Shandong Wu, Ph.D.
Postdoctoral Researcher

University of Pennsylvania, School of Medicine
Department of Radiology
3600 Market St. Suite 360
Philadelphia PA 19104-2643
Tel: 215-615-0825
Fax: 215-349-8972
E-mail: Shandong.Wu at uphs dot upenn dot edu, roboted at gmail dot com
Webpage: http://www.eecs.ucf.edu/~sdwu/


What's New

CV

PDF (6-page)



Bio-sketch


Dr. Wu is currently a Postdoctoral Researcher in the Computational Breast Imaging Group at the University of Pennsylvania School of Medicine (mentor: Dr. Kontos Despina). From Apr. 2009 to Apr. 2011, he was a Postdoctoral Associate in the Computer Vision Lab in the Department of Electrical Engineering and Computer Science at the University of Central Florida (mentor: Dr. Mubarak Shah). Dr. Wu received his Ph.D. degree in Robot Vision from City University of Hong Kong, Hong Kong in 2008 under supervision of Dr. Y. F. Li. He received his B.Sc. and M.Sc. both in Computer Science from Shanghai University, Shanghai, China in 1998 and 2003, respectively. During 1998 to 2004, Dr. Wu was employed jointly, at Shanghai University and Shanghai Robotics Institute, as a Research Assistant, Teaching Assistant, and Lecturer.

Dr. Wu has published more than 30 papers in referred journals and conferences mainly in the fields of Computer Vision, Pattern Recognition, and Robotics, in which more than 90% are the first-author publications.  He has been involved (PI, leading, or participating) in more than 15 research projects granted by the agencies of the US, Hong Kong, and China. Dr. Wu served as the PC member for a number of international conference such as IMECS, WCICA, ICAL, ICMA, ICNC, etc. He is also a regular reviewer for many journals (e.g. PAMI, TMI, TIP, CVIU, ACM MM, SMC-B, RAM, NEUROCOMUTING, AR, MVA, JEI, etc.) and conferences (ICCV, CVPR, ECCV, ICRA, IROS, CASE, HRI, ROBIO, ICIP, etc.). Dr. Wu also has experience in teaching courses on Computer Science.

Dr. Wu is a member of IEEE and RSNA.

Research Interests/Experience

My research is in a broad area in Computer Vision, Biomedical Image Analysis, and robotics. My meta-interest lies in cutting edge computational vision techniques and interdisciplinary research with the goal of leveraging the technical strengths in computational science to boost translational biomedical applications. Currently I am dedicated to the following two main research themes:
  1. Biomedial image analysis: I have growing interests in translating advanced computer vision and imaging processing techniques to meeting practical demands in clinical and biological applications. This enables me to go beyond a pure computational scientist to discovery important and challenging biomedical imaging problems and then propose solutions. I enjoy working very closely with clinicians and biologists. I led a NIH-sponsored project on human brain tumor detection in MR images during my postdoc training at UCF where I investigated a low-rank optimization algorithm for batch processing of brain tumor segmentation and contributed an interactive volumetric tumor segmentation system developed by using ITK and VTK, which is found potentially and practically useful as a computer-aided diagnosis tool for radiologists and oncologists. Biologically, I was applying 3D volumetric reconstruction to examine XGAL+ cells in adult mouse heart. I was also involved in stem cell analysis for drug compound screening through glial cell segmentation and neuron tracking in phase-contrast microscopy images/videos. Currently I focus on breast MR imaging analysis to investigate quantitative risk biomarkers for personalized breast cancer risk assessment in high-risk women screening. Related work includes breast segmentation, fibroglandular tissue segmentation, enhancement segmentation, and the characterization of parenchymal and enhancement patterns.

  2. Visual motion analysis: This entails classic computer vision tasks such as action/behavior/activity/crowd analysis (segmentation, description, perception, recognition, learning, anomaly detection, etc.). My favorite motion feature is "motion trajectory" and I have specific interest in designing trajectory-oriented descriptors. I have been actively investigating clustered particle trajectories for crowd scene analysis and anomaly detection. I am also extending the work for human action recognition with moving cameras using the decomposed motion components from particle trajectories. Previously I pioneered the investigation of a systematic signature-based motion trajectory descriptor in my Ph.D. study. I also correlate the motion analysis with robot learning in context of visual learning by demonstration, where I pursue to enable a robot to actively learn generalized human demonstrated tasks through leveraging the signature descriptor of the demonstrations.
Research Specifics:

