Dynamic Antenna Array Design for Scene Classification Through Fourier-Domain Filtering
| Autor: | Stavros Vakalis, Jeffrey A. Nanzer, Daniel Chen |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Contextual image classification
Computer science business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION 020206 networking & telecommunications 02 engineering and technology Filter (signal processing) Iterative reconstruction Antenna array Microwave imaging Frequency domain Radar imaging 0202 electrical engineering electronic engineering information engineering Computer vision Spatial frequency Artificial intelligence Electrical and Electronic Engineering business |
| Zdroj: | IEEE Transactions on Antennas and Propagation. 69:5953-5962 |
| ISSN: | 1558-2221 0018-926X |
| DOI: | 10.1109/tap.2021.3069521 |
| Popis: | We present a new approach to the classification of scenes in the microwave and millimeter-wave bands that leverages a novel dynamic antenna array concept to capture distinct features in the spatial frequency information of the scene. The spatial frequency information of a scene is obtained through its Fourier transform, and by sampling a subset of this information, key features can be extracted and used for image classification. We demonstrate that a dynamic antenna array can synthesize spatial frequency filters, and that scene classification can be accomplished using the filtered signals without full image reconstruction. We develop a new dynamic antenna array concept using only two antennas to generate a ring-shaped spatial frequency filter and explore the use of this concept for the classification of ground scenes. Natural ground scenes tend to have smoother spatial frequency signals, while, in contrast, features such as buildings and roadways result in sharp broadband spatial frequency responses. Using this design, we demonstrate the ability to classify between two classes of ground scenes: those with man-made structures (buildings, roads, etc.) and those without (natural scenes). We demonstrate the ability of the spatial filters synthesized by the proposed dynamic antenna array to achieve a classification accuracy of 0.971 with an empirical true positive rate of 0.982. The method is broadly applicable to microwave and millimeter-wave sensing at any range. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|