Interactive Physically-based Animation System for Dense Meshes

Interactive image deformation is an important feature of modern image processing pipelines. It is often used to create caricatures and animation for input images, especially photos. State-of-the-art image deformation techniques are based on transforming vertices of a mesh, which is textured by the input image, using affine transformations such as translation, and scaling. However, the resulting visual quality of the output image depends on the geometric resolution of the mesh. Performing these transformations on the CPU often further inhibits performance and quality. This is especially problematic on mobile devices where the limited computational power reduces the maximum achievable quality. To overcome these issue, we propose the concept of an intermediate deformation buffer that stores deformation information at a resolution independent of the mesh resolution. This allows the combination of a high-resolution buffer with a low-resolution mesh for interactive preview, as well as a high-resolution mesh to export the final image. Further, we present a fully GPU-based implementation of this concept, taking advantage of modern OpenGL ES features, such as compute shaders.

Simulating cloth in real-time is a challenging endeavour due to the number of triangles necessary to depict the potentially frequent changes in curvature, in combination with the physics calculations which model the deformations. To alleviate the costs, adaptive methods are often employed to refine the mesh in areas of high curvature, however, they do not often consider a decimation or coarsening of areas which were refined previously. In addition to this, the triangulation and consistency checks required to maintain a continuous mesh can be prohibitively time consuming when attempting to simulate larger pieces of cloth. In this paper we present an efficient edge-based approach to adaptively refine and coarsen a dynamic mesh, with the aim to exploit the varied nature of cloth by trading the level of detail in flat parts for increased detail in the curved regions of the cloth. An edge-based approach enables fast incremental refinement and coarsening, whereby only two triangles need updating on each split or join of an edge. The criteria for refinement includes curvature, edge length and edge collisions. Simple collision detection is performed allowing interactions between the cloth and the other objects in the environment.

In this paper we present GPU-based techniques for the interactive visualization of large unsteady 3D flow fields on uniform grids. We propose a novel dual-core approach to asynchronously stream such fields from the CPU, thus enabling the efficient exploration of large time-resolved sequences. This approach decouples visualization from data handling, resulting in interactive frame rates. Built upon a previously published GPU particle engine for flow visualization we have developed new strategies to compute and to visualize path lines and streak lines on the GPU. To provide additional visual cues, focus+context techniques for polygonal meshes have been integrated. The proposed techniques are used in the visual analysis of the Terashake 2.1 earthquake simulation data, and they have been shown to be very effective in revealing the relevant information in this data.

A general hybrid geometric-physical method that is useful for animating characters with skin, clothing and a MPEG-4 compatible face is presented. This method uses an approximation to the underlying musculo-skeletal structure of the body and face to generate a distance field, used for collision detection purposes, and skin/clothing consisting of a particle-spring mesh. The results obtained deliver animation with plausible dynamics for fairly detailed models at around 60fps on a PIII computer with Nvidia Gforce2 graphics. The main contribution is to show that similar algorithms can be used for skin, clothing and facial animation, at interactive rates.

Rugby union requires players to perform high-intensity locomotor and contact efforts, interspersed with low-intensity activity. Locomotor efforts include accelerations, running and sprinting, while collision efforts include ruck, tackle scrum and maul events. Recent research has quantified the demands of Rugby Union using player-worn microtechnology that contains global positioning systems (GPS) and tri-axial microsensors including accelerometers, magnetometers and gyroscopes. To date, research has extensively reported the locomotor demands of Rugby Union match-play using GPS, documenting total distance covered, high-speed distance, accelerations and running efforts. However, there is a lack of research on the contact events of match-play. A number of authors have investigated the contact demands of Rugby Union using microtechnology and applying non-specific algorithms to determine the number of collision events in Rugby Union. However, two major limitations exist in this approach. Firstly, while these algorithms have been validated for other collision sports (i.e. Rugby League), the unique collision events of Rugby Union mean they are unsuitable for this sport. Secondly, the developed algorithms do not delineate among contact events (i.e. ruck, scrum and maul), resulting in an underestimation of the contact demands of Rugby Union and all collision events being considered equally demanding. Therefore, the total physical demands of Rugby Union match-play are being under reported. Based on the identified gaps in the literature, the purpose of this thesis was to 1) conduct a systematic review of the use of microsensors to quantify sport-specific movements and determine if such devices are potentially capable of detecting collision events in Rugby Union (Study 1), 2) develop a valid algorithm to detect scrum events in training and match-play (Study 2); and 3) develop an algorithm to determine ruck and one-onone tackle events in Rugby Union (Study 3). This program of research was subsequently brought together by the final study (Study 4), which applied the newly developed algorithms with existing methods to uniquely quantify the locomotor and contact demands from both winning and losing teams in 12 elite matches. Results of this research provides a novel insight into the contact demands of elite Rugby Union and additionally provide validated methods to delineate each collision type. Results of this research provide a detailed overview of total physical demands of Rugby Union and provide insight into the different locomotor and collision profiles of winning and losing teams. This research demonstrates a unique application of microsensors and specific algorithms to quantify the collision demands of elite Rugby Union training and matchplay. For the first time, the total locomotor and contact demands of elite Rugby Union match-play, including those of winning and losing teams have been documented. Performance staff can use this information to more effectively monitor the training loads of players and design sport-specific conditioning programs to prepare players for the most demanding passages of match-play.

