The majority of my research and publication efforts are in coordination with the UCD Graphics Laboratory with Dr. Min Choi. This includes lab research and research conducted in coordination with FLIR Thermal Imaging and Laser Technologies Inc. Some of my previous research in software engineering and robotics was conducted at MSUD with Dr. Jody Paul, Dr. Louise F. Gunderson (G2R), and Dr. James Gunderson (G2R).

My current research directions include deformable object simulation (FEM, Continuum Mechanics), inverse-gas simulation (Navier Stokes), along with sensor development research including thermal and wireless imaging. The following publications illustrate contributions of my research towards computer graphics, mechanical engineering, and medical applications.

International Conferences

Abstract – Reliable occluded skeletal posture estimation is a fundamentally challenging problem for vision-based monitoring techniques. This is due to several imaging related challenges introduced by existing depth-based pose estimation techniques that fail to provide accurate joint position estimates when the line of sight between the imaging device and the patient is obscured by an occluding material. In this work, we present a new method of estimating skeletal posture in occluded applications using both depth and thermal imaging through volumetric modeling and introduce a new occluded ground-truth tracking method inspired by modern motion capture solutions. Using this integrated volumetric model, we utilize Convolutional Neural Networks to characterize and identify volumetric thermal distributions that match trained skeletal posture estimates which includes disconnected skeletal definitions and allows correct posture estimation in highly ambiguous cases. We demonstrate this approach by correctly identifying common sleep postures that present challenging cases for skeletal joint estimates obtaining an average classification accuracy of ~94.45%. [PDF] [IEEE Xplore]

Shane Transue, Phuc Nguyen, Tam Vu, and Min-Hyung. Choi, "Thermal-Depth Fusion for Occluded Body Skeletal Posture Estimation", in The Second IEEE Conference on Connected Health: Applications, Systems, and Engineering, pp. tbd. July, 2017.

Abstract – Breathing volume measurement has long been an important physiological indication widely used for the diagnosis and treatment of pulmonary diseases. However, most of existing breathing volume monitoring techniques require either physical contact with the patient or are prohibitively expensive. In this paper we present an automated and inexpensive non-contact, vision-based method for monitoring an individual’s tidal volume, which is extracted from a three-dimensional (3D) chest surface reconstruction from a single depth camera. In particular, formulating the respiration monitoring process as a 3D spacetime volumetric representation, we introduce a real-time surface reconstruction algorithm to generate omni-direction deformation states of a patient’s chest while breathing, which reflects the change in tidal volume over time. These deformation states are then used to estimate breathing volume through a per-patient correlation metric acquired through a Bayesian-network learning process. Through prototyping and implementation, our results indicate that we have achieved 92.2% to 94.19% accuracy in the tidal volume estimations through the experimentation based on the proposed vision-based method. [PDF] [IEEE Xplore]

Shane Transue, Phuc Nguyen, Tam Vu, and Min-Hyung. Choi, "Real-time Tidal Volume Estimation using Iso-surface Reconstruction", in The First IEEE Conference on Connected Health: Applications, Systems, and Engineering, pp. tbd. June, 2016.

Abstract – We present a methodology of accurately reconstructing the deformation and surface characteristics of a scanned 3D model recorded in real-time within a Finite Element Model (FEM) simulation. In our approach we utilize a template-free procedural process for recording a deforming scanned surface, extracting the physical rest-state model, and reconstruction the deformation surface that de fines the objects deformation characteristics over time. Based on this reference animation, we illustrate the ability to accurately replicate the deformation behaviors of an object composed of an unknown homogeneous elastic material using Finite Element Model (FEM) simulation. We then formulate the generation of unknown external forces that de fine the deformation, the geometric, and material parameterization required to achieve the recorded deformation as a non-linear optimization problem. In this formulation the geometric distribution (quality) and density of tetrahedral components is optimized with the elastic material parameters (Young's Modulus and Possion's ratio) of the procedurally generated FEM model to provide the optimal deformation behavior with respect to the recorded surface. [PPTX] [PDF]

Shane Transue and Min-Hyung. Choi, "Deformable Object Behavior Simulation derived from 3D Surface Reconstructions", in 11th International Symposium on Visual Computing, p. 474-485, December, 2015.