Computer Vision : human/robot action/behavior/activity recognition and learning, video surveillance, motion trajectory descriptor, crowded scene analysis, shape representation, chaotic invariants, time series analysis, anomaly detection, motion prediction, visual search;  

Medical/Biological Image Analysis : breast MRI analysis, breast segmentation, fibroglandular tissue segmentation, breast tissue likelihood atlas, DCE-MRI biomarker for risk assessment; brain tumor detection from MRI, subspace decomposition for tumor segmentation, ITK+VTK, tissue symmetry detection; volumetric cell reconstruction; stem cell segmentation and tracking in phase-contrast microscopy images;

Robotics: robot task descriptor, robot learning by demonstration, distributed robot control, computing grid-based multi-robot collaboration, mobile agent, hand-held device-based robot control

　

  Selected Research Projects

     

Videos: Demos for ARG, APHill, UCF sports, and HOHA dataset

　 Recognition of human actions in a video acquired by a moving camera typically requires standard preprocessing steps such as motion compensation, moving object detection and object tracking. The errors from themotion compensation step propagate to the object detection stage,resulting in miss-detections, which further complicates thetracking stage, resulting in cluttered and incorrect tracks.Therefore, action recognition from a moving camera is consideredvery challenging. In this paper, we propose a novel approach whichdoes not follow the standard steps, and accordingly avoids theaforementioned difficulties. Our approach is based on Lagrangianparticle trajectories which are a set of dense trajectoriesobtained by advecting optical flow over time, thus capturing theensemble motions of a scene. This is done in frames of unalignedvideo, and no object detection is required. In order to handle themoving camera, we propose a novel approach based on low rankoptimization, where we decompose the trajectories into theircamera-induced and object-induced components. We performed intensive experiments on multiple benchmark datasets andtwo new aerial datasets called ARG and APHill, and obtainedpromising results.

Publications: Shandong Wu, Omar Oreifej, Mubarak Shah, Action Recognition in Videos Acquired by a Moving Camera Using Motion Decomposition of Lagrangian Particle Trajectories, International Conference on Computer Vision (ICCV 2011) , Barcelona, Spain, Nov. 2011. (pdf)

     

Videos: Particle trajs -> Trimmed trajs -> Represenative trajs

　 A novel method for crowd flow modeling and anomaly detection is proposed for both coherent and incoherent scenes. The novelty is revealed in three aspects. First, it is a unique utilization of particle trajectories for modeling crowded scenes, in which we propose new and efficient representative trajectories for modeling arbitrarily complicated crowd flows. Second, chaotic dynamics are introduced into the crowd context to characterize complicated crowd motions by regulating a set of chaotic invariant features, which are reliably computed and used for detecting anomalies. Third, a probabilistic framework for anomaly detection and localization is formulated.

Videos: (1) Particle advection (particle trajectory)   (2) After trimming of noise and motionless particles   (3) Representative trajectory   (4) 10-frame clip based particle trajectory for a 590-frame sequence   Code (coming soon);   Poster (ppt)    Project page

Publications: Shandong Wu, Brian E. Moore, and Mubarak Shah, “Chaotic Invariants of Lagrangian Particle Trajectories for Anomaly Detection in Crowded Scenes,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010), San Francisco, CA, USA, June 13-18, 2010. (Acceptance rate: 22.3%) (pdf); 

　 　 　

More details

Picture

 　

We are investing the Computer Assisted Identification and Volumetric Quantification of Enhancements for Brain Tumor from MRI Scans in a NIH-sponsored project. I have been designing and developing an interactive software system by using ITK, VTK, and MFC for clinical trails for both radiologists and neuro-oncologists. The current implementation aims at using both T1 pre-contrast and post-contrast modalities to identify the enhancements. However, the system provides an option to use T1 post-contrast individually or other single modality such as T2, etc. The system is featured by the seamless combination of robust computer vision techniques (e.g. registration and segmentation) and the interactive incorporation of clinical knowledge from the users. Another important feature is the automatic and intelligent propagation of the processes operated on a single slice to the rest slices of a scan. Both quantitative and qualitative (3D visualization) tumor assessment can be generated and archived.  The system shows great advantages in accuracy and reproducibility in volumetric tumor assessment as compared to present bi-directional measurement.

 

System Developer: Shandong Wu (sdwu@eecs.ucf.edu), Ph.D., Nicholas Avgeropoulos, M.D., David Rippe, M.D., and Mubarak Shah, Ph.D.

