Abstract

Poster - Plenary

Machine Learning Assisted Multi-Frame Stellar Image Alignment and Change Detection

Reyhan Yağmur Pekşen,
Yeditepe University

Detecting transient celestial events in multi-epoch astronomical imagery is a critical task in observational astronomy, yet manual inspection of large sky survey datasets remains impractical at scale. This project presents an automated image analysis framework that detects spatial and photometric changes between pairs of astronomical frames through a combination of classical image processing and machine learning techniques. Stellar sources are extracted using connected-component analysis on adaptively thresholded images and filtered by area, peak intensity, and Signal-to-Noise Ratio to suppress spurious detections. Inter-frame alignment is achieved via affine transformation estimated through RANSAC over spatially matched star centroids, after which sources are associated across frames using mutual nearest-neighbor search in a KD-tree structure.The system is delivered as an interactive graphical tool supporting FITS and standard image formats, providing real-time detection, classification, and visualization of moving objects and brightness-variable sources.