Electronically Assisted Astronomy with a Jetson Nano

8k views have been reach !

Mr Huang, what is the prize for 8k ? I must say Aston Martin DB9 is really nice !

Jensen (may i call you Jensen), i know you are going to be very busy with GTC 2021. But when the GTC will be over, i hope you will find 5mn to say hello here !


Just a thought on extending this beyond astronomy: Aircraft (at least larger commercial ones) have some anti-collision abilities based on either transponders or radar. However, if optical anti-collision could be added…or at least a collision threat display…this would be far cheaper and smaller/lighter than something based on radar, and simultaneously not relying on the other guy having certain equipment (think of a flock of birds…they rarely comply with FAA or other flight rules). So far as I see this, your project is the start of such tech. The two things most needed to extend your current system would be depth knowledge and (especially) the ability to work from a moving platform (e.g., to work on an aircraft which is flying and maneuvering rather than holding still…but gyros could work in combination with a fixed nose camera to simplify this). When seeing something inspiring (and astronomy is inspiring) it is hard to stop with the first success.

Hello linuxdev,

that is something really difficult to realize. There are many problems to solve and experiments would need a plane ! It’s a very interesting project but i think it is far to complex for me !


Still testing Jetson Sky Survey.

This time, i have solved a problem i had when the ground was in the field of view. Wind can bring movements and generate false positive.
So, my software automatically generate a mask to hide ground structures and prevent detection in this area.

The result :

At 0:44, you can see a beautiful pigeon !



I have just seen GTC with Jensen. Really impressive.

Well, Jensen, no need for you to come here. You won’t have time for that and this is not really important.

The projects Nvidia is involved in are really great. Congrats. Really.

Jensen speach was interesting. In the same time, i am a bit less pride with the pigeon ! We are not in the same category for sure.

8k prize is not anymore required. I will wait for the 100k.


There’s another project for aviation: Instead of worrying about getting rid of ground clutter moving due to wind, use the camera to map wind patterns based on the movement. One would need to cancel out movements which are not from wind, perhaps starting with a local wind speed/direction indicator at the camera. I imagine that if we had a flight instrument which watches landing areas and produces an image similar to a FLIR, but designed to show a wind speed and direction contour (based on plants or other things moving as hints), this too would be a valuable safety tool for aviation. So many things the astronomy tech would be the basis for in other tech industries.

1 Like

Hello Linuxdev,

yes, astronomy image treatment could bring many things. I guess image treatment is very vell developped for military use for example. A friend of mine works for military projects (he uses Jetson products), mainly survey and target seeking. There is also automotive applications. But from what i can see, it is quite impossible for somebody like me to collaborate with them.

I can see 2 possibles explanations for this :

  • my work is crappy
  • pros only work with pros

A guy like me gets 2 major problems :

  • time (i have my job, many things to do with my personal life)
  • money (to get very good results, you have to get very good equipment; software can’t make miracles with crappy captures or equipment).

Thinking about this, those 2 points are not really a problem if you admit your results will never be as good as they could.

I have many projects i could manage with my equipment (which is not so bad). But i have an other problem (so many problems in fact) : the test platform.

Writing an algorithm is ot really complicated. But when you want to test it, you may have many problems. For example, i would like to make night vision for automotive. This means i must make capture using my car at night. You should think it is easy. Well, it is not. Many parameters to manage with a car which is not made for that.

If i want to make something with a plane, well, i have to find a plan to put my equipment in. Etc, etc.

So, it is really hard to get the test plateform to test your idea (idea is not really hard to find) and validate it. And most of time, you work alone without real support and this is not really easy.

I must admit i have choosen the wrong job (but when i made my choices long time ago, it was the good choices; today, it seems my choices were not ver clever). Even with this bad choice (considering everything), i can make some small projects which are interesting (for me so far) but i will never be able to do what i am able to do. But it is not a drama. I just have to make things i know i can manage considering time and money.

That’s why i say very often to people : lear as much as you can, don’t focus too much on objective, try and try again, work to find the best environment which will bring you the best working conditions, try, try and try until things work


And other test, tracking satellites at night. Quite long test to find targets. The sky transparency was very bad, unfortunately.

Many many satellites above us. As i know the observation place, the look at direction and the field of view, i can match what i see with satellites positions (using Pyorbital and NORAD informations). I will probably do that but later because this needs some work.


Back to the Moon. Colours of the Moon’s soils using an old map of the colours of the Moon i made.

Colours improvement with the Jetson Xavier NX and my software :

I think it is really magical to be able to retrieve informations in the camera signal that eye can’t really see.

It’s about the same thing with the bird/planes or satellites detection : most of time, the eye can’t focus on very small detail / small information.


1 Like

The part I like most about the tracking one is that it works just as well with a lot of simultaneous targets.

That last one though (the one with the color enhancing) would make such nice art it is something I could see printed with high quality and hung on the wall. Kind of breathtaking!

Hello Linuxdev,

for now, i have not think about how to use target detection. Maybe i will use this information to calculate trajectory, altitude of the satellite (using angular information considering exposure time), etc. I don’t know. But detection works quite well.

For the Moon, for sure, i huge print could be really cool. I will consider to make one for my future house !


Nothing really new unfortunately but just a small video of the Moon, using a Canon FD 135mm lens, the colour camera and the Jetson Xavier NX (control, capture, treatment).

Even with a small focal length, the result is not that bad.


1 Like


A small example about treatment.

The setup is exactly the same : Celestron C9.25 and IMX178 colour sensor camera.

Result i get 4 years ago :

Result i get now :

What the pictures look like is not the major point. For sure, we can find one picture is more beautiful than the other but this is subjective opinion.

The key point is the number of colours we get. The more colours i get, the more informations i get about the soils.

The only difference with those images come from the treatment. Good tools can bring good results.


Regardless of how subjective one’s view is, I’m certain you could start a space art business. That looks fabulous, and if it were high enough resolution for a larger picture to hang on the wall, you’d have some art fans.

Hello linuxdev,

Yes, those pictures are close to art. I sell some but to be honest, my business dies not work at all.

Since 2018, my sells are about 120$. Yes, it is ridiculous. But, if i consider i am the baddest artist in the world, i think i could be proud of this.

My sells simply don’t work. I don’t know why. I think i will have to keep my job !

If an art gallery want to sell my picture in New York, i am ok !


If I were you I’d create YouTube videos of the art, with links below the video for purchase. The Jetsons certainly help that art, and I suspect that the art itself would eventually bring more attention to the Jetsons as well. If you made videos, then the lossless high resolution art would not be exposed, but the “obvious” art value would be demonstrated.

I will think about that. Thanks for the advices !


The weather is so bad here so i tried to make a video which is supposed to be a kind of summary of my work about colours of the Moon.

Here it is :

For sure, i am not really aware with video editing. Anyway, if you like it, please share it. This will help me to get in touch with many people, hoping they will be interested. Maybe tis will help me to get better quipment to get better results.

Many thanks !


I like all of that video, but the part which looks like close up flying near the surface in 3D is just amazing. Amazing is an understatement, the “3D flight” is beyond beautiful.

Hello linuxdev,

yes, the fly over surface is cool. I did that kind of things a few time ago. I used digital elevation model of the Moon provided by Lunar Reconnaissance Orbiter and Kaguya satellite (you can find them in the USGS web site).

I was also very pleased with the “real” flyover video treatment. The colour video was good enough to bring informations about colours of the soils. That’s too bad i can’t get the RAW data to process them.

So many things we could do and so few we really do. But the actual results are quite good. Still have to work !