Haptic technology is quite recent and therefore in many cases it is difficult to simulate real contacts or interactions with a high sensation of realism. Collision response methods that calculate the force-feedback tend to cause haptic instabilities when the normal direction changes abruptly. In consequence, collision events in sharp corners are often difficult to render properly by means of haptic devices. This paper describes a collision response method that first distinguishes the nature of the corners -concave or convexand then provides an adequate solution in each case to avoid haptic instabilities.

Mankind has been using ships for more than 5,000 years and has developed a range of related technologies. However, despite such a long history, compared to aircraft with a history of approximately one century, the pace of progress has been markedly slow. Even though technological progress of ships or the installation of various navigation equipment have been achieved, seaborne collisions have occurred quite frequently. This study analyzed the TCAS( Traffic Collision Avoidance System) that has contributed to the prevention of collisions with other transport methods including aircraft to suggest a collision avoidance system that can be deployed for ships. To apply the technologies applied to aircraft that move in 3D to ships that move in 2D, the difference in the operational environment between the two modes was analyzed to identify elements that need to be applied to ships. The suggested display of data on the collision prevention system is one that manipulates the augmented reality display device used in automobiles that over the past few years has undergone rapid development. Based on the presentation of technological elements that need to be considered when adopting the SCAS or the Seaborne Collision Avoidance System as suggested in this study, the authors hope to contribute to the prevention of collisions.

The pavement skid resistance, namely the vehicle tire-road contact friction, plays important roles on pavement response, pavement deterioration, traffic operation, and traffic safety. One may find higher tire-road friction surface course is needed or applicable as an effective safety countermeasure for reducing potential traffic crashes and/or collision severity at certain roadway locations, in particular, the ramp emerging or weaving area where foreseeable sideswipe or rear-end collisions might take place between a merging vehicle and another vehicle from the outside lane. A driver should decelerate when merging become difficult as trained in a licensing procedure for avoiding potential collisions. If the required deceleration length is shortened upon avoiding a potential collision, the number of collisions at the onramp should decrease and the collisions if occurred would be less severe. The tire-road frictional coefficient if raised high enough could dramatically reduce the required deceleration length prior to collision, and the high friction surface treatment (HFST) has been introduced in practice for the last decade to shorten the required distance to stop a moving vehicle for both wet and dry road surface conditions. It is of significant interest to examine quantitatively, based on a proposed integrated framework, the effect of a high frictional surface course on reducing the needed deceleration distance, in particular, on an onramp merging/weaving section constructed with limited right of way.

The 2D real time optimally adapting meshes (ROAM) algorithm has had wide success in the field of terrain visualization, because of its efficient error-controlling properties. In this paper, we propose a generalization of ROAM in 3D suitable for scalable volume visualization. Therefore, we perform a straightforward 2D/3D analogy, replacing the triangle of 2D ROAM by its 3D equivalent, the tetrahedron. Although work in the field of hierarchical tetrahedral meshes was widely undertaken, the produced meshes were not used for volumetric rendering purposes. We explain how to compute a bounded error inside the tetrahedron to build a hierarchical tetrahedral mesh and how to refine this mesh in real time to adapt it to the viewing conditions. We further show how to achieve cell sorting in linear time, thus yielding real time view-dependent display of the volumetric object. We present examples of large volume data sets and compare our approach with a similar one. Our results outline the high quality and computational efficiency of our approach.

Parallel rendering of large polygonal models with transparency is challenging due to the need for alpha-correct blending and compositing, which is costly for very large models with high depth complexity and spatial overlap. In this paper we compare the performance of raster-based rendering methods on mesh models of neurons using two applications, one of which is specifically tailored to the neuroscience application domain, the other a general purpose visualization tool with domain specific additions. The first implements both sort-first and sort-last and uses a scene graph style traversal to cull objects, and dual depth peeling for order independent transparency, whilst the other uses a simpler brute force data-parallel approach with sort last composition. The advantages and trade offs of these approaches are discussed.We present the optimized algorithms needed to achieve interactive frame rates for a non-trivial, real-world parallel rendering scenario. We show that a generic data visualization application can provide competitive performance when optimizing its rendering pipeline, with some loss of capability over an optimized domain-specific application.

