- Project under Physics and Astronomy Club IIT Roorkee.
- Our goal is to automatically detect the constellations in an image based on the 88 constellations division by the IAU.
- Implemented an algorithm to detect constellations in the night sky.
- Used OpenCV with Python.
The algorithm first applies thresholding to the image, blacking out the stars below a certain threshold. Then it detects the 3 brightest stars in the image(the larger the star is i.e. the more pixels it covers, the brighter it is) and forms a triangle between them. Then, the angles of this triangle are measured and stored as the template for that constellation. For an unknown image, it repeats the same steps as above, and then compares the angles of the unknown image with the angles of the known templates, and predicts the constellation for which the error between predicted angles and template angles is minimum.
- Creating templates/dataset for all 88 constellations.
- Creating a UI using Streamlit(an open-source python framework for building web apps for Machine Learning and Data Science.).
- Optimizing detection algorithm to use more than three stars(implementing the paper Constellation Detection by Xiaoge Liu, Suyao Ji, Jinzhi Wang).