Electronically Assisted Astronomy with a Jetson Nano

Hello,

clouds are back. As i have some free time, i tried something new with my colour camera : try to change it into a monochrome camera.

My sensor (IMX485) is a bayer sensor (to get colour information).

Colour is extract from 4 pixels. In my case, the order is :
R G1
G2 B

So, 1 red pixel, 2 green pixels (G1 & G2) and 1 blue pixel.

For sure, photosite doesn’t get the entire signal (only red, or green, or blue). That’s why colour sensor are about 1/3 sensitive as mono sensor (which gets the entire signal).

Most time, you can get grey image with a colour sensor. The camera applies a small calculus like :
Luminance = Red0.2126 +Green0.7152+Blue*0.0722 (or something like that). In that case, the result intensity is quite the same result you will get with the colour image.

Let’s try something different :
Luminance = R + (G1+G2)/2 +B

Quite simple. So, let’s see the result.

First, the classical colour image (gain 100, exposure 11ms) :

Now, the classical grey transformation. The luminance is the same :

Now, the new transformation, which is supposed to raise the luminance :

As you can see, it works quite well.

What would be the camera setting to get the same luminance with classical grey transformation ?

To get the same result, the exposure time must be 43ms (compare it to the 11ms exposure time with the new transformation).

It is the same with colour image :

So, with the new grey transformation, i can get much more signal than the normal signal i can get with the colour sensor. I must say i always use the sensor information so i don’t bring false information).

With this method, i will be able to make deep sky survey video with (much) lower exposure time (for example, 100ms exposure time instead of 300-400ms exposure time) and keep my colour sensor instead of buying a new monochrome sensor camera.

For sure, i will lost the colour. As deep sky survey means quite long exposure time, i guess i could colourize my monochrome capture with the RAW colour image. Just have to verify that.

An important thing to say : monochrome sensor is better for resolution and details (1 pixel for the mono sensor, 4 pixels for the colour sensors). I can’t make miracles each time.

Well, that’s all. Clear sky !

Alain

Hello,

quite a sad day today. 3 years ago, a achieved my 2 axis motorized mount (see the first post of this topic).

It was interesting project. Several months for conception, 3 weeks to make the mount, several months to make it work and improve the software (mount control).

Now, i must say the mount is a bit old and i have new issues each time i use it (mechanical and electrical problems). I had to face the truth : my mount needs retirement.

So, today, i decided to disassemble it. This is the end.

So pictures to say good bye to my mount :

https://scontent-cdt1-1.xx.fbcdn.net/v/t39.30808-6/287632758_1774097952941548_1653363584799481301_n.jpg?_nc_cat=101&ccb=1-7&_nc_sid=730e14&_nc_ohc=p7-CyfI9wYMAX-Uj137&_nc_ht=scontent-cdt1-1.xx&oh=00_AT9u_-O9V1uAePa2mDUiL7X5P3lLfC_1MZATtQEBq1Q94w&oe=62ABF494

And the last picture : my mount totally disassembled

It took me 3 hours to totally disassemble it. So sad.

If i had time (i mean no need to work), i would have make a new mount (more simple, more accurate, etc.). But as i don’t get much free time, i will probably buy a small altaz mount (like Skywatcher AZ-GTI) in order to get time to improve my treatment software (JetsonSky) and make sky survey.

So long my mount. You were really cool.

Alain

1 Like

Hello,

some time ago, i tried to get automatic satellite detection with JetsonSky. It worked quite well but it was very sensible with noise and background illumination.

I did not give up.

Looking around, i founded an promising OpenCV function : SimpleBlobDetector

So, i tried to use it in JetsonSky and i must say it works really well.

An example :

This function will bring me new opportunities. I am quite happy with it.

For sure, those opportunities will need more CPU and GPU power. AGX Orin looks more and more sexy.

Have a nice day.

Alain

Related to previous post, an other test.

First part : Hercules great cluster (M13)
Second part : classical wide field survey

Alain

1 Like

Hello,

some new improvements of satellites tracking. This time, i can get the trajectories. Still some improvements to come.

This new result is really promising.

Alain

This is rather cool. I am curious though, there were a few which did not track, but went smoothly across most of the field of view. Do you have a guess as to why those were skipped, e.g., not bright enough?

Yes, some satellites are not detected. The main problem is the noise. If i catch low brightness satellites, i will also catch many false satellites (mainly noise). I will try to improve this.

Alain

1 Like

Some new tests.

The method is interesting but i still get issues.

The problem is always the same : noise. To catch satellites, i need to set the camera gain at high value. So, i got noise.

I need some time to get improvements. But the test result is quite cool.

The treatment is really heavy and hard to be made with GPU. A big CPU is required.

Alain

Great job with the detections of satellites.

I’m interested on plate solving to know the center coordinates of the video. Did you implement astrometry in real time video?

Hello Carlos,

astrometry is something i am interested in. I plan to use it but i must say it is something quite complicated.

Alain

I have received my new mount : Skywatcher AZ-GTI.



Quite the same as my old home made mount but i must admit it looks … better.

I will be able to use many interesting software. I guess this new mount will be cool. First test (very quick test) is OK. Just have to wait for clear sky now.

Alain

3 Likes

Hello,

today, i made daylight test with the new mount :


Everything is ok … with my laptop under Windows.

I still have issues with linux (Jetson Xavier NX). The mount control with linux system is less easy than Windows system. I still have some work to make the mount work with linux system.

