Sensor Fusion with Asynchronous Decentralized Processing for 3D Target Tracking with a Wireless Camera Network
Abstract
:1. Introduction
1.1. Related Work
1.2. Our Proposal
1.3. Contributions
1.4. Paper Structure
2. Problem Formulation
- -
- Variable k instants or time steps within a time window: they are integer multiples of the camera inter-frame interval and represent the capture instants of target centroids, and the instants of LoS vectors are received from neighboring cameras during the time window about the frame capture instant of a local camera. They are stored in the i-th camera node;
- -
- Frame capture instants : the moment when a frame is captured and defines a time window;
- -
- Here we assume that the communication is ideal among cameras, i.e., there is no communication delays.
- First: We propose to exchange among neighboring cameras LoS vector measurements to the t-th target from the i-th camera at instant within the time window period, and the data are represented in the world coordinate system {W}. The measured (referring to the target centroid in the image plane in pixel coordinates) is transformed into with the camera construction parameters as described in Section 2.1. The fused 3D position measurement results from averaging intersecting pairs of vector constraints derived from LoS vectors of neighboring j-th cameras pointing at the same t-th target according to a LoS-vector 3D proximity determination algorithm. Local target state estimation uses the fused 3D position measurement to update and predict the state estimate in the next frame capture instant .
- Second: For comparison, exchange among cameras and , where , the 3D fused measurement for each target, transmitted and instantaneously received at instants within the fusion window with configurable p size after the time window. This is the used in a sequential approach in SA-PF.
- Third: Again, for comparison, exchange among cameras and , where , information pairs the estimates for each target, transmitted and instantaneously received at instants within the fusion window with configurable p size after the time window. This is used in the batch approach in BAF.
2.1. Obtaining
2.1.1. Target Centroid Detection in the Image Plane in Pixels and Identification Number (Id) Assignment
2.1.2. Camera Calibration
2.1.3. The Optical Center and Point Represented in the World Coordinate System {W}: A Preamble for Obtaining the LoS Vector
2.1.4. LoS Vector and Parametric Equation of the Line
2.1.5. Proximity Metric between Vector Constraints Originating from LoS to a Target
2.1.6. Asynchronous Exchange of Information Vectors
2.1.7. Obtaining the 3D Position Measurement at Capture Instant
2.2. Dynamic Target Model: Linear Gaussian Model
2.3. Target Information Fusion within the Fusion Window [15]
- (1)
- BAF fusion: The t-th target information pair is propagated from frame capture instant to , thus yielding () for transmission. Define to be the reception instant at which camera receives an information pair estimate () in the fusion window. We use the Euclidean distance applied to the corresponding 3D centroid position to indicate whether the target is the same. Then, all the received information pairs’ estimates are propagated backward in time to for Kullback–Leibler Average (KLA) fusion with the local i-th camera information pair estimate (), and the result propagated to the next frame capture instant .
- (2)
- SA-PF fusion: As stated before, sends the fused 3D measurement and covariance at instant to the neighboring cameras without time propagation from to . At various instants , receives (,) from neighboring cameras, only once from each . Here as well we use Euclidean distance applied to 3D position measurement to indicate whether the target is the same. Fusion occurs in this filter by propagating particles’ state estimates from the capture instant to the distinct receiving instants and sequentially updates the respective weights with (,). At the end of the fusion window, updates the state estimate at and propagates to the next frame capture instant .
