the video producer via their corporate site is a bit more than normal behaviour.
 
Last edited:
YUV422p is the last decision. YUV420p is not much faster but the colours are like in "Gone with the Wind" movie. For white warm and black warm palettes, of course, doesn't matter but it better is to minimize transformations. For the still picture, I'll switch mode to 14-bit temperature data, which looks green and awful. ;)
 
All in sync and proportional.

The next is the temperature data on the screen.DSC_0005.JPG
 
Due to the specific of the lens, will be necessary to prepare some kind of distance meter. Manual focus in the middle of the air is difficult and not necessary for now. Simple focus on the ground to a known distance, with a possibility to position on the same distance for inspection, looks to be enough.

Still looking for:
- pallette data arrays;
- some kind of Rjpeg, better from Autel;
- recommendations from the thermal guys on how should look the picture on the ground station screen. Take into account, the pallette is always dynamic.
- I have some ideas about video and still pictures captured by the pan slider on the ST10+ (the right one when it is connected to the H 480).

- probably the idea to prepare it as a CGO3+ simulator will be rejected. Nothing from the menu on the ST16 will be useable in this situation and so on. To people, who want to make a fake CGO3+ camera. Just look for one more feed in the communication stream between the camera and the ground station instead of the CGI commands to the camera and the RTSP downlink. Redundancy data to the copter I mean...
 
50 000 frames sent, zero dropped zero sync issues. Will not touch it anymore in regards to "get and send".
 
  • Like
Reactions: Steve Carr
I can explain the basic steps if someone is interesting to play with these things. A thermal camera is only one of the possible sources. About to share something ready-made I can't. All is complicated, will not work on a foreign desk from the first time, I'm not ready and don't plan to be ready to give explanations and so on.

In a very simplified explanation:
Two sources of signal UVC camera and Pi camera. About the Pi camera the progress is zero, just didn't attach it jet. The next is the FFmpeg from YUYV422 to YUV420p packed as H.264 via v4l2 to the loopback device. The next is transmitting as RTSP to 192.168.110.1:554/cam1/h264. For AP I'm using the cheapest 5G router on the market connected to the ETH0 to not mess the primary WLAN on the Pi. This will be, for now, one of the steps will be to add an RTL8720DN module to a Banana Pi M2 zero.
if you can transmit to 192.168.110.1:554/cam1/h264 and view it on the ST10, can you transmit to rtsp://192.168.42.1/live and view it on the ST16 by selecting the CGO3 camera?

Dylan
 
Still not tested, but probably this will be not successful. The ST16 checks the camera type before opening the stream. At least it expects proper CGI responses. Now I'm doing some other things. When I finish them I'll add new values on the router and test them. On ST16S with H+ firmware the trick will go because it eats everything. This all is not on the near agenda but I'll try somewhen.
 
  • Like
Reactions: Enrico Jorysch
- some kind of Rjpeg, better from Autel;
IAW the last update from Yuneec, some files from 520E with thermal data will be welcome too. In any case, I'll write a raw data binary file, which can be read with some dedicated software. The possibility to transform it to another form, accepted from the majors is not so bad idea.
 
The RPi NoIR ver. 2 camera is wasted money for nothing. Hard recommendation to not buy it at all. Very slow, see exactly nothing in the darkroom, where I can see everything, the picture is jumping and pixelized... As was added, five-time faster was removed from the system.
 
The final run is here. I have some difficulties with correction coefficients to the measured points in the frame. Can someone wrap a library to put fake data to see what these clever guys did?

PM only! I'm not authorized to share these things with the public, but if someone can help... The question is active for today-tomorrow. If the thing will take more time, I'll do it alone.
 
The question about the temperature calculation is closed. Now I'm waiting for some electronics from China to prepare a prototype. The T2Search camera has a lot of useful capabilities, but most of them are usable only in interaction with the ground station. I mean point selection to see the temperature, palette change, make a shot, and so on. Because the idea is not closed to only this camera, the best approach will be to add a camera script module in the existing FlightMode.apk in the ST16. After that everyone can prepare its camera with the functionality which he/she wants.

I'm not a programmer at all but can write almost everything on any programming language if I have a program skeleton. I'm sure, there are some android guys, who can give an initial push on this project.
 
  • Like
Reactions: Chris Koch

New Posts

Members online

Forum statistics

Threads
20,955
Messages
241,587
Members
27,284
Latest member
csandoval