Search

This study focuses on 3D human motion recovery from a sequence of video frames by using the exemplar-based approach. Conventionally, human pose tracking requires two stages: 1) estimating the 3D pose for a single frame, and 2) using the current estimated pose to predict the pose in the next frame. This usually involves generating a set of possible poses in the prediction state, then optimizing the mapping between the projection of the predicted poses and the 2D image in the subsequent frame. The computational complexity of this approach becomes significant when the search space dimensionality increases. In contrast, we propose a robust and efficient approach for direct motion estimation in video frames by extracting dense appearance and motion features in spatio-temporal space. We exploit three robust descriptors - Histograms of Oriented Gradients, Histograms of Optical Flow and Motion Boundary Histograms in the context of human pose tracking for 3D motion recovery. We conducted comparative analyses using individual descriptors as well as a weighted combination of them. We evaluated our approach using the HumanEva-I dataset and presented both quantitative comparisons and visual results to demonstrate the advantages of our approach. The output is a smooth motion that can be applied in motion retargeting. MOE (Min. of Education, S’pore) Accepted version

Different regularities have been used in the reconstruction of a 3D object from a single-view line drawing. These regularities are not all equally informative in the reconstruction process: certain regularities may correspond mainly to noise, not information; some may overlap with each other or are not too relevant to the reconstruction. This paper studies these regularities comprehensively, so as to select the most effective set that can give robust and reliable 3D reconstruction. The selection is made through a method called automatic relevance determination (ARD), which employs the Bayesian framework and support vector regression estimation. The proposed method is able to identify the worst regularities according to their ARD parameters and eliminate them. The effectiveness of this pruning is evaluated by model validation. The regularity set so obtained is effective for general 3D reconstruction. The experimental results show that the regularity set selected can reduce the reconstruction complexity and produce satisfactory reconstruction performance.

The geometry of an object is formed by its surface boundary over a topological structure. In a companion paper we proposed a deformable volume model in conjunction with an isosurface evaluation in a new approach to surface boundary construction. This paper further discusses the new approach, concentrating on modeling the topological structure of objects. We propose a formal method to describe the topology of objects using a Context-Free Grammar Gpd. Such topology descriptions can be interpreted by a Push-Down Automaton Mtg to generate template objects with the desired topological structures. The template objects will be refined to match the geometry of actual objects.

The objectives of this research programme are to: 1. Establish the fundamental infrastructure - software and hardware - required to support advanced research in and application of rapid design, prototyping and tooling. 2. Facilitate research and development to solve the problems identified, including integration of the processes involved and the study of new methods for rapid prototyping and tooling. 3. Develop new applications using rapid prototyping, especially medical applications. 4. Establish the core expertise in the field and generate new knowledge and processes. 5. Produce trained personnel who can be deployed in local industry. 6. Serve as a resource centre for Singapore industries and institutions in the use of the technologies.

The concept of groupware is not new. This paper discusses issues about realizing it and proposes a collaborative programming environment as an approach. To explain how to establish such an environment, various aspects of collaborative programming such as modules, data templates and wrappers are introduced. We also examine the syntax and semantics of visual languages and show how programming becomes straightforward, yet flexible in visual programming with these modules. Collaborative programming is illustrated by the different phases of the development of an application system. Some advanced topics are discussed, and the paper is concluded by indicating that great effort should be made towards a sophisticated collaborative programming environment. >

This project, ARC 6/94, is concerned with the technology and process development of rapid 3D object forming system and materials and process technology. Its purpose is to meet increasing demanding requirement from manufacturing industries to shorten the design-to-product cycle time and improve design and product in shorter time.

This paper discusses a method that combines 5-axis NC machining and rapid prototyping for manufacturing complex shape parts. The combination allows NC machining to deal with surfaces in a part that cannot be accessed directly by a 5-axis NC machine. It splits the part into fully accessible subparts (layers), and machines them one by one. The subparts are subsequently joined together to form the complete part.

