| |
 |
| Paper: |
A deep learning approach to the discovery of anonymous Kuiper Belt Objects in wide-field imaging |
| Volume: |
535, Astronomical Data Analysis Software and Systems XXXI |
| Page: |
95 |
| Authors: |
Lee, A.; Kavelaars, J. J.; Teimoorinia, H.; Fraser, W. C.; Ashton, E. |
| Abstract: |
We demonstrate the use of deep learning methods to detect Kuiper Belt Objects (KBOs) in wide-field imaging data sets. Synthetic KBOs were added to time series observations. Those observations were divided into sub-image pairs to create a two-channel training set. Convolutional neural networks were applied to extract useful features from the sub-image pairs and predict the existence of KBOs within the training set pairs. ImageNet classification architectures were effective at predicting the existence of a KBO within a sub-image pair. Our deep learning approaches effectively detect moving sources down to a signal-to-noise ratio of 5 per image in self-validating tests. Our goal is to apply this deep learning-based detection method to arbitrary time-series observations. |
|
|
|
|
|