Abstract – In this paper we present an adaptive and intuitive methodology for controlling the localized deformations of physically simulated objects using an intuitive pattern-based control interface. To maximize the interactive component presented in this approach we consolidate existing feedback mechanisms in deformable-body control techniques to provide intuitive editing metaphors for stretching, bending, twisting, and compressing simulated objects. The resulting movements created by these control metaphors are validated using imposed behavior evaluation and the effectiveness of this approach is demonstrated through interactively generated compound movements that introduce complex local deformations of objects in existing physical animations. [PPTX] [PDF]

Shane Transue and Min-Hyung Choi, "Interactive Control of Deformable-object Animations through Control Metaphor Pattern Adherence", in International Conference on Computer Graphics Theory and Applications, April, 2015.

Abstract – Throughout the course of several years, significant progress has been made with regard to the accuracy and performance of pair-wise alignment techniques; however when considering low-resolution scans with minimal pairwise overlap, and scans with high levels of symmetry, the process of successfully performing sequential alignments in the object reconstruction process remains a challenging task. Even with the improvements in surface point sampling and surface feature correspondence estimation, existing techniques do not guarantee an alignment between arbitrary point-cloud pairs due to statistically-driven estimation models. In this paper we define a robust and intuitive painting-based feature correspondence selection methodology that can refine input sets for these existing techniques to ensure alignment convergence. Additionally, we consolidate this painting process into a semi-automated alignment compilation technique that can be used to ensure the proper reconstruction of scanned models. [POSTER] [PDF]

Shane Transue and Min-Hyung Choi, "Intuitive Alignment of Point-clouds with Painting-based Feature Correspondence", in 10th International Symposium on Visual Computing, December, 2014.

Abstract – The process of accurately aligning 3D range scans to reconstruct a virtual model is a complex task in generic circumstances. Yet by exploiting the data characteristics common to many mobile 3D scanning devices, we propose a two phase alignment solution that improves the alignment provided by the iterative closest point (ICP) algorithm. Current approaches target how the ICP algorithm aligns two range scans based on modifying minimization functions, sampling functions, and point correspondence techniques. However, while these approaches have provided subtle improvements in the alignment process, the ICP algorithm is still incapable of aligning low resolution range scans with very little overlap. Based on our proposed algorithm, we are able to increase the accuracy of the alignment provided by the ICP algorithm by 40% on low resolution scan pairs and we demonstrate the versatility of this approach by accurately aligning a variety scan pairs with small overlap regions. [PPTX] [PDF]

Shane Transue and Min-Hyung Choi, "Enhanced Pre-conditioning Algorithm for the Accurate Alignment of 3D Range Scans", in International Conference on Image Processing, Computer Vision, & Pattern Recognition, July, 2013.


Abstract – Respiration volume has been widely used as an important indication for diagnosis and treatment of pulmonary diseases and other health care related issues such as critically ill patients neonatal ventilation, post-operative monitoring and various others. Most of existing technologies for respiration volume monitoring require physical contact with the human body. While wireless-based approaches have also been discussed in the literature, there are still limitations in terms of estimation accuracy and time efficiency preventing these approaches from being realized in practice. In this paper, we present an automated, wireless-based, vision-supervised system, calledWiKiSpiro, for monitoring an individual’s respiration volume. In particular, we present a system design encompassing a wireless device, motion tracking and control system, and a set of methods to accurately derive breathing volume from the reflected signal and to address challenges caused by body movement and posture changes. We present our preliminary results of WikiSpiro, and identify possible challenges for future research and development. [PDF] - Best Paper Award

Phuc Nguyen, Shane Transue, Min-Hyung Choi, Ann C. Halbower, and Tam Vu. "WikiSpiro: Non-contact Respiration Volume Monitoring during Sleep". Proceedings of the Eighth Wireless of the Students, by the Students, and for the Students Workshop, Pages 27-29, New York, NY, 2016.