Various tomographic imaging systems start data processing with a series of 2D images acquired from the target objects. To describe the geometry of complex structures, such as in biomechanical and orthopaedic investigation, we need methods to reconstruct 3D surface boundaries of the objects from 2D contours in the images. Of primary importance is an effective and efficient data structure for representing these free-form surfaces. This paper presents a representation scheme for free-form surface patch by a 3D curvilinear lattice. Coupled with the Marching Cubes algorithm, the new data structure provides a uniform representation for different types of surfaces and its output can be visualized using a normal graphic renderer. The method considerably reduces memory required for boundary generation compared to the conventional way where iso-surfaces are evaluated from a volume. It also speeds up the surface rendering procedure by the reduction of surface triangulation. We illustrate its abilities in handling surfaces with different topological structures.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Compared with other medical imaging modalities, ultrasonic imaging has its advantages of safety, low cost, and very importantly, real-time performance in data acquisition. To match the real-time data acquisition, interactivity is considered as one of the crucial characteristics for 3D ultrasonic imaging systems. This paper proposes an incremental refinement method in a multi-process system with an eye on ultimately running real-time 3D ultrasonic imaging applications. We use deformable geometric models as templates of the target biomedical objects. With a matching and refining procedure, the geometric data will be incrementally refined on the arrival of each injected boundary from 2D ultrasonic slice. All the data processing tasks are distributed to a set of processes which are connected in a graphic environment. Each process incrementally reads the data received from its upstream and produces new data for its downstream. By simultaneously running different processes, the system gradually transforms the 3D templates to match the detected ultrasonic images.

Convex programming has played an important role in studying a wide class of applications arising from computer science, statistics, industrial engineering, and management. Moreover, the advent of big-data analytics has resulted in very large-scale structural convex programming problems, thereby necessitating the research towards devising fast numerical algorithms for solving such problems. The goal of this thesis is to propose some efficient augmented Lagrangian-based splitting methods for solving the convex programming problem, provide an in-depth study of the convergence and convergence rate properties of these algorithms, and discuss computational results that establish the efficiency of these methods. Towards this end, we first consider the linearly constrained separable convex minimization problem, whose objective function consists of the sum of m individual convex functions in the absence of any coupling variables. This work shows that a straightforward Jacobian decomposition of the augmented Lagrangian method is globally convergent if the involved functions are further assumed to be strongly convex. Furthermore, we also analyze the sublinear and linear convergence rate under some corresponding assumptions. Then, we propose a proximal partially parallel splitting method for solving separable convex optimization problems. The proposed method is a hybrid mechanism that combines the nice features of Gauss-Seidel and Jacobian decomposition methods, while simultaneously adopting the predictor-corrector strategy to ensure convergence. Furthermore, the worst-case convergence rate O(1/t) of the proposed method is obtained under both ergodic and non-ergodic conditions. The efficiency of the proposed algorithm is also demonstrated by solving several instances of the Robust PCA problem. Next, continuing in the same vein, we present a portfolio of linearized versions of this method, which can be gainfully employed when the underlying subproblems possess some structural properties, and showcase the computational effectiveness of this method on standard test instances taken from the literature. Doctor of Philosophy (MAE)

The objective of this thesis is to develop a sketching-based CAD prototype for 3D shape modeling which can effectively and efficiently convert an input sketch into a 3D geometric model. In this thesis, the work concentrates on general, planar objects reconstruction from a single view line drawing. It addresses some key problems in the optimization-based 3D reconstruction. Firstly, the basic issues in formulating the compliance function are explored under the framework of support vector machine regression (SVR). Secondly, automatic relevance determination, Bayesian framework and SVR are coupled together to form a regularity selection method. An optimal regularity set is established from the commonly used regularities. Thirdly, a hybrid intelligent 3D reconstruction system is built to handle various line drawings. Line drawings are classified into 4 classes using the technique of support vector machine multi-classification. Effective reconstruction is performed by the “reconstruction experts” which are tuned to deal with each class specifically. Finally, the issues in the topological analysis of a line drawing are investigated, especially in distinguishing manifolds from non-manifolds.The experimental results confirm that the proposed methods can reliably resolve the 3D reconstruction issues. The contribution of this research is to improve the efficiency and effectiveness of 3D reconstruction with the proposed methods by applying advanced artificial intelligence techniques. DOCTOR OF PHILOSOPHY (MPE)

Variability is the greatest enemy of a manufacturing system, which causes unstable performances and varies the productivity. This report focuses on reducing the variability and improving the performance of a semiconductor manufacturing process. Other methods to improve the performance of a semiconductor manufacturing process includes improving machine's process time, standard work of operators and other ergonomics factors; these are included in this study. A simulation model of the manufacturing semiconductor cell was created using the Arena software. It was based on actual production process data from the author's company Heptagon. To get different performances of the manufacturing system, some scenarios of variations had been applied on the simulation model. The results were used in further analysis and discussion to investigate the factors that produce the best improvement or cause the worst productivity in the manufacturing system. The author proposed a set of solutions to improve the performance of the semiconductor manufacturing system. These include improvements on number of operators, walking speed of operators, process analysis time reduction and machine modification time reduction. Master of Science (Manufacturing Systems & Engineering)

