3D object recognition using ISAR image
Katsura Kawakami, H. Tanaka, Kazuhiko Yamamoto · Society of Instrument and Control Engineers of Japan · 2004
We propose a 3D object recognition method using inverse synthetic aperture radar images, which doesn't need any aspect information. It employs a simple support vector machine of linear kernel. Numerical experiments clarified that when the movements of the targets are ideally or practically compensated, it recognizes cones with up to 90% accuracy and cylinders with up to 100% accuracy.