Ground Moving Target Detection for Airborne Radar Using Machine Learning Approaches

Authors

Rafi Ahmed, Florida International UniversityFollow

Document Type

Dissertation

Degree

Doctor of Philosophy (PhD)

Major/Program

Electrical and Computer Engineering

First Advisor's Name

Dr. Hai Deng

First Advisor's Committee Title

committee chair

Second Advisor's Name

Dr. Ou Bai

Second Advisor's Committee Title

committee member

Third Advisor's Name

Dr. Jean H. Andrian

Third Advisor's Committee Title

committee member

Fourth Advisor's Name

Dr. B. M. Golam Kibria

Fourth Advisor's Committee Title

committee member

Keywords

Airborne radar, Radar clutter, Radar signal processing, Moving target indication, Feature extraction, Machine learning

Date of Defense

7-1-2022

Abstract

Airborne radar faces many challenges to suppress unknown interferences from ground reflections to detect slow-moving targets. In this dissertation work, a feature-based machine learning approach is proposed to effectively classify target and interference such as ground clutter without actually removing them using traditional methods. Multiple features are considered for developing the target/clutter classification algorithms of airborne radars with digital arrays. The features we use for classification include the clutter proximity measures and target geometric feature.

The proximity feature is extracted to distinguish target, and clutter in location in the Doppler-angle domain for airborne radar. The Euclidean distance between a signal and the locus of the expected clutter ridge is known as clutter proximity feature. The distance feature value is generated for each non-zero signal pixel in the angle-Doppler domain of the radar data. Ground moving target and clutter signals are classified and recognized based on the feature for target detection without removing clutters in traditional filtering methods. The proposed feature method is especially effective for target detection in inhomogeneous clutter environment.

In some radar operational scenarios, a single feature might not be enough, and we further introduce the geometric features as well as proximity feature to the machine-learning target detection method to improve the target detection performance. Several geometric features such as block size, roundness ratio, and bending energy are used to extract the relevant geometric information indicating target and clutter geometric differences. The extracted features are then utilized to classify target and clutters reliably and robustly.

The effectiveness of the proposed feature-based methods is validated by the simulation results based on typical airborne radar systems. This study also demonstrates substantial performance improvement over traditional target detection methods such as space-time adaptive processing (STAP) and Beam-Doppler Image Feature Recognition (BDIFR) methods.

Identifier

FIDC010828

Recommended Citation

Ahmed, Rafi, "Ground Moving Target Detection for Airborne Radar Using Machine Learning Approaches" (2022). FIU Electronic Theses and Dissertations. 5101.
https://digitalcommons.fiu.edu/etd/5101