ISAR Dataset for the Recognition of Conical Targets with Micro-Motion

Area of Science:

• Radar signal processing
• Deep learning applications
• Electromagnetic scattering

Background:

• Deep learning for ballistic micro-motion target recognition is a growing field.
• Current research relies on privately simulated datasets, hindering objective performance evaluation.
• Variations in electromagnetic simulation methods significantly affect data, even with identical parameters.

Purpose of the Study:

• To address the lack of public datasets for micro-motion target recognition.
• To demonstrate the impact of electromagnetic simulation details on dataset generation.
• To introduce a novel, publicly available dataset for performance benchmarking.

Main Methods:

• Development of a unified process for generating target radar echoes.
• Utilizing fully polarimetric Inverse Synthetic Aperture Radar (ISAR) principles.
• Inclusion of aspect angle sequence and static electric field data.

Main Results:

• Simulation method details critically influence micro-motion target data.
• The proposed unified process provides a consistent framework for echo generation.
• The ISAR Micro-Motion Dataset (IMD) offers a valuable resource for the research community.

Conclusions:

• A publicly available dataset is crucial for advancing micro-motion target recognition research.
• Understanding simulation nuances is key to generating reliable radar echo data.
• The IMD facilitates standardized evaluation of deep learning algorithms.