Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Classification of synthesised ISAR images of small complex targets
Thesis (MEng)--Stellenbosch University, 2023.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA en_ZA |
| Published: |
Stellenbosch : Stellenbosch University
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613777759830016 |
|---|---|
| access_status_str | Open Access |
| author | Stewart-Burger, Cullen |
| author2 | Ludick, Danie |
| author_browse | Ludick, Danie Stewart-Burger, Cullen |
| author_facet | Ludick, Danie Stewart-Burger, Cullen |
| author_sort | Stewart-Burger, Cullen |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2023. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/127318 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA en_ZA |
| last_indexed | 2026-06-10T12:41:32.562Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/127318 Classification of synthesised ISAR images of small complex targets Stewart-Burger, Cullen Ludick, Danie Botha, Matthys Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering. Inverse synthetic aperture radar Machine learning Algorithms Thesis (MEng)--Stellenbosch University, 2023. ENGLISH ABSTRACT: This study examines the application and comparison of several machine learning algorithms to the problem of classifying inverse synthetic aperture radar (ISAR) images of electrically small, geometrically complex targets. These algorithms include k-nearest neighbours, logistic regression, a fully connected neural network and a capsule network. A novel classifier is proposed, utilizing a capsule network with a reconstruction sub-network, to perform open-set classification. A dataset of synthetic ISAR images was created from simulated electromagnetic (EM) target returns and used to train and test the models. The EM simulation process was performed using a method of moments solver to compute the backscattering from models of the targets, which are represented as triangular meshes. Specifically considering geometrically small targets with low radar cross-sections, particular attention is paid to the performance of the classifiers when signals are received in the low signal-to-noise ratio regime. The use of a capsule network is found to be highly effective for both closed-set and open-set classification tasks, out-performing the other traditional machine learning and deep learning based classifiers investigated in this study (logistic regression, support vector machines, k-nearest neighbours and fully connected neural networks). The proposed method of comparing the images formed by the capsule network’s reconstruction subnetwork to the input image is demonstrated to be an effective technique for identifying observations of ISAR images of targets that do not belong to any known classes, i.e. targets which are “unknown” to the classifier. Additionally, it is demonstrated that the use of the zero-mean normalised cross-correlation coefficient to compare the input and reconstructed images makes the proposed open-set recognition method more resilient to noisy inputs when compared to the use of the mean-squared error between the images. This addresses a commonly overlooked problem that an operational radar’s automatic target recognition algorithm is not guaranteed to have been trained for all the possible target types that it will sense in the surveillance volume. The proposed classifier achieves an F1-score of greater than 0.9 for a test set containing two known and two unknown classes with signal-to-noise ratios of 6 dB and above. AFRIKAANS OPSOMMING: Hierdie studie ondersoek die toepassing en vergelyking van verskeie masjienleer-algoritmes wat gebruik word om ISAR beelde van elektries klein dog geometries kompleks teikens te herken. Hierdie algoritmes sluit die volgende tegnieke in: logistiese regressie, ondersteuningsvektor-masjiene, k-naaste bure en volledig gekoppelde neurale netwerke. ’n Nuwe klassifiseerder word voorgestel wat gebruik maak van ’n kapsule netwerk met ’n rekonstruksie-subnetwerk om oopstel klassifikasie uit te voer. ‘n Datastel van sintetiese ISAR beelde is geskep wat gebaseer is op die gesimuleerde lectromagnetiese teiken refleksies en is gebruik vir die opleiding en toets can die herkennings algoritmes. Die EM simulasie proses het gebruik gemaak van die “method of moments” oplossings tegniek om die terug gekaatste sein vanaf die teikens te bereken. Die geometrie van die teikens in hierdie simulasie is voorgestel as ‘n stel gekoppelde driehoekies. Daar word spesifiek oorweeging gegee aan elektriese klein teikens met la¨e radardeursnitte, en daar word aandag geskenk aan die prestasie van die klassifiseerders wanneer hulle seine met la¨e sein-tot-ruis verhoudings ontvang as inset. Resulte dui aan dat die gebruik van ’n kapsule netwerk baie doeltreffend is vir beide geslotestel en oopstel klassifikasietake, en beter presteer as die ander tradisionele masjienleer- en diep-leer gebaseerde klassifiseerders wat in hierdie studie ondersoek is (logistiese regressie, ondersteuningsvektor-masjiene, k-naaste bure en volle gekoppelde neurale netwerke). Die voorgestelde tegniek om die beelde te vergelyk wat deur die kapsule netwerk se rekonstruksiesubnetwerk gevorm word met die invoerbeeld, is bewys om doeltreffend te wees om ISAR beelde van teikens te identifiseer wat nie tot enige bekende klasse behoort nie. Dit wil sˆe teikens wat onbekend is vir herkennings algoritmes hiermee identifiseer word. Boonop word daar aangetoont dat die gebruik van die nulgemiddeld-genormaliseerde kruiskorrelasie-ko¨effisi¨ent om die invoer en rekonstrureerde beelde te vergelyk, die voorgestelde oop-stel erkenning meer bestand maak teen ruiserige insette in vergelyking met die gebruik van die gemiddelde-kwadraat-fout tussen die beelde. Hierdeur word ’n algemene uitdaging in outomatiese teiken-herkenning aangespreuk waar ’n operasionele radarstelsel nie gewaarborg is om opgelei te wees op alle tiepes teikens wat in die die sensor se ruimte gewaar kan word nie. Die voorgestelde klassifiseerder behaal ’n F1-telling van meer as 0.9 vir ’n toetsstel wat twee bekende en twee onbekende klasse bevat, met sein-tot-ruis verhoudings van 6 dB en ho¨er. Masters 2023-03-03T12:07:59Z 2023-05-18T07:15:49Z 2023-03-03T12:07:59Z 2023-05-18T07:15:49Z 2023-03 Thesis http://hdl.handle.net/10019.1/127318 en_ZA en_ZA Stellenbosch University xiii, 117 pages : illustrations. application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Inverse synthetic aperture radar Machine learning Algorithms Stewart-Burger, Cullen Classification of synthesised ISAR images of small complex targets |
| title | Classification of synthesised ISAR images of small complex targets |
| title_full | Classification of synthesised ISAR images of small complex targets |
| title_fullStr | Classification of synthesised ISAR images of small complex targets |
| title_full_unstemmed | Classification of synthesised ISAR images of small complex targets |
| title_short | Classification of synthesised ISAR images of small complex targets |
| title_sort | classification of synthesised isar images of small complex targets |
| topic | Inverse synthetic aperture radar Machine learning Algorithms |
| url | http://hdl.handle.net/10019.1/127318 |
| work_keys_str_mv | AT stewartburgercullen classificationofsynthesisedisarimagesofsmallcomplextargets |