Abstract
The need to classify targets and features in high-resolution imagery is of interest in applications such as detection of landmines in ground penetrating radar and tumors in medical ultrasound images. Convolutional neural networks (CNNs) trained using extensive datasets are being investigated recently. However, large CNNs and wavelet scattering networks (WSNs), which share similar properties, have extensive memory requirements and are not readily extendable to other datasets and architectures—and especially in the context of adaptive and online learning. In this paper, we quantitatively study several quantization schemes on WSNs designed for target classification using X-band synthetic aperture radar (SAR) data and investigate their robustness to low signal-to-noise ratio (SNR) levels. A detailed study was conducted on the tradeoffs involved between the various quantization schemes and the means of maximizing classification performance for each case. Thus, the WSN-based quantization studies performed in this investigation provide a good benchmark and important guidance for the design of quantized neural networks architectures for target classification.
Highlights
Feature extraction and classification are essential ingredients in imagery analysis in myriad applications: remote sensing, military, nondestructive testing, ultrasound, medical, cell analysis, etc
Expanding upon our preliminary work in [44], this paper systematically explores the application of quantized wavelet scattering networks (WSNs) to target classification of synthetic aperture radar (SAR) imagery for a large range of signal-to-noise ratio (SNR) conditions; in particular, we used the MSTAR SAR dataset to validate the techniques presented in this paper
Due to its structural similarity with Convolutional neural networks (CNNs), the WSN-based quantization studies performed in this study may provide a good benchmark for future work in the quantization of CNN-based neural networks
Summary
Feature extraction and classification are essential ingredients in imagery analysis in myriad applications: remote sensing, military, nondestructive testing, ultrasound, medical, cell analysis, etc. The foundation of a WSN, the wavelet scattering transform, is itself an effective instrument in feature extraction due to its provision of translation invariance, stability, and the ability to linearize small diffeomorphisms that result from its layered architecture of scattering wavelets It is even used as a preprocessing measure wherein a WSN performs preliminary feature extraction prior to the training of a deep neural network (DNN) for localization [30]. The roto-translation properties of the WSN were incorporated in a convolutional architecture to construct a rotation invariant CNN for image classification [40] Another scale- and rotation-invariant feature extraction method, the speeded-up robust features (SURF) is a local feature detector and descriptor that utilizes multi-scale representation based on box filters [41].
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