Abstract – To increase the level of interactivity in the generation of physically-based animations, we present an adaptive and intuitive methodology for controlling the localized deformation of physically simulated objects using an intuitive motion-based control interface. We achieve this control through the dynamic recording of physically simulated deformable objects and the development of high-level motion controls that provide effective manipulation techniques for altering the animation of deformable objects. To maximize the interactive component presented in this approach we consolidate existing feedback mechanisms in deformable-body control techniques to provide an intuitive simulation editing environment. We introduce the notion of control metaphors as the abstract formulations of primitive motions enacted by deformable-bodies when external forces modify the physical state of the object. As an application of this proposed control methodology we develop a practical solution for interjecting local deformations into dynamically recorded deformable objects. The effectiveness of this approach is demonstrated through interactively generated compound movements that introduce complex local deformations of the objects in existing physically-based animations. Additionally we validate the resulting movements imposed by the control metaphors by using directed behavior demonstrations through physical animations.

Shane Transue, "Interactive Control of Deformable-object Animations with Intuitive Motion Pattern Adherence", Master's Thesis, University of Colorado Denver, May, 2014.

Technical Reports

Abstract – Surface reconstruction from images based on the premise that geometric structures can be extracted using Shape From Shading (SFS) techniques is a well studied area within digital image processing. Most existing shape from shading methods focus on improving the quality of the surface that can be extracted based on improving surface curvature and depth estimations, however a new application of these techniques that has not been extensively studied is the application to surface deformation reconstructions of volumetric objects. The primary objective of this project is to: (1) explore current shape from shading techniques for surface reconstruction, (2) apply these techniques to deformation recording, and (3) improve surface curvature estimates of depth-images. Within the implementation of this project, several different techniques within digital image processing were used to reconstruct the animation of a deforming surface recorded using a Microsoft Kinect2. These include: intensity image acquisition, finite-difference based spatial convolution for image derivatives, normal map generation, and displacement map generation through field integration. The generated surfaces illustrate that the surface of a recorded deforming object can be accurately reconstructed using the proposed implementation. [PDF]

Abstract – In this work we introduce and construct a framework of parallel algorithms that facilitate the parameterization and simulation of incompressible particle-based fluids. This parallel simulation framework is composed of three components: (1) the parameterization of the initial states of the simulated particles derived from Fourier analysis, (2) a set of discrete and continuous collision detection algorithms for particle and convex object interactions based on a newly introduced cluster-based event system, and (3) an impulse-based collision response algorithm for rigid-particle interaction. In the implementation of this framework we utilized both grouped SIMD facilitated through the Compute Unified Device Architecture (CUDA) and MIMD techniques through OpenMP to provide parallel implementations of efficient algorithms for each of these components. The collected results are demonstrated through the parameterization of waveforms to drive particle visualizations and the interactive simulation of particle-based fluid animation. Through the parallel implementations of our algorithms, we have been able to reduce our overall simulation time by approximately 80% demonstrating a 5.37x increase in performance compared to the sequential implementation.

Shane Transue and Shannon Steinmetz, "Parallel Fourier Parameterization and Impluse-based Cluster Collisions in Incompressible Particle-based Fluid Dynamics", in Parallel and Distributed Computing, University of Colorado Denver, 2014.

Abstract – With the development of the uniform processor designs implemented in modern graphics processing units (GPUs) released by NVIDIA and AMD, the expansive development of parallel languages and frameworks allows us to create high performance parallel applications that exploit the architectural characteristics of these devices. Specifically, we look at the modern GPU memory hierarchy, the development of parallel algorithms in OpenCL and how fast on-chip local memory can be targeted to significantly increase the performance of modern parallel applications. We develop a standard benchmark, standard matrix multiplication in OpenCL, to analyze the desired performance gains achieved by using this memory. We have found that through the analysis of the elapsed execution time of our benchmark, the utilization of local on-chip memory versus global memory utilization has provided ~90% performance increase.

Shane Transue, Alejandro Alonso, and Sukthana Pongma, "Performance Analysis of GPU Memory Architectures with Standard Matrix-Multiplication in OpenCL", in Advanced Computer Architecture, University of Colorado Denver, 2013.