With windows system, i can control the mount with Skywatcher app and i can also get the pointing position of the mount in Stellarium (i can also point a target in Stellarium and set the mount pointing position on this target). It’s really cool.

I hope i will be able to get the same with Jetson Xavier NX.

With this system, it will be very easy to make a deep sky journey with specific targets. From what i have seen, i can bring mount target coordinates in my software (JetsonSky). If i can do this, i will be able to show pointing coordinates in the videos (azimuth, elevation).

Maybe i will also be able to make satellites tracking.

With this system, i will use my (very) old Canon FD 135mm F/D2.5 lens.

From what i know, something is coming from Santa Clara here in Bretagne. Something with strong CPU and GPU. This extra power will bring me new possibilities (heavy multi task). In that case, i will have to think about fix installation (running many software is easier with fix installation and at least 2 screens to get comfortable conditions to watch the sky).

Just have to wait.

Alain

Hello,

beautiful surprise today when i went back home from work : a very beautiful box (black box) from Nvidia Santa Clara.

Inside the box, an amazing Nvidia Jetson AGX Orin dev kit (the big one).

Some pictures :




The AGX Orin is now working (software update). Now, i will try to get a good software configuration in order to get my programs running without problem. With this AGX Orin, i think many new things will be possible.

I want to thanks Dusty_nv ! He supports me for years now and he is the coolest guy who works for Nvidia.

Dusty, if you read this, many many thanks !

Well, i must go to see if AGX Orin software update is ok.

Clear sky !

Alain

3 Likes

Hello,

i have just finished to configure the AGX Orin (Python libraries, pycuda, camera SDK etc.).

Well, JetsonSky works now with AGX Orin. No problem.

From what i can see, AGX Orin is much faster than my laptop (i7-8750H & GTX1060). That’s crazy. Really.

The AGX Orin at work :


Alain

1 Like

Hello,

as Jetson AGX Orin is very powerful, i can use it as a standalone computer for many uses.

Looking at the sky is interesting but is also very interesting to know what you are looking at.

So, i have to install some software like KStars. KStars is freely licensed, open source, cross-platform Astronomy Software by KDE.

It provides an accurate graphical simulation of the night sky, from any location on Earth, at any date and time.

If you are interested with this, you can take a look here :

KStars is also supposed to be open to astronomy equipment using INDI drivers :

In order to install KStars, you can do this :

sudo apt-add-repository ppa:mutlaqja/ppa
sudo apt-get update
sudo apt-get install indi-full kstars-bleeding

It works for me except some issues i do not really understand.

I guess KStars would work with Xavier NX and Nano but i am not sure. Ubuntu is not the same version.

The Jetson AGX Orin running KStars :


Alain

2 Likes

An other planetarium i use very often : Stellarium

You can find some information here :

The problem : Stellarium must be compiled from source in order to run with Jetson computer.

The howto is here :

I have just compiled the source code and now, AGX Orin can run Stellarium :


I do love this soft. I am very happy AGX Orin can run it !

Alain

1 Like

Beautiful apps! Makes me kind of wish I had large high res monitors on the walls just for space artwork. I’ll probably do that right after I make my first billion…

1 Like

Hello linuxdev,

i hope you will earn your first million soon !

It is really cool that AGX Orin can support that kind of software. This will allow to have several software running at the same time and for astro observation, it is really useful.

For now, the only weak point i have found with the AGX Orin is that there is only 1 video output connector. AGX Orin can easily manage 2 monitors and this would have been helpful.

Alain

1 Like

Hello,

still working to get the good setup for AGX Orin.

Back to KStars which is really interesting.

It could more suitable to build KStars for the Jetson Nano/XavierNX/AGXOrin. To do this :

sudo apt-add-repository ppa:mutlaqja/ppa
sudo apt-get -y install build-essential cmake git libstellarsolver-dev libeigen3-dev libcfitsio-dev zlib1g-dev libindi-dev extra-cmake-modules libkf5plotting-dev libqt5svg5-dev libkf5xmlgui-dev libkf5kio-dev kinit-dev libkf5newstuff-dev kdoctools-dev libkf5doctools-dev libkf5notifications-dev qtdeclarative5-dev libkf5crash-dev gettext libnova-dev libgsl-dev libraw-dev libkf5notifyconfig-dev wcslib-dev libqt5websockets5-dev xplanet xplanet-images qt5keychain-dev libsecret-1-dev breeze-icon-theme

mkdir -p ~/Projects/build/kstars
cd ~/Projects
git clone Education / KStars · GitLab
cd build/kstars
cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=RelWithDebInfo ~/Projects/kstars
make -j16
sudo make install

This works for the AGX Orin. I guest it will work for the other Jetsons.

Using INDI drivers, it is possible to control the mount with KStars. This means you know exactly were the camera is looking and what you are looking at.

I made some tests and everything works fine.

Here is the complete setup (indoor tests):

On this screen capture, the mount (Skywatcher Alt-Az) is looked on the Bode’s galaxy (M81). You just have to choose the target with KStars and send a GOTO command to the mount and that’s it. If you manually move the mount, KStars will get the mount position and display it on the map.

Just need now to make some tests at night. I will also try to get the mount position in JetsonSky to put those coordinates in the video.

Alain

I do see kstars available on the Orin from the standard “apt” install. I did add this and it works, so I am curious if there is a reason why it might be better to build from source?