Commercial vehicle classification from spectrum parted linked image test‐attributed synthetic aperture radar imagery. Issue 3 (1st March 2016)

• Record Type: Journal Article
• Authors: Saville, Michael A.; Saini, Dalvir K.; Smith, Jared
• Is Part Of: IET radar, sonar & navigation. Volume 10:Issue 3(2016)
• Page Start: 569
• Page End: 576
• Publication Date: 2016-03-01
• Subjects: synthetic aperture radar -- radar imaging -- image classification -- road vehicle radar -- electromagnetic wave scattering -- radar polarimetry -- AWGN; commercial vehicle classification -- spectrum parted linked image test algorithm -- attributed scattering centre -- wideband polarimetric synthetic aperture radar imagery -- additive white Gaussian noise -- ASC extraction method -- query sets -- SNR approach -- signal‐to‐noise ratio -- attributed synthetic aperture radar imagery; Signal processing -- Periodicals; Radar -- Periodicals; Sonar -- Periodicals; Electronics in navigation -- Periodicals; Navigation -- Periodicals; 621.3848
• DOI: 10.1049/iet-rsn.2015.0342
• Languages: English
• ISSNs: 1751-8784

Abstract

Abstract : The primitive type attribute as recently reported in the spectrum parted linked image test algorithm is used for the first time in a 10‐vehicle classification experiment. Attributed scattering centres (ASCs) from wideband polarimetric synthetic aperture radar imagery are surveyed and two highly complementary sets of attributes are compared in a nearest neighbour classifier. The classifier performance for each set of attributes is shown to be over 90% with 10° sub‐apertures and 10 dB additive white Gaussian noise which had not been considered in earlier works. Using a common image formation and ASC extraction method to ensure that only the pixel attributes differed, two different sets of complementary attributes are compared under precisely the same conditions. In addition, the query sets of attributed images are formed from azimuth and elevation angles that are not in the set of training angles. The results show that the classifier performance degrades gracefully as the signal‐to‐noise ratio (SNR) decreases below 10 dB and that the sensitivity to aspect angle is nearly the same for all vehicle classes as the SNR approaches 10 dB and above. The primary limitation of the approach is the use of wide‐band, wide‐aperture, and polarimetric radar data.