Protein complex conformation prediction is crucial for studying the biological systems and thus for drug design. However it is very hard to do the prediction because of the flexibility of the backbone and side-chains of the receptor and ligand. In our work, the “divide and conquer” scheme has been utilized for protein-protein docking, with an initial stage which focuses on generating a reasonable list of near-native candidate docking conformations, and a refinement step which aims to rank the hit within the first several positions by incorporating the protein flexibility into the docking procedure. In the initial docking stage, we present a new framework for an efficient and rigid-body docking in view of geometric complementarity. In our approach, the protein surfaces are firstly segmented into several local surface patches. Then the geometric complementarity can be determined by matching these surface patches. Based on the geometric property, we convert the geometric complementarity matching to geometric similarity comparing. During the matching process, we extract the shape feature for each surface patch, thus the geometric similarity comparing problem will be further simplified as a histogram matching problem. After finding the patch pairs with geometric complementarity, a list of docking candidates can be generated. Finally, the candidate solutions are ranked by a scoring function to filter out the near-native candidate conformations. In the refinement step, to deal with the protein flexibility and possible conformational changes during binding, many approaches have been proposed. Some of them allow some degree of overlap on the interface region and some algorithms simulate the soft docking by smoothing the protein surfaces. In our approach, we will deal with the side-chain flexibility by sampling its conformational space. This will be carried out in our future work. In this work, our major contribution is that we propose to apply an innovative three-dimensional shape descriptor, Spherical Harmonics Descriptor (SHD) in matching. The main property of SHD is rotation invariance that helps to avoid undergoing a very thorough series of rotations between each pair of surface patches. This can greatly reduce the computational cost. The experimental results illustrate the high efficiency and accuracy of our method. Master of Engineering (SCE)

Reconstructing the 3D object depicted in a line drawing consistent with human perception is an interesting and important topic in computer vision. Normally, the reconstruction process requires two main steps: topological analysis and 3D recovery. This thesis makes three novel contributions in the three steps: (1) Decomposition of the line drawing of a complex object into a collection of drawings of simple manifolds; (2) Finding the faces of the object in the drawing; (3) Reconstruction of the object in both parallel and perspective projections. The work here is restricted to drawings of planar polyhedra. The decomposition of a complex line drawing into drawings of simpler manifolds occurs at (1) non-manifold vertices, (2) along non-manifold edges and (3) across internal faces. The algorithms for the first two are simple. The decomposition across an internal face has two stages: basic decomposition and the subsequent repair. Our results show that the algorithm can deal with objects with planar faces. Reconstructing of the 3D object is much easier and much more efficient from a simpler drawing than a complex one. Knowing the faces of the object in a drawing is essential to the 3D reconstruction of the object. A new face finding algorithm much more efficient than the existing ones is presented here. This algorithm also contains two stages: decomposition and face forming. In the first stage, a line drawing is decomposed into sub-loops, which are connected but usually open sequence of edges that must belong to the same face. Faces are formed from the sub-loops in the next stage. The method has the capability to differentiate an internal face from a real one once it is formed. After obtaining the faces, the next step is to reconstruct the 3D model. Here, we present a novel, simple and direct method, the cubic corner method (CCM). A cubic corner is a trihedral vertex where the three incident edges are mutually perpendicular. The 3D coordinates of the three end vertices relative to the given vertex can be obtained analytically given the 2D coordinates of the four vertices in the drawing. The algorithm starts with a user-identified cubic corner in the drawing and uses the face information obtained previously to recover the geometry of the entire object. It works for both parallel and perspective drawings; the only difference is that, for perspective, the method needs to find the focal length first, to establish the linear relationship between a vertex on the image plane and its corresponding point on the object. The method generates accurate results for accurate drawings. For inaccurate drawings, it generates plausible results which may not be good enough. A novel hybrid method, combining CCM and the well-known optimization-based method (OBM), overcomes the problem. The optimization process in OBM requires an initial guess to the 3D object; a guess close to the final solution could improve the result of OBM significantly. In the hybrid method, CCM supplies this initial guess to OBM, which then produces a result better than either method on its own. This thesis also presents an alternative efficient method that recovers 3D objects from perspective drawings by reversing the process one uses for drawing a perspective drawing manually. The method does not require the focal length and, depending on the topology of the drawing itself, may not need the faces of the object. This is a significant advantage as face identification is not a trivial task. Future work is required to resolve some of the shortcomings in the algorithms presented in the thesis. Besides, there is a need to overcome the geometric restriction of planar polyhedra only, and extend the geometric coverage to objects that contain both planar and curved faces. Also, the ability to recover line drawings accurately opens the possibility of recovering objects in photographs, if we can extract edges accurately from photographs. This brings forth the need for a stitching algorithm that merges reconstructed results together to form the whole object as the complete view of an object normally requires multiple photographs. This can then lead to some useful applications such as the reconstruction of engineering objects accurately, without using the expensive 3D laser scanner. Much work remains before we can arrive at that. DOCTOR OF PHILOSOPHY (MAE)

