Chairs: Prof. Mihai Dutcu (DLR), Assoc. Prof. Zhongling Huang (Northwestern Polytechnical University), Prof. Zhe Zhang (AIR, CAS)
Abstract:
In radar imaging, processing
and image interpretation fields, the past traditional approaches based on
scientific theories have developed for years. Recently, the rapid advancement
of artificial intelligence technologies and the growing accessibility of radar
data have generated significant interest in the field of machine learning and
deep neural networks. The prevailing methodologies mostly revolve around the
utilization of powerful deep learning algorithms in the field of computer
vision, which have demonstrated notable success. Nonetheless, it continues to
confront the obstacles of limited annotated data, inconsistencies in the
physical predictions, and a dearth of interpretability.
Theory-guided data science aims
to investigate the realm of knowledge discovery and pattern recognition by
effectively utilizing the existing data while remaining cognizant of the
fundamental scientific knowledge. This session calls for papers aiming to
ground novel knowledge guided AI approaches, which integrate the domain
knowledge, electromagnetic theory, physical models of radar with deep neural
networks, for radar image processing and interpretation.
The broad topics of interest:
l Cascade or fusion of physical
model and DNNs
l Mutual substitution of DNN and
physical mode
l Physical model informed
interpretable deep neural network
l Radar knowledge guided model
initialization and transfer learning
l Radar knowledge inspired regularization
l Theoretical constraints for
model optimization
l Radar knowledge guided design
of model architecture
l Deep learning aided 2D/3D/4D
radar imaging
The applications include (but
are not limited to):
l Target detection, tracking, and
recognition
l Radar image simulation and
generation
l PolSAR image classification
l 2D/3D SAR imaging
l SAR imaging semantic
segmentation and change detection
l 3D/4D automative radar imaging
and applications
l Radar target characterizing
l Artificial intelligent
oceanography