skip to main content
DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Performance evaluation of two optical architectures for task-specific compressive classification

Abstract

Many optical systems are used for specific tasks such as classification. Of these systems, the majority are designed to maximize image quality for human observers. However, machine learning classification algorithms do not require the same data representation used by humans. We investigate the compressive optical systems optimized for a specific machine sensing task. Two compressive optical architectures are examined: an array of prisms and neutral density filters where each prism and neutral density filter pair realizes one datum from an optimized compressive sensing matrix, and another architecture using conventional optics to image the aperture onto the detector, a prism array to divide the aperture, and a pixelated attenuation mask in the intermediate image plane. We discuss the design, simulation, and trade-offs of these systems built for compressed classification of the Modified National Institute of Standards and Technology dataset. Both architectures achieve classification accuracies within 3% of the optimized sensing matrix for compression ranging from 98.85% to 99.87%. The performance of the systems with 98.85% compression were between an F/2 and F/4 imaging system in the presence of noise.

Authors:
 [1];  [2];  [2];  [2];  [2];  [2]
  1. Sandia National Lab. (SNL-NM), Albuquerque, NM (United States); Univ. of Arizona, Tucson, AZ (United States)
  2. Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Publication Date:
Research Org.:
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Sponsoring Org.:
USDOE National Nuclear Security Administration (NNSA)
OSTI Identifier:
1595428
Report Number(s):
SAND-2020-0102J
Journal ID: ISSN 0091-3286; 681873
Grant/Contract Number:  
AC04-94AL85000; NA-0003525
Resource Type:
Accepted Manuscript
Journal Name:
Optical Engineering
Additional Journal Information:
Journal Volume: 59; Journal Issue: 05; Journal ID: ISSN 0091-3286
Publisher:
SPIE
Country of Publication:
United States
Language:
English
Subject:
47 OTHER INSTRUMENTATION; compressive sensing; compressive classification; task-specific; MNIST

Citation Formats

Redman, Brian J., Dagel, Amber L., Galiardi, Meghan A., LaCasse, Charles F., Quach, Tu-Thach, and Birch, Gabriel C. Performance evaluation of two optical architectures for task-specific compressive classification. United States: N. p., 2020. Web. doi:10.1117/1.OE.59.5.051404.
Redman, Brian J., Dagel, Amber L., Galiardi, Meghan A., LaCasse, Charles F., Quach, Tu-Thach, & Birch, Gabriel C. Performance evaluation of two optical architectures for task-specific compressive classification. United States. doi:10.1117/1.OE.59.5.051404.
Redman, Brian J., Dagel, Amber L., Galiardi, Meghan A., LaCasse, Charles F., Quach, Tu-Thach, and Birch, Gabriel C. Tue . "Performance evaluation of two optical architectures for task-specific compressive classification". United States. doi:10.1117/1.OE.59.5.051404. https://www.osti.gov/servlets/purl/1595428.
@article{osti_1595428,
title = {Performance evaluation of two optical architectures for task-specific compressive classification},
author = {Redman, Brian J. and Dagel, Amber L. and Galiardi, Meghan A. and LaCasse, Charles F. and Quach, Tu-Thach and Birch, Gabriel C.},
abstractNote = {Many optical systems are used for specific tasks such as classification. Of these systems, the majority are designed to maximize image quality for human observers. However, machine learning classification algorithms do not require the same data representation used by humans. We investigate the compressive optical systems optimized for a specific machine sensing task. Two compressive optical architectures are examined: an array of prisms and neutral density filters where each prism and neutral density filter pair realizes one datum from an optimized compressive sensing matrix, and another architecture using conventional optics to image the aperture onto the detector, a prism array to divide the aperture, and a pixelated attenuation mask in the intermediate image plane. We discuss the design, simulation, and trade-offs of these systems built for compressed classification of the Modified National Institute of Standards and Technology dataset. Both architectures achieve classification accuracies within 3% of the optimized sensing matrix for compression ranging from 98.85% to 99.87%. The performance of the systems with 98.85% compression were between an F/2 and F/4 imaging system in the presence of noise.},
doi = {10.1117/1.OE.59.5.051404},
journal = {Optical Engineering},
number = 05,
volume = 59,
place = {United States},
year = {2020},
month = {1}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record

Save / Share:

Works referenced in this record:

Convolutional Neural Networks for Large-Scale Remote-Sensing Image Classification
journal, February 2017