Safety Management Systems (SMS) have been the primary driver to ensure safety of personnel in the workplace and to achieve the goal of "Zero Harm", which majority of the Liquefied Natural Gas (LNG) operators embrace. However, people are still being seriously harmed in the workplace despite of stringent safety policies, rules, procedures and processes. This dissertation explores the integration of Human Factors (HF) with SMS to create a safer workplace and to achieve the Zero Harm goal. A comprehensive approach that encompassed the entire life cycle of an LNG facility, from its conceptualization, design, construction, to its operation and maintenance was taken. The analysis also focused on potential errors that humans may make from time to time, to minimise the consequences on harm to people, damage to assets, or polluting the environment. Due to the broadness of the subject area, this study focuses only the integration of HF and SMS in the design phase. The approach and results of this study may then be applied to the other phases of the life cycle. An in-depth review of current SMS, Incident Database and case studies was taken to identify HF issues leading and/or contributing to the incidents. It showed that up to 84% of identified root causes and accidents were attributed to Human Factors. In pursuit of the Zero Harm Goal, the implementation of a systematic approach in integrating HF and SMS into the entire Lifecycle of a new LNG facility eliminates or reduces the 84% related causes attributed to Human Factors and can potentially have a significant impact on achieving the Zero Harm Goal in the workplace and cost savings from injuries and retrofitting work. By early intervention to integrate the proposed model with each phase of the project's lifecycle, human performance is expected to improve, as the operators' specific needs with respect to the operation of the plant, equipment and systems would have been resolved. Master of Science (Human Factors Engineering)

The objective of this research is to recover 3D freeform surfaces depicted in a single line drawing, with the aim of taking conceptual design sketches and converting them into 3D models. This work is a continuation of the project “3D Surface Reconstruction from a 2D Line Drawing” (SRLD) done at Nanyang Technological University. SRLD is a system to reconstruct planar objects. The work in this thesis extends the capability of SRLD by recovering freeform shapes. In this thesis, a novel method of reconstructing freeform surface patches from 2D line drawings is introduced. The study focuses on drawings containing four edges that form the boundary of surface patches. The four edges form two pairs of opposite edges which are used as the basis of interpolation to create two sets of backbone curves. After recovering all the 3D curves of one set of backbones, a freeform surface is fitted over them. Controls are provided within the system to adjust the shape of a surface; this is a useful tool given the non-unique nature of the result. An optimization-based method for reconstructing surface patches, cylinder-like shapes and their combination with polyhedron from its projection on a drawing plane is also presented. The rough drawing is first transformed into a straight line drawing. Then, a polyhedron is rebuilt from the straight line drawing. Finally, planar objects are converted back to freeform shapes. Our implementation of the method produces results that are consistent with human perception of the curves as surfaces. Due to the inherent inaccuracies in the input, the models reconstructed are likely to be aesthetically unpleasing. A post-processing is necessary to improve these curved models aesthetically. Our work focuses on improving objects with symmetrical curved boundaries. Symmetrical curved boundaries are detected and fitted to two exactly symmetrical curves on the drawing plane. And then a planar object is reconstructed and beautified from its converted straight line drawing. Finally, curved surfaces are recovered. Experiments show that our method is able to improve shapes with planar and curved surfaces reconstructed from its projection on a drawing plane well. Doctor of Philosophy (MAE)

There is an insufficient coverage of existing strategy to cater the highly personalised nature of industrial equipment in customer's requirement and service delivery. The right after sales service strategy adoption are vital to enable industrial equipment manufacturer maintaining. its competitive advantages and delivering the best overall company performance. The objective of this dissertation project is to study, formulate a competitive after sales service strategy and system guide for industrial equipment manufacturers. A practical system guide further gives industrial manufacturer a comprehensive view in devising and assessing the most appropriate after sales strategy. Literature in general after sales and industrial equipment specific are explored. After sales strategic profile and framework, service configuration and process as well as its performance measurements system literatures are carefully analysed to establish a theoretical framework for competitive strategy in after sales service. It will be used as a reference for developing the after sales competitive strategic system guide. A survey is conducted to verify the proposed strategy and system guide relevancy to real industrial application. Questionnaires obtained from respondents of two industrial equipment manufacturers are studied and compared to the proposed strategy and system guide. The survey shows the relevancy between this study and the real industrial application. Both companies apply although not closely, the proposed system guide in devising their strategic objectives, Their after sales configuration, process and delivery show the relevancy to the proposed system guide. Future research with larger respondents across industry sectors, state-ofthe- art data testing and verification can be initiated to provide more comprehensive study. It is important to explore the new business concept to adapt to ever changing world of the service industry. Master of Science (Computer Integrated Manufacturing)

