Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group
Journal Home About Issues in Progress Current Issue All Issues Early Posting Feature Issues

3D far-field Lidar sensing and computational modeling for human identification

A. Glandon, L. Vidyaratne, N. K. Dhar, B. O. Familoni, N. Sadeghzadehyazdi, S. T. Acton, and K. M. Iftekharuddin

Open Access Open Access

Abstract

3D sensors offer depth sensing that may be used for task-specific data processing and computational modeling. Many existing methods for human identification using 3D depth sensors primarily focus on Kinect data, where the range is very limited. This work considers a 3D long-range Lidar sensor for far-field imaging of human subjects in 3D Lidar full motion video (FMV) of “walking” action. 3D Lidar FMV data for human subjects are used to develop computational modeling for automated human silhouette and skeleton extraction followed by subject identification. We propose a matrix completion algorithm to handle missing data in 3D FMV due to self-occlusion and occlusion from other subjects for 3D skeleton extraction. We further study the effect of noise in the 3D low resolution far-field Lidar data in human silhouette extraction performance of the model. Moreover, this work addresses challenges associated with far-field 3D Lidar including learning with a limited amount of data and low resolution. Moreover, we evaluate the proposed computational algorithm using a gallery of 10 subjects for human identification and show that our method is competitive with the state-of-the-art OpenPose and V2VPose skeleton extraction models using the same dataset for human identification.

© 2023 Optica Publishing Group

Full Article  |  PDF Article

Corrections

2 January 2024: A correction was made to the author listing.

More Like This
LiDAR-camera-system-based unsupervised and weakly supervised 3D object detection

Haosen Wang, Tiankai Chen, Xiaohang Ji, Feng Qian, Yue Ma, and Shifeng Wang
J. Opt. Soc. Am. A 40(10) 1849-1860 (2023)

3D object detection through fog and occlusion: passive integral imaging vs active (LiDAR) sensing

Kashif Usmani, Timothy O’Connor, Pranav Wani, and Bahram Javidi
Opt. Express 31(1) 479-491 (2023)

Enhancing 3D human pose estimation with NIR single-pixel imaging and time-of-flight technology: a deep learning approach

Carlos Osorio Quero, Daniel Durini, Jose Rangel-Magdaleno, Jose Martinez-Carranza, and Ruben Ramos-Garcia
J. Opt. Soc. Am. A 41(3) 414-423 (2024)

View More...

Data availability

Data underlying the results presented in this paper are not publicly available as the data are considered sensitive by the US Army NVESD.

Cited By

Optica participates in Crossref's Cited-By Linking service. Citing articles from Optica Publishing Group journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (8)
Fig. 1.
Fig. 1. Proposed flow diagram of computational model for silhouette and skeleton extraction.

Download Full Size | PDF

Fig. 2.
Fig. 2. Single person silhouette extraction (a) Raw data. (b) Naïve extraction. (c) Noise removed. (d) All limbs.

Download Full Size | PDF

Fig. 3.
Fig. 3. Multiple person silhouette extraction (a) Raw data. (b) Two persons silhouette extraction.

Download Full Size | PDF

Fig. 4.
Fig. 4. Silhouette size to demonstrate model operating limits.

Download Full Size | PDF

Fig. 5.
Fig. 5. Examples of challenging sensor operation regions. (a) Subject too close to sensor. (b) Subject too far from sensor. (c) Subject middle range with noise.

Download Full Size | PDF

Fig. 6.
Fig. 6. Example Lidar input video frame with 3D skeleton output. (a) Single subject skeleton output. (b) Multiple subject skeleton output.

Download Full Size | PDF

Fig. 7.
Fig. 7. Comparison of occlusion completion methods.

Download Full Size | PDF

Fig. 8.
Fig. 8. ROC curves for single and multiple subject identification.

Download Full Size | PDF

Tables (2)

Tables Icon

Table 1. Best Accuracy for Single Subject per Frame

View Table | View all tables in this article
Tables Icon

Table 2. Performance Comparison with Existing Methods

View Table | View all tables in this article

Equations (12)

Equations on this page are rendered with MathJax. Learn more.

prop 1 ( x , y , t ) = n e a r f i e l d < r a n g e ( x , y , t ) < max t r a n g e ( x , y , : ) − ϵ b a c k g r o u n d .
prop 2 ( x , y , t ) = m e d i a n p r o p 1 − α c l o s e < r a n g e ( x , y , t ) < m e d i a n p r o p 1 + α f a r
prop 3 ( x , y , t ) = ( m e d i a n p r o p 1 − α c l o s e < r a n g e ( x , y , t ) < m e d i a n p r o p 1 + α r e l a x e d ) ; A N D i n t e n s i t y ( x , y , t ) > i n t e n s i t y n o i s e .
prop 4 ( x , y , t ) = p r o p 1 A N D ( p r o p 2 O R p r o p 3 ) .
prop 5 ( x , y , t , k ) = p r o p 1 ( x , y , t ) A N D r a n g e ( x , y , t ) ∈ s e t k .
prop 6 ( x , y , t , k ) = p r o p 5 ( x , y , t , k ) A N D x > x o f f s e t ;
prop 7 ( x , y , t , k ) = u n i o n f i n d ( p r o p 6 ( x , y , t , k ) ∀ t , k ) ;
prop 8 ( x , y , t , k ) = p r o p 7 ( x , y , t , k ) O R × ( x > x o f f s e t A N D p r o p 5 ( x , y , t , k ) ) .
y ¯ ( t ) = ∑ y ( t ) | s i l h o u e t t e ( x , y , t ) n ,
z ¯ ( t ) = ∑ z ( x , y , t ) | s i l h o u e t t e ( x , y , t ) n .
P = [ X 11 Y 11 Z 11 X 12 Y 12 Z 12 X 1 N Y 1 N Z 1 N X 21 Y 21 Z 21 X 22 Y 22 Z 22 X 2 N Y 2 N Z 2 N ⋮ ⋱ X M 1 Y M 2 Z M 3 X M 2 Y M 2 Z M 2 X M N Y M N Z M N ]
X h e i g h t = X p i x e l s ⋅ Z .
Select as filters
Select Topics Cancel
Top
       
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.
Privacy | Terms of Use