3. Asynchronous Bayesian State Estimation
Algorithm1. Algorithm to investigate the effect on target tracking of further fusion other than exchanged LoS vectors and local estimation |
Input: I–images taken by camera Output: estimated state and the root mean square error computation |
for r Monte-Carlo simulation runs for each camera 1-If t-th target detection is confirmed after three occurrences in images, its position estimate is computed as follows: for each detected target 2-Detect target centroids in image plane within the time window period ; 3-Transform measurements into stored information vectors ; 4-Asynchronous transmission/reception of information vectors among cameras and , ; 5-Compute 3D position measurements and covariances within the time window from received LoS vectors, and interpolate all of them into temporal alignment with the frame capture instant to account for frame processing delay and camera asynchrony; 6-Local estimation of the target position and velocity in 3D space ; 7-Investigate the added fusion phase with either the BAF or the SA-PF Asynchronously transmit/receive information pairs’s estimates (BAF) or fused 3D position measurements (SA-PF) among cameras and , ; 8- is the resulting 3D position component of the target state estimate; 9-Perspective projection of the 3D position estimate of each target being tracked onto the 2D image planes of each neighboring camera, expressed in pixel coordinates: ; end for; end for; 10-Average root mean square error to compare performance in steps 6 and 7. end for. |
3.1. Batch Asynchronous Filter (BAF)-Lower Load Mode (LLM) [15]
3.2. Sequential Asynchronous Particle Filter [17]
4. Simulation Results
4.1. Evaluation Setup
4.2. Discussions
- (a)
- Variable —each target has 7 sigma points assigned to the local LoS vector pointing to the t-th target in the image plane;
- (b)
- Variable —each target with its 7 sigma points assigned to the received LoS vector pointing to the received t-th target in the image plane;
- (c)
- Intersection search of pairs of LoS from the set of 49 LoS vectors providing points in 3D space (Equations (6) and (8));
- (d)
- Checks which 3D points pass the distance criterion m;
- (e)
- Selects the 7 LoS vectors from the set of 49 LoS vectors that pass the above test with smallest distance ; and
- (f)
- UT computation: 3D position measurement mean and covariance obtained at a capture instant of the i-th camera.
- (a)
- For the number of targets varying from 1 up to 90 targets (i.e., from 1 × 1 up to 90 × 90 targets), the elapsed processing time shows an exponential tendency;
- (b)
- From 91 up to 262 targets (i.e., from 91 × 91 up to 262 × 262 targets) we observe a linear tendency, and over 262 × 262 targets, the elapsed processing time presents an exponential tendency.
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Abbreviations
WCN | Wireless Communication Networks |
BAF | Batch Asynchronous Filter |
BAF-LLM | Batch Asynchronous Filter – Lower Load Mode |
SA-PF | Sequential Asynchronous Particle Filter |
LoS | Line Of Sight |
FoV | Field Of View |
Appendix A
References
- Sutton, Z.; Willet, P.; Bar-Shalom, Y. Target Tracking Applied to Extraction of Multiple Evolving Threats from a Stream of Surveillance Data. IEEE Transac. Comput. Soc. Syst. 2021, 8, 434–450. [Google Scholar] [CrossRef]
- Song, B.; Kamal, A.T.; Soto, C.; Ding, C.; Farrel, J.A.; Roy-Chowdhury, A.K. Tracking and Activity Recognition Through Consensus in Distributed Camera Networks. IEEE Transac. Imag. Proc. 2010, 19, 2564–2579. [Google Scholar] [CrossRef] [PubMed]