  • Maggiori, Emmanuel; Tarabalka, Yuliya; Charpiat, Guillaume
  • IEEE Transactions on Geoscience and Remote Sensing, Vol. 55, Issue 2
  • DOI: 10.1109/TGRS.2016.2612821

Compressive imaging system design using task-specific information
journal, January 2008

  • Ashok, Amit; Baheti, Pawan K.; Neifeld, Mark A.
  • Applied Optics, Vol. 47, Issue 25
  • DOI: 10.1364/AO.47.004457

Design and evaluation of task-specific compressive optical systems
conference, May 2019

  • Redman, Brian J.; Birch, Gabriel C.; LaCasse, Charles F.
  • Computational Imaging IV
  • DOI: 10.1117/12.2520187

Task-specific information for imaging system analysis
journal, January 2007

  • Neifeld, Mark A.; Ashok, Amit; Baheti, Pawan K.
  • Journal of the Optical Society of America A, Vol. 24, Issue 12
  • DOI: 10.1364/JOSAA.24.000B25

On Projection Matrix Optimization for Compressive Sensing Systems
journal, June 2013

  • Li, Gang; Zhu, Zhihui; Yang, Dehui
  • IEEE Transactions on Signal Processing, Vol. 61, Issue 11
  • DOI: 10.1109/TSP.2013.2253776

Direct inference on compressive measurements using convolutional neural networks
conference, September 2016

  • Lohit, Suhas; Kulkarni, Kuldeep; Turaga, Pavan
  • 2016 IEEE International Conference on Image Processing (ICIP)
  • DOI: 10.1109/ICIP.2016.7532691

An Ensemble of Fine-Tuned Convolutional Neural Networks for Medical Image Classification
journal, January 2017

  • Kumar, Ashnil; Kim, Jinman; Lyndon, David
  • IEEE Journal of Biomedical and Health Informatics, Vol. 21, Issue 1
  • DOI: 10.1109/JBHI.2016.2635663

Optical systems for task-specific compressive classification
conference, September 2018

  • Birch, Gabriel C.; Quach, Tu-Thach; Galiardi, Meghan
  • Optics and Photonics for Information Processing XII
  • DOI: 10.1117/12.2321331

Distributed Compressed Estimation Based on Compressive Sensing
journal, September 2015

  • Xu, Songcen; de Lamare, Rodrigo C.; Poor, H. Vincent
  • IEEE Signal Processing Letters, Vol. 22, Issue 9
  • DOI: 10.1109/LSP.2015.2400372

Optical architectures for compressive imaging
journal, January 2007


Single-pixel imaging via compressive sampling
journal, March 2008

  • Duarte, Marco F.; Davenport, Mark A.; Takhar, Dharmpal
  • IEEE Signal Processing Magazine, Vol. 25, Issue 2, p. 83-91
  • DOI: 10.1109/MSP.2007.914730

On the mean accuracy of statistical pattern recognizers
journal, January 1968


Compressive Sensing [Lecture Notes]
journal, August 2007


Learning to Sense Sparse Signals: Simultaneous Sensing Matrix and Sparsifying Dictionary Optimization
journal, July 2009

  • Duarte-Carvajalino, J. M.; Sapiro, G.
  • IEEE Transactions on Image Processing, Vol. 18, Issue 7
  • DOI: 10.1109/TIP.2009.2022459

Unsupervised Deep Feature Extraction for Remote Sensing Image Classification
journal, March 2016

  • Romero, Adriana; Gatta, Carlo; Camps-Valls, Gustau
  • IEEE Transactions on Geoscience and Remote Sensing, Vol. 54, Issue 3
  • DOI: 10.1109/TGRS.2015.2478379

Random Forests
journal, January 2001


Optimized Projection Matrix for Compressive Sensing
journal, July 2010

  • Xu, Jianping; Pi, Yiming; Cao, Zongjie
  • EURASIP Journal on Advances in Signal Processing, Vol. 2010, Issue 1
  • DOI: 10.1155/2010/560349

The smashed filter for compressive classification and target recognition
conference, February 2007

  • Davenport, Mark A.; Duarte, Marco F.; Wakin, Michael B.
  • Electronic Imaging 2007, SPIE Proceedings
  • DOI: 10.1117/12.714460

Multi-column deep neural network for traffic sign classification
journal, August 2012


Optimized Projections for Compressed Sensing
journal, December 2007