skip to main content

This content will become publicly available on October 11, 2022

Title: What You Can Learn by Staring at a Blank Wall
We present a passive non-line-of-sight method that infers the number of people or activity of a person from the observation of a blank wall in an unknown room. Our technique analyzes complex imperceptible changes in indirect illumination in a video of the wall to reveal a signal that is correlated with motion in the hidden part of a scene. We use this signal to classify between zero, one, or two moving people, or the activity of a person in the hidden scene. We train two convolutional neural networks using data collected from 20 different scenes, and achieve an accuracy of 94% for both tasks in unseen test environments and real-time online settings. Unlike other passive non-line-of-sight methods, the technique does not rely on known occluders or controllable light sources, and generalizes to unknown rooms with no recalibration. We analyze the generalization and robustness of our method with both real and synthetic data, and study the effect of the scene parameters on the signal quality.
Authors:
; ; ; ; ; ;
Award ID(s):
1816209
Publication Date:
NSF-PAR ID:
10301901
Journal Name:
Int. Conf. Computer Vision (ICCV-2021)
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Non-Line-Of-Sight (NLOS) imaging aims at recovering the 3D geometry of objects that are hidden from the direct line of sight. One major challenge with this technique is the weak available multibounce signal limiting scene size, capture speed, and reconstruction quality. To overcome this obstacle, we introduce a multipixel time-of-flight non-line-of-sight imaging method combining specifically designed Single Photon Avalanche Diode (SPAD) array detectors with a fast reconstruction algorithm that captures and reconstructs live low-latency videos of non-line-of-sight scenes with natural non-retroreflective objects. We develop a model of the signal-to-noise-ratio of non-line-of-sight imaging and use it to devise a method thatmore »reconstructs the scene such that signal-to-noise-ratio, motion blur, angular resolution, and depth resolution are all independent of scene depth suggesting that reconstruction of very large scenes may be possible.

    « less
  2. In this paper, we present a multiple concurrent occupant identification approach through footstep-induced floor vibration sensing. Identification of human occupants is useful in a variety of indoor smart structure scenarios, with applications in building security, space allocation, and healthcare. Existing approaches leverage sensing modalities such as vision, acoustic, RF, and wearables, but are limited due to deployment constraints such as line-of-sight requirements, sensitivity to noise, dense sensor deployment, and requiring each walker to wear/carry a device. To overcome these restrictions, we use footstep-induced structural vibration sensing. Footstep-induced signals contain information about the occupants' unique gait characteristics, and propagate through themore »structural medium, which enables sparse and passive identification of indoor occupants. The primary research challenge is that multiple-person footstep-induced vibration responses are a mixture of structurally-codependent overlapping individual responses with unknown timing, spectral content, and mixing ratios. As such, it is difficult to determine which part of the signal corresponds to each occupant. We overcome this challenge through a recursive sparse representation approach based on cosine distance that identifies each occupant in a footstep event in the order that their signals are generated, reconstructs their portion of the signal, and removes it from the mixed response. By leveraging sparse representation, our approach can simultaneously identify and separate mixed/overlapping responses, and the use of the cosine distance error function reduces the influence of structural codependency on the multiple walkers' signals. In this way, we isolate and identify each of the multiple occupants' footstep responses. We evaluate our approach by conducting real-world walking experiments with three concurrent walkers and achieve an average F1 score for identifying all persons of 0.89 (1.3x baseline improvement), and with a 10-person "hybrid" dataset (simulated combination of single-walker real-world data), we identify 2, 3, and 4 concurrent walkers with a trace-level accuracy of 100%, 93%, and 73%, respectively, and observe as much as a 2.9x error reduction over a naive baseline approach.« less
  3. Non-line-of-sight (NLOS) imaging is a rapidly advancing technology that provides asymmetric vision: seeing without being seen. Though limited in accuracy, resolution, and depth recovery compared to active methods, the capabilities of passive methods are especially surprising because they typically use only a single, inexpensive digital camera. One of the largest challenges in passive NLOS imaging is ambient background light, which limits the dynamic range of the measurement while carrying no useful information about the hidden part of the scene. In this work we propose a new reconstruction approach that uses an optimized linear transformation to balance the rejection of uninformativemore »light with the retention of informative light, resulting in fast (video-rate) reconstructions of hidden scenes from photographs of a blank wall under high ambient light conditions.« less
  4. Monitoring the compliance of social distancing is critical for schools and offices to recover in-person operations in indoor spaces from the COVID-19 pandemic. Existing systems focus on vision- and wearable-based sensing approaches, which require direct line-of-sight or device-carrying and may also raise privacy concerns. To overcome these limitations, we introduce a new monitoring system for social distancing compliance based on footstep-induced floor vibration sensing. This system is device-free, non-intrusive, and perceived as more privacy-friendly. Our system leverages the insight that footsteps closer to the sensors generate vibration signals with larger amplitudes. The system first estimates the location of each personmore »relative to the sensors based on signal energy and then infers the distance between two people. We evaluated the system through a real-world experiment with 8 people, and the system achieves an average accuracy of 97.8% for walking scenario classification and 80.4% in social distancing violation detection.« less
  5. Single-photon avalanche diodes (SPADs) are a rapidly developing image sensing technology with extreme low-light sensitivity and picosecond timing resolution. These unique capabilities have enabled SPADs to be used in applications like LiDAR, non-line-of-sight imaging and fluorescence microscopy that require imaging in photon-starved scenarios. In this work we harness these capabilities for dealing with motion blur in a passive imaging setting in low illumination conditions. Our key insight is that the data captured by a SPAD array camera can be represented as a 3D spatio-temporal tensor of photon detection events which can be integrated along arbitrary spatio-temporal trajectories with dynamically varyingmore »integration windows, depending on scene motion. We propose an algorithm that estimates pixel motion from photon timestamp data and dynamically adapts the integration windows to minimize motion blur. Our simulation results show the applicability of this algorithm to a variety of motion profiles including translation, rotation and local object motion. We also demonstrate the real-world feasibility of our method on data captured using a 32x32 SPAD camera.« less