- Olfati-Saber, R. Distributed Kalman filtering for sensor networks. In Proceedings of the 2007 46th IEEE Conference on Decision and Control, New Orleans, LA, USA, 12–14 December 2007; pp. 5492–5498. [Google Scholar] [CrossRef]
- Solomon, C.; Breckon, T. Fundamentos de Processamento de Imagens uma Abordagem práTica com Exemplos em Matlab; LTC: Rio de Janeiro, Brazil, 2013. (In Portuguese) [Google Scholar]
- Chagas, R.; Waldmann, J. Difusão de medidas para estimação distribuída em uma rede de sensores. In Proceedings of the XI Symposium on Electronic Warfare, São José dos Campos, Brazil, 29 September–2 October 2009; pp. 71–75. (In Portuguese). [Google Scholar]
- Beaudeau, J.; Bugallo, M.F.; Djurić, P.M. Target tracking with asynchronous measurements by a network of distributed mobile agents. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Kyoto, Japan, 25–30 March 2012; pp. 3857–3860. [Google Scholar] [CrossRef]
- Casbeer, D.W.; Beard, R. Distributed information filtering using consensus filters. In Proceedings of the American Control Conference, St. Luis, MO, USA, 10–12 June 2009; pp. 1882–1887. [Google Scholar] [CrossRef]
- Garcia-Fernandez, A.; Grajal, J. Asynchronous Particle Filter for tracking using non-synchronous sensor networks. Sig. Proc. 2011, 91, 2304–2313. [Google Scholar] [CrossRef]
- Wymeersch, H.; Zazo, V.S. Belief consensus algorithms for fast distributed target tracking in wireless sensor networks. Sig. Process. 2014, 95, 149–160. [Google Scholar] [CrossRef] [Green Version]
- Ji, H.; Lewis, F.L.; Hou, Z.; Mikulski, D. Distributed information-weigthed Kalman consensus filter sensor networks. Automatica 2017, 77, 18–30. [Google Scholar] [CrossRef] [Green Version]
- Khaleghi, B.; Khamis, A.; Karray, F.O.; Razavi, N.S. Multisensor data fusion: A review of the state-of-the-art. Inf. Fusion 2013, 14, 28–44. [Google Scholar] [CrossRef]
- Taj, M.; Cavallaro, A. Distributed and Decentralized Multicamera Tracking. IEEE Sign. Process. Mag. 2011, 28, 46–58. [Google Scholar] [CrossRef]
- Hlinka, O.; Hlawatsch, F.; Djurić, P.M. Distributed Sequential Estimation in Asynchronous Wireless Sensor Networks. IEEE Sig. Proc. Lett. 2015, 22, 1965–1969. [Google Scholar] [CrossRef]
- Olfati-Saber, R. Kalman-Consensus Filter: Optimality, stability, and performance. In Proceedings of the 48h IEEE Conference on Decision and Control (CDC) Held Jointly with 2009 28th Chinese Control Conference, Shanghai, China, 15–18 December 2009; pp. 7036–7042. [Google Scholar] [CrossRef]
- Katragadda, S.; Cavallaro, A. A batch asynchronous tracker for wireless smart-camera networks. In Proceedings of the 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), Lecce, Italy, 29 August–1 September 2017; pp. 1–6. [Google Scholar] [CrossRef]
- Yang, R.; Bar-Shalom, Y. Full State Information Transfer Across Adjacent Cameras in a Network Using Gauss Helmert Filters. J. Advan. Inform. Fus. 2022, 17, 14–28. [Google Scholar]
- Zhu, G.; Zhou, F.; Xie, L.; Jiang, R.; Chen, Y. Sequential Asynchronous Filters for Target tracking in wireless sensor networks. IEEE Sens. J. 2014, 14, 3174–3182. [Google Scholar]
- Olfati-Saber, R.; Sandell, N.F. Distributed tracking in sensor networks with limited sensing range. In Proceedings of the 2008 American Control Conference, Seattle, WA, USA, 11–13 June 2008; pp. 3157–3162. [Google Scholar] [CrossRef]