Wrinkle simulation can reflect the texture and simulation effect of clothes. In this paper, we realize a new cloth wrinkle enhancement method based on the coarse mesh simulation. Firstly, we extend the deformation tensor to generate the mesh vertex stretch tensor, so as to obtain one measuring factor for the three-dimensional cloth mesh vertex deformation. Then, we introduce a new method to extract the wrinkle baseline in the 2D mesh which is stretching from 3D coarse mesh and mapping it to 3D coarse mesh. Finally, we use the convolution surface as framework to simulate the wrinkle shape. We also propose and realize the wrinkle surface adjustment method. Experiment result proves the effectiveness of this method, namely this method can significantly improve the cloth wrinkle effect based on the coarse mesh simulation.

Two level penalty finite element methods for the stationary incompressible magnetohydrodynamics problem

Representational accuracy of the isosurface meshes produced by well-known marching isosurfacing methods is considered. Accuracy is determined versus the isosurface given by trilinear interpolation and quantified using a new measure of volumetric divergence. The new measure is an accurate estimate of spatial discrepancy between a produced mesh and the isosurface of trilinear interpolation. Through experimental testing using the measure, suitability of the considered methods to well-model the isosurface of trilinear interpolation's behavior on grid cell interiors is also established.

Unmanned Aerial Vehicles are becoming increasingly popular for a broad variety of tasks ranging from aerial imagery to objects delivery. With the expansion of the areas, where drones can be efficiently used, the collision risk with other flying objects increases. Avoiding such collisions would be a relatively easy task, if all the aircrafts in the neighboring airspace could communicate with each other and share their location information. However, it is often the case that either location information is unavailable (e.g. flying in GPS-denied environments) or communication is not possible (e.g. different communication channels or non-cooperative flight scenario). To ensure flight safety in this kind of situations drones need a way to autonomously detect other objects that are intruding the neighboring airspace. Visual-based collision avoidance is of particular interest as cameras generally consume less power and are more lightweight than active sensor alternatives such as radars and lasers. We have therefore developed a set of increasingly sophisticated algorithms to provide drones with a visual collision avoidance capability. First, we present a novel method for detecting flying objects such as drones and planes that occupy a small part of the camera field of view, possibly move in front of complex backgrounds, and are filmed by a moving camera. In order to be solved this problem requires combining motion and appearance information, as neither of the two alone is capable of providing reliable enough detections. We therefore propose a machine learning technique that operates on spatio- temporal cubes of image intensities where individual patches are aligned using an object-centric regression-based motion stabilization algorithm. Second, in order to reduce the need to collect a large training dataset and to manual annotate it, we introduce a way to generate realistic synthetic images. Given only a small set of real examples and a coarse 3D model of the object, synthetic data can be generated in arbitrary quantities and further used to supplement real examples for training a detector. The key ingredient of our method is that the synthetically generated images need to be as close as possible to the real ones not in terms of image quality, but according to the features, used by a machine learning algorithm. Third, though the aforementioned approach yields a substantial increase in performance when using Adaboost and DPM detectors, it does not generalize well to Convolutional Neural Networks, which have become the state-of-the-art. This happens because, as we add more and more synthetic data, the CNNs begin to overfit to the synthetic images at the expense of the real ones. We therefore propose a novel deep domain adaptation technique that allows efficiently combining real and synthetic images without overfitting to either of the two. While most of the adaptation techniques aim at learning features that are invariant to the possible difference of the images, coming from different sources (real and synthetic). Unlike those methods, we suggest modeling this difference with a special two-stream architecture. We evaluate our approach on three different datasets and show its effectiveness for various classification and regression tasks.

Speed of information has become critical in daily life, as evidenced by strong consumer demand for the fastest information technologies, despite the extremely high cost incurred in terms of system usage. To satisfy this requirement, radio frequency technology is not as expensive as other complex ones. This work focuses on acquiring the fastest information by applying radio frequency identification (RFID) system that is integrated with a cloud computing system. Detecting numerous tags simultaneously always leads to the possibility of collisions occurring. In particular, a collision problem occurs for both a tag and a reader. The collision may result in either undetected tags ID or misinterpretation of a different code. This work proposes a novel RFID-based system that implementing the anticollision feature of a mobile reader involves solving the problem of collision. Performance results indicate that the Slotted ALOHA protocol for 1,000 tags requires nearly 2.7 sec, while the proposed method requires only around 1.4 sec. Overall comparison performance reveals that the proposed method decreases the reading time by nearly 50% over that of the Slotted ALOHA method.