کنفرانس مهندسی برق ایران

صفحه اصلی / سی و یکمین کنفرانس بین المللی مهندسی برق

SAR Images Clustering Based on Modified Nonlinear Orthogonal non-Negative Matrix Factorization (NMF)

نویسندگان :

Mahdi Jowkar dehouei¹ Soolmaz Khazandi² Yaser Norouzi³

1- دانشگاه صنعتی امیرکبیر(پلی تکنیک تهران) 2- دانشگاه صنعتی امیرکبیر(پلی تکنیک تهران) 3- دانشگاه صنعتی امیرکبیر(پلی تکنیک تهران)

کلمات کلیدی :

SAR،Clustering،NMF،Orthogonal،Spectral

چکیده :

This study presents the results of unsupervised learning (clustering) on Synthetic Aperture Radar (SAR) images. We can perform Automatic Target Recognition (ATR) in SAR images thanks to fast and accurate clustering. Data can be automatically clustered based on similar properties. It was common to present Nonnegative Matrix Factorization (NMF) linearly. We employed a nonlinear NMF with orthogonal properties in this presentation. The outcomes of spectral clustering for subspaces and NMF matrix factorization are equivalent, as demonstrated by mathematical and empirical experiments. The kernel-based orthogonal in subspace clustering results in this method with updates were superior to those in other methods. We benchmarked our methodology using the dataset for stationary target acquisition and recognition.