Robinson Czajkowski’s research while affiliated with University of South Florida and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (5)


Two-Edge-Resolved 3d Non-Line-of-Sight Imaging: A Fisher Information Equalized Discretization
  • Conference Paper

April 2024

·

7 Reads

Robinson Czajkowski

·


Two-edge-resolved NLOS imaging scenario and hidden scene representation
a Depiction of the imaging scenario and proposed projected-elevation spherical coordinate. With the origin at the upper-left corner of the door frame, a hidden scene point is identified by its range ρ, azimuth θ, and projected-elevation ψ. b Shows the projected-elevation ψ in the proposed projected-elevation spherical coordinate system, it is the projection of the conventional elevation angle of spherical coordinates onto the xz-plane and is such that tan(ψ)=tan(φ)sec(θ)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tan (\psi )=\tan (\varphi )\sec (\theta )$$\end{document}. (For clarity, the z-axis is flipped from (a) to point upward.) c Elemental surface representation resulting from 10 equal divisions of azimuth and projected-elevation axes with fixed range, ρ. Indicated by the red dot is an example surface element whose centre is at (ρ, θ, ψ) = (1, 11π/40, 13π/40) and angular extents equal π/20 along azimuth and projected-elevation. d, e Depict the changes in the observed measurement due to a hidden point source (red dot) moving from its position in (d) to a new position in (e) such that its range and projected-elevation angle are fixed and only its azimuthal angle changes. The light from a hidden scene point is occluded by the doorway edges to create an illuminated region of trapezoidal shape on the ceiling. The observation in (d) has an illuminated trapezoidal region whose slanted edge is steeper than that of (e) because the azimuthal angle of the point source increases from (d) to (e); the heights of the illuminated trapezoid portions in (d) and (e) are otherwise equal because the projected-elevation angle is unchanged.
Two-step reconstruction procedure
a Measured penumbra photograph. b A linear inverse problem (LIP) is solved, per colour channel, to recover the azimuth and projected-elevation representation (i.e., the shapes) of hidden scene objects, with the entire hidden scene assumed to be confined to a single fixed range. c Colour visualisation of the initial angular reconstruction. d The reconstruction is analysed for connected surface elements which most likely belong to the same cluster. Four such clusters were identified, one for each of the three objects and a fourth (bottom right) for spurious elements. e A non-linear inverse problem (NLIP) is solved to estimate four ranges and four global radiosities, one for each cluster identified in (d). f Multiple views of the 3D colour reconstruction produced by incorporating the estimated ranges and solving a resulting total-variation-regularised LIP to obtain smoother estimates; a 3D view (second column), plan view (first column, top) and side view (first column, bottom) of the final reconstruction. (Best viewed in colour. The reconstruction procedure is also explained in Supplementary Movie 1).
Colour 3D reconstructions of three hidden scenes
a Play scene. b Work scene. c USF scene. For each scene, the first column shows the NLOS measurement photograph (top), and a line-of-sight photograph of the true scene (bottom); the second column shows the 3D full-colour reconstruction; and the last column shows the side (top) and plan (bottom) views of the 3D reconstruction. The reconstructions shown here were obtained with Jmax=4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${J}_{\max }=4$$\end{document} for (a) and (b), and Jmax=5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${J}_{\max }=5$$\end{document} for (c).
Reconstructions under increasing levels of visible side illumination
a Photograph of the hidden scene with two mannequins striking different poses: A red-green mannequin in a T pose and a white mannequin in a marching pose measured at 0.86 m and 1.04 m, respectively, from the origin. b With visible side illumination turned off, the ranges of the mannequins are estimated to be 0.86 m for the T pose mannequin and 1.10 m for the marching mannequin. c Moderate amount of visible side illumination introduced; here, the ranges of the mannequins are to be 0.85 m for the T pose mannequin and 1.10 m for the marching mannequin. d High amount of visible side illumination introduced; the T pose mannequin is reconstructed as two disjoint clusters of elemental surfaces (representing upper and lower halves of the mannequin) with estimated ranges of 0.90 m and 0.92 m, and the marching mannequin is estimated to be at 0.98 m from the origin. The reconstructions shown here were obtained with Jmax=3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${J}_{\max }=3$$\end{document}. Best viewed in colour.
Two-edge-resolved three-dimensional non-line-of-sight imaging with an ordinary camera
  • Article
  • Full-text available

February 2024

·

143 Reads

·

8 Citations

We introduce an approach for three-dimensional full-colour non-line-of-sight imaging with an ordinary camera that relies on a complementary combination of a new measurement acquisition strategy, scene representation model, and tailored reconstruction method. From an ordinary photograph of a matte line-of-sight surface illuminated by the hidden scene, our approach reconstructs a three-dimensional image of the scene hidden behind an occluding structure by exploiting two orthogonal edges of the structure for transverse resolution along azimuth and elevation angles and an information orthogonal scene representation for accurate range resolution. Prior demonstrations beyond two-dimensional reconstructions used expensive, specialized optical systems to gather information about the hidden scene. Here, we achieve accurate three-dimensional imaging using inexpensive, and ubiquitous hardware, without requiring a calibration image. Thus, our system may find use in indoor situations like reconnaissance and search-and-rescue.

Download



Citations (3)


... Many imaging techniques use reflective surfaces to detect objects hidden behind occluding structures [7][8][9][10]. In [10], a three-dimensional image of a scene hidden behind an occluding structure was reconstructed from an ordinary photograph of a matte LOS surface illuminated by the hidden scene. ...

Reference:

The CornerGuard: Seeing around Corners to Prevent Broadside Collisions
Two-edge-resolved three-dimensional non-line-of-sight imaging with an ordinary camera

... Passive NLOS methods, NLOS methods that do not require an active illumination source, primarily model the surface scattering problem from the distribution of scattered light intensity. 23,24 These methods typically employ detectors such as CCD/CMOS to capture scattered light information where the object is illuminated by ambient light or self-emission. 25 Due to the absence of expensive equipment like pulsed lasers or single-pixel detectors, passive NLOS methods offer advantages in terms of easy availability and deployment in practical environments. ...

Turning Door Frames into Cameras for 3D Non-line-of-sight Imaging
  • Citing Conference Paper
  • September 2022

... Passive imaging techniques measure the intensity at the relay surface without the use of any light source. One class of methods utilizes the aperture created by the visible obstructor ( Fig. 1) to extract useful information about the hidden scene [1][2][3][4][5][6]. For example, a vertical edge from this obstructor maps different angular regimes in the hidden scene to different contiguous areas on the relay surface, and this can be used to track [2] and reconstruct [3][4][5] a 2D projection of the hidden object. ...

Double Your Corners, Double Your Fun: The Doorway Camera
  • Citing Conference Paper
  • August 2022