Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2511.23383 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918222774665216 |
|---|---|
| author | Shrivastava, Rajit |
| author_facet | Shrivastava, Rajit |
| contents | This thesis presents an innovative framework for the automated detection and characterization of galactic bars, pivotal structures in spiral galaxies, using the YOLO-OBB (You Only Look Once with Oriented Bounding Boxes) model. Traditional methods for identifying bars are often labor-intensive and subjective, limiting their scalability for large astronomical surveys. To address this, a synthetic dataset of 1,000 barred spiral galaxy images was generated, incorporating realistic components such as disks, bars, bulges, spiral arms, stars, and observational noise, modeled through Gaussian, Ferrers, and Sersic functions. The YOLO-OBB model, trained on this dataset for six epochs, achieved robust validation metrics, including a precision of 0.93745, recall of 0.85, and mean Average Precision (mAP50) of 0.94173. Applied to 10 real galaxy images, the model extracted physical parameters, such as bar lengths ranging from 2.27 to 9.70 kpc and orientations from 13.41$^\circ$ to 134.11$^\circ$, with detection confidences between 0.26 and 0.68. These measurements, validated through pixel-to-kiloparsec conversions, align with established bar sizes, demonstrating the model's reliability. The methodology's scalability and interpretability enable efficient analysis of complex galaxy morphologies, particularly for dwarf galaxies and varied orientations. Future research aims to expand the dataset to 5,000 galaxies and integrate the Tremaine-Weinberg method to measure bar pattern speeds, enhancing insights into galaxy dynamics and evolution. This work advances automated morphological analysis, offering a transformative tool for large-scale astronomical studies. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_23383 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Identifying bars in galaxies using machine learning Shrivastava, Rajit Astrophysics of Galaxies This thesis presents an innovative framework for the automated detection and characterization of galactic bars, pivotal structures in spiral galaxies, using the YOLO-OBB (You Only Look Once with Oriented Bounding Boxes) model. Traditional methods for identifying bars are often labor-intensive and subjective, limiting their scalability for large astronomical surveys. To address this, a synthetic dataset of 1,000 barred spiral galaxy images was generated, incorporating realistic components such as disks, bars, bulges, spiral arms, stars, and observational noise, modeled through Gaussian, Ferrers, and Sersic functions. The YOLO-OBB model, trained on this dataset for six epochs, achieved robust validation metrics, including a precision of 0.93745, recall of 0.85, and mean Average Precision (mAP50) of 0.94173. Applied to 10 real galaxy images, the model extracted physical parameters, such as bar lengths ranging from 2.27 to 9.70 kpc and orientations from 13.41$^\circ$ to 134.11$^\circ$, with detection confidences between 0.26 and 0.68. These measurements, validated through pixel-to-kiloparsec conversions, align with established bar sizes, demonstrating the model's reliability. The methodology's scalability and interpretability enable efficient analysis of complex galaxy morphologies, particularly for dwarf galaxies and varied orientations. Future research aims to expand the dataset to 5,000 galaxies and integrate the Tremaine-Weinberg method to measure bar pattern speeds, enhancing insights into galaxy dynamics and evolution. This work advances automated morphological analysis, offering a transformative tool for large-scale astronomical studies. |
| title | Identifying bars in galaxies using machine learning |
| topic | Astrophysics of Galaxies |
| url | https://arxiv.org/abs/2511.23383 |