 

Publications: (1)AANS/CNS 2009; (2) In preparation (journal)
　 　 　

PLoSONE

 　 Three-dimensional computer imaging of adrenergic cell-derived myocardium in the adult mouse heart reveals predominantly left-sided distribution. We examined the three-dimensional (3-D) distribution of XGAL+ cells using computer-aided reconstruction of our 2-D photo microphotographs.  This 3-D representation confirmed our 2-D results showing that adrenergic derived cells populate a vast majority of the left heart.  Our 3-D representation is the first look at the structural architecture and localization of adrenergic derived cells within the adult mouse myocardium.  This work is a collaboration with Dr. Steven Ebert at UCF biomedical school.

Publications: (1) AHA2009; (2) AHA2010;  (3) Journal: PLoS ONE 2011
　 　 　

　 We propose a novel motion trajectory signature descriptor and study its rich descriptive invariants to benefit effective motion trajectory recognition. Invariant is a key measure to the flexibility and effectiveness of a descriptor. Substantial descriptive invariants can be deduced from the proposed trajectory signature. We first present the signature definition and its robust implementation. Then the signature’s invariants are elaborated. A nonlinear inter-signature matching algorithm is developed to measure the signature’s similarity for trajectory recognition. Experiments are conducted to recognize human sign language, in which both synthetic and real data are used to verify the signature’s invariants.

Publications: Shandong Wu and Y. F. Li, “On Signature Invariants for Effective Motion Trajectory Recognition,” The International Journal of Robotics Research, vol. 27, no. 8, pp. 895-917, Aug. 2008. (pdf)
　 　 　

    

　 We propose a novel 3-D motion trajectory signature descriptor and develop three signature descriptions for motion characterization. The flexible descriptions give the signature high functional adaptability to meet various application requirements in trajectory representation, perception and recognition. The full signature, optimized signature and cluster signature are firstly defined for trajectory representation. Then we explore the motion perception from a single signature, inter-signature matching and the generalization of a cluster signature. Furthermore, three solutions for signature recognition are investigated corresponding to different signature descriptions. The conducted experiments verified the signature’s capabilities and flexibility. The signature’s application to robot learning is also discussed.

Videos: (1) DTW based inter-trajectory matching   (2) Cluster signature   (3) Full signature instantiation from cluster signature

Publications:  Shandong Wu and Y. F. Li, “Flexible Signature Descriptions for Adaptive Motion Trajectory Representation, Perception and Recognition,” Pattern Recognition, vol. 42, no. 1, pp. 194-214, Jan. 2009. (pdf)   (Also related: ICRA2008, IROS2009)
　 　 　

　 For an effective descriptor sometimes it is desired to have mutual description functions regarding the describing and un-describing abilities to support some applications like robot learning. Hence, opposite to describing a motion trajectory using the signature, this paper focuses on the un-describing problem, that is, reproducing a trajectory instance from a given signature description. The moving frame technique is used in formulating the trajectory reproduction method. A nonlinear signature matching-based metric is also developed to measure the quality of reproductions. This work can serve as a core component for trajectory based robot learning by visual demonstration.

Videos: (1) Frenet Frame for trajectory reproduction

Publications: Shandong Wu and Y. F. Li, “Motion Trajectory Reproduction from Generalized Signature Description,” Pattern Recognition, vol. 43, no. 1, pp. 204-221, Jan. 2010. (pdf)   (Also related: IROS2008)
　 　 　

　 The development of intelligent hand-held devices provides a new solution for remote robot control. Using the intelligent hand-held devices, a user can perform controlling robots from anywhere at anytime. In this paper, a general architecture for remote robot control using hand-held devices is proposed. The technical solutions and implementation details are discussed according to current technologies. The PT500 industry robot is used in the experiment and it is shown that the intelligent hand-held devices extremely extend the controllable scope in both time and space domains.

Publications: Shandong Wu and Yimin Chen, “Remote Robot Control Using Intelligent Hand-held Devices,” in Proc. IEEE International Conference on Computer and Information Technology (CIT 2004), Wuhan, China, Sep. 2004, pp. 587-592. (pdf)
　 　 　

　 The new generation Internet information infrastructure--computing grid, presents a new solution for robot control. In this paper, we present our research on computing grid-based robot system which aims at organizing and integrating robot’s control capabilities to construct a Robot-Control-Grid (RCG). Not only a single robot can be connected to the RCG, but also multiple robots located in different places can collaborate in the RCG to perform distributed robot tasks. A simple portal for accessing and controlling a robot is designed for end users.

