Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Scattering-induced entropy boost for highly-compressed optical sensing and encryption

Xinrui Zhan, Xuyang Chang, Daoyu Li, Rong Yan, Yinuo Zhang, Liheng Bian

TL;DR

A novel image-free sensing framework for resource-efficient image classification, where the required number of measurements can be reduced by up to two orders of magnitude, and represents a significant breakthrough in high-throughput machine intelligence for scene analysis with low bandwidth, low costs, and strong encryption.

Abstract

Image sensing often relies on a high-quality machine vision system with a large field of view and high resolution. It requires fine imaging optics, has high computational costs, and requires a large communication bandwidth between image sensors and computing units. In this paper, we propose a novel image-free sensing framework for resource-efficient image classification, where the required number of measurements can be reduced by up to two orders of magnitude. In the proposed framework for single-pixel detection, the optical field for a target is first scattered by an optical diffuser and then two-dimensionally modulated by a spatial light modulator. The optical diffuser simultaneously serves as a compressor and an encryptor for the target information, effectively narrowing the field of view and improving the system's security. The one-dimensional sequence of intensity values, which is measured with time-varying patterns on the spatial light modulator, is then used to extract semantic information based on end-to-end deep learning. The proposed sensing framework is shown to obtain over a 95\% accuracy at sampling rates of 1% and 5% for classification on the MNIST dataset and the recognition of Chinese license plates, respectively, and the framework is up to 24% more efficient than the approach without an optical diffuser. The proposed framework represents a significant breakthrough in high-throughput machine intelligence for scene analysis with low bandwidth, low costs, and strong encryption.

Scattering-induced entropy boost for highly-compressed optical sensing and encryption

TL;DR

A novel image-free sensing framework for resource-efficient image classification, where the required number of measurements can be reduced by up to two orders of magnitude, and represents a significant breakthrough in high-throughput machine intelligence for scene analysis with low bandwidth, low costs, and strong encryption.

Abstract

Image sensing often relies on a high-quality machine vision system with a large field of view and high resolution. It requires fine imaging optics, has high computational costs, and requires a large communication bandwidth between image sensors and computing units. In this paper, we propose a novel image-free sensing framework for resource-efficient image classification, where the required number of measurements can be reduced by up to two orders of magnitude. In the proposed framework for single-pixel detection, the optical field for a target is first scattered by an optical diffuser and then two-dimensionally modulated by a spatial light modulator. The optical diffuser simultaneously serves as a compressor and an encryptor for the target information, effectively narrowing the field of view and improving the system's security. The one-dimensional sequence of intensity values, which is measured with time-varying patterns on the spatial light modulator, is then used to extract semantic information based on end-to-end deep learning. The proposed sensing framework is shown to obtain over a 95\% accuracy at sampling rates of 1% and 5% for classification on the MNIST dataset and the recognition of Chinese license plates, respectively, and the framework is up to 24% more efficient than the approach without an optical diffuser. The proposed framework represents a significant breakthrough in high-throughput machine intelligence for scene analysis with low bandwidth, low costs, and strong encryption.
Paper Structure (12 sections, 2 equations, 4 figures, 1 table)

This paper contains 12 sections, 2 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. The reported scattering-induced image-free sensing and optical encryption.
  4. Scattering procedure
  5. Structural light illumination
  6. Sensing network
  7. Experimental setup
  8. Results
  9. Conclusion and discussion
  10. Author Contributions
  11. Funding
  12. Acknowledgment

Figures (4)

  • Figure 1: The framework of the reported method. (a) A scenario illustration in car plate license sensing. (b) Optical transformations in the workflow. (c) Entropy boost induced by scattering. We use a light-like image to symbolize the transformation of information. The target's different areas of information are mixed into a section after scattering, which produces a significant entropy boost. Therefore, we propose to only extract the regional field of view (FOV) of the scattering image for optical sensing and encryption. The left figure presents the quantitative analysis of the scattering-induced entropy boosts by calculating the average entropy of each image dataset. (d) Analysis of different scattering strengths. The 'entropy' on the x-axis represents the average entropy of the training images with different scattering intensities, and the 'accuracy' on the y-axis is the sensing accuracy on the test dataset at the same sampling rate with a regional FOV. For single-pixel sensing, we select only a 50% field width at a 0.05 sampling rate. For multitarget recognition, the corresponding parameters are set to 8% and 0.1.
  • Figure 2: Single-target image-free sensing results. (a) Sensing accuracy at different sampling rates, including computational simulations on the MNIST and Fashion-MNIST datasets (with 'Monte-0.8' scattering intensity) and physical experiments on the MNIST dataset. (b) Sensing accuracy in the regional FOV. In this work, the sampling rate is the ratio of the coupled measurement length to the size of the modulated field area. The term 'field width' refers to the width of the modulated target area (the number of active columns or rows during modulation), which can be narrowed with a scattering-induced entropy boost. (c) Different strategies to extract the regional FOV data. The black blocks in the modulation patterns have values of zero where the light cannot pass through, and the white blocks have values of one. The gray line shows the adjustment that turns a one into a zero. (d) Sensing accuracy with different simulated scattering strengths on the MNIST dataset.
  • Figure 3: Multitarget sensing results. (a) Simulation results on the SCCPD dataset with different scattering strengths at different sampling rates. (b) Experimental results on the SCCPD dataset at different sampling rates and different field widths.
  • Figure 4: Modulation sensing results for multitarget recognition. (a) Examples of random patterns and optimal patterns. (b) Sensing results with learning-based optimal patterns, where 'LBO' represents the learning-based optimal and the 'width scalability' is the scale ratio of the selected regional FOV to the original light width. (c) Sensing results when modulated patterns are replaced by the diffuser at different numerical precisions.