Investigation on the scattering characteristics and unsupervised clustering of 3D printed samples. Issue 9 (24th November 2020)

• Record Type: Journal Article
• Authors: Elsaadouny, Mostafa; Barowski, Jan; Jebramcik, Jochen; Rolfes, Ilona
• Editors: Macháč, J.; Švanda, M.; Vbra, J.
• Is Part Of: International journal of microwave and wireless technologies. Volume 12:Issue 9(2020)
• Page Start: 862
• Page End: 869
• Publication Date: 2020-11-24
• Subjects: Radar, -- SAR imaging, -- image processing, -- 3D printing, -- machine learning, -- unsupervised clustering; Wireless communication systems -- Periodicals; Microwave circuits -- Periodicals; Radio frequency -- Periodicals; 621.381305
• DOI: 10.1017/S1759078720000823
• Languages: English
• ISSNs: 1759-0787

Abstract

Abstract: In this work, the scattering characteristics of 3D-printed samples are being investigated by using a single-polarized and a cross-polarized radar system. The 3D-printed technology participates in a wide range of applications nowadays. The idea of synthetic aperture radar (SAR) has been utilized to investigate the reflected electromagnetic energy from the 3D-printed samples by setting each of the radar systems in a fixed position and the mounting sample on an x - y positioning table which has been used to achieve rectangular-scan mode for SAR. The data have been ported and processed by the matched filter approach. For better image interpretation, the data have been further processed by the median filter in order to reduce noise level while preserving the main image details. Afterwards, the data have been further investigated for determining and classifying any possible defects. This process has been accomplished by deploying the unsupervised learning concept to cluster the SAR responses into two groups, compromising the defected positions responses and the non-defected responses. The obtained results of both radar sensors have been compared and evaluated using different quality assessment factors. Moreover, unsupervised learning techniques have been investigated and the obtained results show a high degree of efficiency in clustering the SAR responses.