Publications: Shandong Wu and Y.F. Li, “Grid-based Robot Control,” in Proc. International Conference on High Performance Computing and Applications (HPCA 2004), Shanghai, China, Aug. 2004, pp. 501-508. (extended version pdf)
　 　 　

　 Distributed computing technologies are gaining increasing interests in constructing robot control systems. This paper proposes a loosely distributed software model by coupling Agent and CORBA for supporting multi-robot collaboration. Within this architecture the Agent control units reside at semantic level while CORBA serves as providing functional interfaces to Agent at the syntax level. In addition, this study also investigates a robot oriented Mobile Intelligent Agent Model (MIAM) that is equipped with modules of core function, migration ability, intelligent engine and communication interfaces.

Publications: (1)High Technology Letters (journal, pdf); (2) ICIIP 2004 (pdf); (3) Computer Engineering and Applications, 2004 (journal, in Chinese, pdf)
　 　 　
Selected Publications (30+ publications)

Refereed journal papers

  1. Kingsley Osuala, Kathleen Telusma, Saad M. Khan, Shandong Wu, Mubarak Shah, Candice Baker, Sabikha Alam, Ibrahim Abukenda, Aura Fuentes, Hani B. Seifein, and Steven N. Ebert, "Distinctive left-sided distribution of adrenergic-derived myocytes in the adult mouse heart," PLoS ONE 2011.
  2. Shandong Wu and Y.F. Li, “Motion Trajectory Reproduction from Generalized Signature Description,” Pattern Recognition, vol. 43, no. 1, pp. 204-221, Jan. 2010.
  3. Shandong Wu and Y.F. Li, “Flexible Signature Descriptions for Adaptive Motion Trajectory Representation, Perception and Recognition,” Pattern Recognition, vol. 42, no. 1, pp. 194-214, Jan. 2009.
  4. Shandong Wu and Y.F. Li, “On Signature Invariants for Effective Motion Trajectory Recognition,” The International Journal of Robotics Research, vol. 27, no. 8, pp. 895-917, Aug. 2008.
  5. Shandong Wu, Yimin Chen, and Yongyi He, “A Loosely Coupled Control Architecture Based on Agent and CORBA for Multiple Robots,” High Technology Letters, Vol. 9, No. 4, pp. 17-20, Dec. 2003.
  6. Shandong Wu, Yimin Chen, and Yongyi He, “A Remote Robot Control System Based on Mobile Agent,” Computer Engineering and Applications, Vol. 40, No. 35, pp. 229-232, Dec. 2004. (in Chinese)

