Core Concepts
The core message of this article is to develop an efficient framework based on a hierarchical visual Transformer with a sliding window technique to search for strong lensing systems within entire multi-band images captured by the China Survey Space Telescope (CSST).
Abstract
The article presents a comprehensive framework for detecting strong gravitational lensing systems in multi-band images from the China Survey Space Telescope (CSST). The key highlights are:
The authors generate simulated CSST images containing strong lensing systems using an advanced extragalactic catalog (CosmoDC2) and a CSST image simulator. This simulated data is used to train and evaluate the detection algorithm.
The data pre-processing pipeline includes image cropping, deconvolution using a deep learning-based PSF-NET, and grayscale transformation using the asinh function. These steps enhance the visibility of faint lensing features.
The detection algorithm is based on a hierarchical visual Transformer with a sliding window technique. This architecture can effectively capture both local and global features of strong lensing systems, even in images with varying numbers of channels.
The detection framework achieves precision and recall rates of 0.98 and 0.90 respectively on the CSST mock data. When applied to real observation data from the DESI Legacy Imaging Surveys and Euclid Early Release Observations, the method discovered 61 new strong lensing system candidates.
The authors identify limitations in the simplified galaxy morphology assumptions within the simulation, which can lead to false positives. This underscores the need for further improvements to handle the complexity of real observation data.
Overall, the proposed framework demonstrates a promising approach for efficiently detecting strong gravitational lensing systems in the upcoming CSST multi-band imaging survey.
Stats
The article does not contain any explicit numerical data or statistics. The key figures and metrics reported are:
Precision and recall rates of 0.98 and 0.90 respectively on the CSST mock data.
61 new strong lensing system candidates discovered in real observation data from the DESI Legacy Imaging Surveys and Euclid Early Release Observations.
Quotes
The article does not contain any direct quotes.