In this dissertation, research efforts had been conducted for air cargo operational defects process capturing and its timely rectification for smooth operation. Research was conducted by making field observations, interview with people working in the industry, understanding of air cargo operations and its operational issues and problems. A review of research and conference papers was also carried out. The results were consistent that a combination of network and discrete GUI system ware will greatly augment air cargo operations and enable air cargo operators to have a strong awareness of issues that can impact cargo flights. One of the other major changes is the relentless process of integrating the avionics systems, which is blurring the division between some of the functions of the flight management system and EFB. The current EFB processes information from aircraft sensors and navigation equipment. There is a graphical interface between the system and the crew who are using the equipment. Majority of the function is mostly to protect the integrity of the Flight Management System from virus contamination by the constantly updated information contained in digital charts, notices to airmen, aviation information systems, real-time weather information services, flight manuals. This also including weight and balance, performance calculation criteria and checklists. One area, which is not yet covered by the EFB in the aviation industry, is the main deck cargo handling system status display. This information is crucial to cargo airline operations due to their importance of revenue generation. The current main deck cargo handling system currently has no displays of the electrically powered units (PDU) that move the pallets of cargo. These powered drive unit facilitate movement of cargo throughout the main deck compartment. It is accomplished by two different type of PDU. First, is by power drive wheels or second, by power drive unit actuators. All of these are currently installed in the main deck floor of the airplane. The PDUs are controlled from the cargo control panels installed on the sidewalls on each side of the compartment. All of these PDU defects are recorded in paper logbook for. Due to the schedule constraint in flight transit down time, logbooks are often not recorded properly. This resulted in inaccurate PDU defect recording and in worst-case scenario, eventual cargo offloading. These are heavy losses to the cargo airline on accumulation or occurrences. One way to improve this situation will be converting these defect logbook recordings into EFB defect entries for real time defect display and monitoring via graphic user interface. Another advantage of EFB logbook is its paperless operations, which saves weight. Master of Science (Mechanical Engineering)

When humans and automation work together, the human typically takes a supervisory role and rely on the automation to complete certain tasks. However, there are times when the human supervisor cannot understand what the automation is doing. This is complicated by the fact that automation can sometimes fail, which makes the actions of the automation difficult to comprehend. This thesis presents information for calibrating operator's trust and dependence on automation for improving decision performance. The Motivation-Expectation Space (MES) representation is proposed as a tool to aid the monitoring of automation. The literature review suggested that the four components in the MES (goals, means, causes, and effects) are fundamental in how people naturally strategize decisions. The review also revealed that the MES can support most of the information requirements needed for bridging automation trust and dependence. A case study was performed to demonstrate how data obtained through a cognitive task analysis (CTA) could be translated to the MES representation. This was previously demonstrated in a petrochemical context where work constraints are limited primarily to the laws of physics (e.g. temperature). In this thesis, the translation was repeated for an in-car navigation task to gain insights in the MES application for a context with intentional constraints. The case study concluded with (1) a specific definition for each relation depicted in the MES, and (2) the design of the MES representation for an in-car navigation aid. A three group between-subject experiment was conducted to test the effectiveness of the information and orthogonal structure in the MES. The information in the MES was presented in either the orthogonal MES format or textual formats. DOCTOR OF PHILOSOPHY (MAE)

The study aims at establishing the topological and geometrical relationships that govern the design and construction of pop-up laminar structures. It also proposes a systematic, computer-aided approach for pop-up designs. A pop-up classification model and graph representation schemes are established after a comprehensive study on the available pop-up structures. The folding and development of pop-up structures are analyzed and represented as Mountain-Valley (MV) graphs. The MV graphs illustrate relationships between layers and creases and define the boundary regions of feasible pop-up design. Topological conditions for governing the composition of the structures are also established using vertex-edge graphs. In addition, conditions for flat folding are derived after investigation into the geometry of the structures. While the graphs enable simple representation of pop-up structures without the complexity of geometric designs, the mathematical relationships permit the validation of the structures’ composition and their ability to flat fold. A modular, object-oriented software prototype has been built, and it demonstrates the creation of pop-up designs using the graph structures and the mathematical relationships developed in this research. DOCTOR OF PHILOSOPHY (MAE)