Refereed conference papers

  1. Shandong Wu, Omar Oreifej, Mubarak Shah, "Action Recognition in Videos Acquired by a Moving Camera Using Motion Decomposition of Lagrangian Particle Trajectories," International Conference on Computer Vision (ICCV2011), Barcelona, Spain, 6-13 Nov. 2011.
  2. Shandong Wu, Mubarak Shah, “Chaotic Invariants of Lagrangian Particle Trajectories for Anomaly Detection in Crowded Scenes,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010), San Francisco, CA, USA, June 13-18, 2010.
  3. Shandong Wu, Y.F. Li, and Jianwei Zhang, “Probabilistic Cluster Signature for Modeling Motion Classes,” IEEE International Conference on Intelligent Robots and Systems (IROS 2009), St. Louis, MO, USA, October 11-15, 2009.
  4. Shandong Wu, Y.F. Li, and Jianwei Zhang, “Motion Descriptor: A Motion Trajectory Signature,” IEEE International Conference on Information and Automation (ICIA 2009), Zhuhai/Macau, China, June 22 - 25, 2009, pp. 346-351.
  5. Shandong Wu, Y.F. Li, and Jianwei Zhang, “Signature Based Task Description and Perception for Motion Trajectory Oriented Robot Learning,” IEEE International Conference on Robotics and Biomimetics (ROBIO 2008), Bangkok, Thailand, Dec. 14 -17, 2008, pp. 1123-1128.
  6. Shandong Wu, Y.F. Li, and Jianwei Zhang, “Invariant Signature Description and Trajectory Reproduction for Robot Learning by Demonstration,” in Proc. IEEE International Conference on Intelligent Robots and Systems (IROS 2008), Nice, France, Sep. 22-26, 2008, pp. 4060-4065.
  7. Shandong Wu, Y.F. Li, and Jianwei Zhang, “A Hierarchical Motion Trajectory Signature Descriptor,” in Proc. IEEE International Conference on Robotics and Automation (ICRA 2008), Pasadena, CA, USA, May 2008, pp. 3070-3075.
  8. Shandong Wu, Y.F. Li, and Jianwei Zhang, “A Viewpoint Invariant Signature Descriptor for Curved Shape Recognition,” in Proc. IEEE International Conference on Robotics and Biomimetics (ROBIO 2007), Sanya, P. R. China, Dec. 15 -18, 2007, pp. 1121-1126.
  9. Shandong Wu and Y.F. Li, “On Signature Invariants for Motion Trajectory Recognition,” in Proc. International Conference on Advanced Robotics (ICAR 2007), Jeju Island, Korea, Aug. 2007, pp. 532-537.
  10. Shandong Wu, Y.F. Li, and Jianwei Zhang, “Signature Descriptor for Free Form Trajectory Modeling,” in Proc. IEEE International Conference on Integration Techniques (ICIT 2007), Shenzhen, P. R. China, Mar. 2007, pp. 167-172.
  11. Shandong Wu and Y.F. Li, “Computing Grid-based Robot Control,” in Proc. International Conference on High Performance Computing and Applications (HPCA 2004), Shanghai, P. R. China, Aug. 2004, pp. 501-508.
  12. Shandong Wu and Yimin Chen, “Remote Robot Control Using Intelligent Hand-held Devices,” in Proc. IEEE International Conference on Computer and Information Technology (CIT 2004), Wuhan, P. R. China, Sep. 2004, pp. 587-592.
  13. Shandong Wu, Yimin Chen, and Yongyi He, “MIAM: A Robot Oriented Mobile Intelligent Agent Model,” in Proc. International Conference on Intelligent Information Processing (ICIIP 2004), Zhongzhi Shi and Qing He, eds., N.Y.: Springer, Beijing, P. R. China, Oct. 2004, pp. 51-54.

Posters

  1. Eric J. Leach, Pingkun Yan, David J. Rippe, Nicholas Avgeropoulos G, Shandong Wu, and Mubarak Shah, “A Knowledge-Based Framework for Quantifying Enhancing Brain Tissue,” 2009 Joint Meeting of the Society for Neuro-Oncology and the AANS/CNS Section on Tumors, New Orleans, Louisiana, USA, 22-24 Oct. 2009.
  2. Kingsley Osuala, Kathleen Telusma, S Kahn, Shandong Wu, Mubarak Shah, and Steven Ebert, “ICA Cells Predominantly Contribute to Left Myocardial Development in the Adult Mouse Heart as Revealed by 3-D Computer Imaging,” American Heart Association (AHA) Scientific Sessions 2009, Orlando, Florida, USA, Nov. 14-17, 2009.
  3. K Osuala, K Telusma, S Kahn, S Wu, M Shah, C Baker, K. Pfeifer and SN Ebert, “Non-uniform distribution of adrenergic-derived cells in left ventricular myocardium:  Implications for Tako-Tsubo Syndrome”, American Heart Association (AHA) Scientific Sessions 2010, Chicago, USA, Nov. 13–17, 2010.

Theses

  1. Shandong Wu, “Signature Descriptions for Motion Trajectory Representation, Perception, Recognition and Reproduction,” Ph.D. Thesis, Department of Manufacturing Engineering and Engineering Management, City University of Hong Kong, Hong Kong, 2008.
  2. Shandong Wu, “Distributed Software Architecture on PC Platform for Collaborative Multiple Robots Control,” M.Sc. Thesis, School of Computer Engineering and Science, Shanghai University, 2003. (in Chinese)
  3. Shandong Wu, “Industrial Robot Control and Task Teaching Based on PC and Windows,” B.Sc. Thesis, School of Computer Engineering and Science, Shanghai University, 1998. (in Chinese)

  Personal

   Hobbies:

   Swimming, fishing, badminton, table tennis, tennis, golf, hiking

   Photos:

Eden

My son

BLOG of My Son

　
              
Yummy! I like fishing I believe I can fly... Don't peck me, pigeon ;)
　 　   　
  with Dr.Mubarak Shah
With my beloved wife With my Dad, Mom, and elder Sister With Dr. Y. F. Li With Dr. Mubarak Shah

   My Colleges:

penn　     

     

  This page has been visited several times since 12 June, 2009.
  Last updated: 28 Mar., 2012.