- Julier, S.; Uhlmann, J.; Durrant-Whyte, H.F. A new method for the nonlinear transformation of means and covariances in filters and estimators. IEEE Trans. Autom. Cont. 2000, 45, 477–482. [Google Scholar] [CrossRef] [Green Version]
- Battistelli, G.; Chisci, L. Kullback-Leibler average, consensus on probability densities, and distributed state estimation with guaranteed stability. Automática 2014, 50, 707–718. [Google Scholar] [CrossRef]
- Katragadda, S.; Regazzoni, C.S.; Cavallaro, A. Average consensus-based asynchronous tracking. In Proceedings of the 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, USA, 5–9 March 2017; pp. 4401–4405. [Google Scholar] [CrossRef]
- Kushwaha, M.; Koutsoukos, X. Collaborative 3D Target Tracking in Distributed Smart Camera Networks for Wide-Area Surveillance. J. Sens. Actuator Netw. 2013, 2, 316–353. [Google Scholar] [CrossRef] [Green Version]
- Matteo, Z.; Matteo, L.; Marangoni, T.; Mariolino, C.; Marco, T.; Manuela, G. 3D Tracking of Human Motion Using Visual Skeletonization and Stereoscopic Vision. Front. Bioeng. Biotech. 2020, 8. [Google Scholar] [CrossRef]
- Zivkovic, Z. Wireless smart camera network for real-time human 3D pose reconstruction. Comp. Vis. Im. Underst. 2010, 114, 1215–1222. [Google Scholar] [CrossRef]
- Giordano, J.; Lazzaretto, M.; Michieletto, G.; Cenedese, A. Visual Sensor Networks for Indoor Real-Time Surveillance and Tracking of Multiple Targets. Sensors 2022, 22, 2661. [Google Scholar] [CrossRef] [PubMed]
- Yan, J.; Liu, H.; Pu, W.; Bao, Z. Decentralized 3-D Target Tracking in Asynchronous 2-D Radar Network: Algorithm and Performance Evaluation. IEEE Sens. J. 2017, 17, 823–833. [Google Scholar] [CrossRef]
- PETS 2009 Benchmark Data. Available online: http://cs.binghamton.edu/~mrldata/pets2009 (accessed on 21 August 2022).
- Rottmann, M.; Maag, K.; Chan, R.; Hüger, F.; Schlicht, P.; Gottschalk, H. Detection of False Positive and False Negative Samples in Semantic Segmentation. In Proceedings of the 2020 Design, Automation & Test in Europe Conference & Exhibition (DATE), Grenoble, France, 9–13 March 2020; pp. 1351–1356. [Google Scholar] [CrossRef]
- Fang, K.; Xiang, Y.; Li, X.; Savarese, S. Recurrent Autoregressive Networks for Online Multi-Object Tracking. In Proceedings of the IEEE Winter Conference on Applications of Computer Vision, Lake Tahoe, NV, USA, 12–15 March 2018; pp. 466–475. [Google Scholar] [CrossRef]
- Cavallaro, A.; Aghajan, H. Multi-Camera Networks: Principles and Applications; Academic Press: Cambridge, MA, USA, 2009. [Google Scholar]
- Corke, P. Robotics, Vision, and Control; International Publishing AG: Cham, Switzerland, 2017. [Google Scholar]
- Hartley, R.; Zisserman, A. Multiple View Geometry in Computer Vision; Cambridge University Press: Cambridge, UK, 2003. [Google Scholar]
- Lao, D.; Sundaramoorthi, G. Minimum Delay Object Detection from Video. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Republic of Korea, 27 October–2 November 2019; pp. 5096–5105. [Google Scholar] [CrossRef] [Green Version]
- Matsuyama, T.; Ukita, N. Real-time multitarget tracking by a cooperative distributed vision system. Proc. IEEE 2002, 90, 1136–1150. [Google Scholar] [CrossRef]
- Konstantinova, P.; Udvarev, A.; Semerdjiev, T. A study of a Target Tracking Algorithm Using Global Nearest Neighbor Approach. In Proceeding of the 4th International Conference on Computer Systems and Technology-CompSysTech, Rousse, Bulgaria, 19–20 June 2003; pp. 290–295. [Google Scholar]
