scopenet

Intro

ScopeNet a low-cost hardware, firmware and online service allowing hobbyist astronomers worldwide to automate and share their telescopes online.

Astronomical observations requested from users globally are optimally dispatched to the most appropriate telescope(s) worldwide. Image processing algorithms produce crisper, better quality images by stacking and registering photographs of the same object taken from multiple scopes in the network, enhancing the performance of a set of low cost recording devices.

The clever part of this project is to move all the complexity in to open source software run locally and online to reduce the hardware costs.

Progress so far:

• We have written a python script that given an astronomical object of interest and the current time and location of the shared telescope, obtains the azimuth and altitude of that the telescope needs to orient to.

• We have demonstrated Arduino controlled servos and stepper motors for controlling telescopes.

• We have demonstrated how using a low cost web-cam with a 3D printed mounting and vision processing software can produce a stable and crisp image.

• We have begun developing a universal API for interfacing these devices all over the world. There is a simple protocol between the API and the actuators which enables people to use whichever type of motors they want.

• We have developed a web portal for the general public to access past images and request images.to be captured.

For the more complex background computations a normal desktop or laptop computer is used. Using Python and its libraries we will make a cross platform application that can just as easily be expanded as it can be shared. The intensive image processing computations are taken care of by this application – this consists of correlating the images and stacking them, to remove noise out of the final result. When the image has been captured and processed by the client it is sent to the server and can be viewed by the whole world.

The server is PHP based with a MySQL back end. The interface is intentionally simple allowing the most novice user to find the planet or distant sun they want to view, see past images that have been taken, and request a new image be taken by someone’s telescope if desired.

Plans for A.I. ScopeNet automation

ScopeNet will be continuously self-optimising, learning which locations and conditions produce the best photos as judged by its users.

Each telescope intermittently requests a task. This is currently allocated on a round robin approach. We intend to build a powerful ScopeNet A.I.. The network will be trained with reinforcement learning to continuously improve the quality and variety of images taken. Based on the accumulating experience and user-awarded scores for past jobs, the network will be able to increasingly well predict the value of each job request on the network, and select the best jobs and best telescopes/locations from which to make those observations.

Project Information

Resources