- Munkres, J. Algorithms for the assignment and transportation problems. J. Soc. Indus. Appl. Mathem. 1957, 5, 32–38. [Google Scholar] [CrossRef] [Green Version]
- Tsai, R.Y. A Versatile Camera Calibration Techniques for High-Accuracy 3D Machine Vision Metrology Using Off-the-Shelf TV Cameras and Lenses. IEEE J. Robot. Autom. 1987, 3, 323–344. [Google Scholar] [CrossRef] [Green Version]
- Camargo, I.; Boulos, P. Geometria Analítica um Tratamento Vetorial, 3rd ed.; Pearson: São Paulo, Brazil, 2005. (In Portuguese) [Google Scholar]
- Ferreira, P.P.C. Cálculo e Análise Vetoriais com Aplicações Práticas; Moderna: Rio de Janeiro, Brazil, 2012; Volume 1. (In Portuguese) [Google Scholar]
- Wan, E.A.; Van Der Merwe, R. The unscented Kalman filter for nonlinear estimation. In Proceedings of the IEEE 2000 Adaptive Systems for Signal Processing, Communications, and Control Symposium (Cat. No.00EX373), Lake Luise, AB, Canada, 4 October 2000; pp. 153–158. [Google Scholar] [CrossRef]
- Bar-Shalom, Y.; Li, X.R.; Kirubarajan, T. Estimation with Applications to Tracking and Navigation: Theory, Algorithms and Software; Wiley Online Library: Hoboken, NJ, USA, 2002. [Google Scholar]
- Katragadda, S. Available online: http://www.eecs.qmul.ac.uk/~andrea/software.htm (accessed on 10 March 2022).
- Thrun, S.; Burgard, W.; Fox, D. Probabilistic Robotics; MIT Press: Massachusetts, UK, 2005. [Google Scholar]
Ref. | Acronym | Characteristics | Fusion Method | |||
---|---|---|---|---|---|---|
Processing Delays | Aligns Received Data to Local Capture Instant | Data Exchanged among Camera Nodes | Position Representation | |||
[16] | GHF | no | no | State estimates | Polar coordinates | No fusion |
[25] | Decentralized EKF | no | no | State estimates | Cartesian coordinates | Sequential |
[26] | DAT2TF | no | no | State estimates | Polar coordinates | Sequential |
[15] | BAF-LLM | yes | yes | State estimates | Cartesian coordinates | Batch |
[17] | SA-PF | no | no | Measurements | Cartesian coordinates | Sequential |
This paper [15] | BAF-LLM—local estimation 1 | yes | yes 3 | (1) LoS vectors to measure 3D positions | Cartesian coordinates | Batch |
BAF-LLM—with further fusion | yes | yes | (2) LoS vectors to measure 3D positions and information pairs’ estimates | Cartesian coordinates | Batch | |
This paper [17] | PF—local estimation 2 | no | yes 3 | (1) LoS vectors to measure 3D positions | Cartesian coordinates | Batch |
SA-PF—with further fusion | no | no | (2) LoS vectors to measure 3D positions and fused 3D position measurements | Cartesian coordinates | Sequential |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Di Gennaro, T.M.; Waldmann, J. Sensor Fusion with Asynchronous Decentralized Processing for 3D Target Tracking with a Wireless Camera Network. Sensors 2023, 23, 1194. https://doi.org/10.3390/s23031194
Di Gennaro TM, Waldmann J. Sensor Fusion with Asynchronous Decentralized Processing for 3D Target Tracking with a Wireless Camera Network. Sensors. 2023; 23(3):1194. https://doi.org/10.3390/s23031194
Chicago/Turabian StyleDi Gennaro, Thiago Marchi, and Jacques Waldmann. 2023. "Sensor Fusion with Asynchronous Decentralized Processing for 3D Target Tracking with a Wireless Camera Network" Sensors 23, no. 3: 1194. https://doi.org/10.3390/s23031194
APA StyleDi Gennaro, T. M., & Waldmann, J. (2023). Sensor Fusion with Asynchronous Decentralized Processing for 3D Target Tracking with a Wireless Camera Network. Sensors, 23(3), 1194. https://doi.org/10.3390/s23031194