Articulated character animation can be performed by manually creating and rigging a skeleton into an unfolded 3D mesh model. Such tasks are not trivial, as they require a substantial amount of training and practice. Although methods have been proposed to help automatic extraction of skeleton structure, they may not guarantee that the resulting skeleton can help to produce animations according to user manipulation. We present a sketching-based skeleton extraction method to create a user desired skeleton structure which is used in 3D model animation. This method takes user sketching as an input, and based on the mesh segmentation result of a 3D mesh model, generates a skeleton for articulated character animation. In our system, we assume that a user will properly sketch bones by roughly following the mesh model structure. The user is expected to sketch independently on different regions of a mesh model for creating separate bones. For each sketched stroke, we project it into the mesh model so that it becomes the medial axis of its corresponding mesh model region from the current viewer perspective. We call this projected stroke a “sketched bone”. After pre-processing user sketched bones, we cluster them into groups. This process is critical as user sketching can be done from any orientation of a mesh model. To specify the topology feature for different mesh parts, a user can sketch strokes from different orientations of a mesh model, as there may be duplicate strokes from different orientations for the same mesh part. We need a clustering process to merge similar sketched bones into one bone, which we call a “reference bone”. The clustering process is based on three criteria: orientation, overlapping and locality. Given the reference bones as the input, we adopt a mesh segmentation process to assist our skeleton extraction method. To be specific, we apply the reference bones and the seed triangles to segment the input mesh model into meaningful segments using a multiple-region growing mechanism. The seed triangles, which are collected from the reference bones, are used as the initial seeds in the mesh segmentation process. We have designed a new segmentation metric [1] to form a better segmentation criterion. Then we compute the Level Set Diagrams (LSDs) on each mesh part to extract bones and joints. To construct the final skeleton, we connect bones extracted from all mesh parts together into a single structure. There are three major steps involved: optimizing and smoothing bones, generating joints and forming the skeleton structure. After constructing the skeleton model, we have proposed a new method, which utilizes the Linear Blend Skinning (LBS) technique and the Laplacian mesh deformation technique together to perform skeleton-driven animation. Traditional LBS techniques may have self-intersection problem in regions around segmentation boundaries. Laplacian mesh deformation can preserve the local surface details, which can eliminate the self-intersection problem. In this case, we make use of LBS result as the positional constraint to perform a Laplacian mesh deformation. By using the Laplacian mesh deformation method, we maintain the surface details in segmentation boundary regions. This thesis outlines a novel approach to construct a 3D skeleton model interactively, which can also be used in 3D animation and 3D model matching area. The work is motivated by the observation that either most of the existing automatic skeleton extraction methods lack well-positioned joints specification or the manually generated methods require too much professional training to create a good skeleton structure. We dedicate a novel approach to create 3D model skeleton based on user sketching which specifies articulated skeleton with joints. The experimental results show that our method can produce better skeletons in terms of joint positions and topological structure.

This thesis explores more efficient methods for visualizing point data sets on three-dimensional (3D) displays. Point data sets are used in many scientific applications, e.g. cosmological simulations. Visualizing these data sets in {3D} is desirable because it can more readily reveal structure and unknown phenomena. However, cutting-edge scientific point data sets are very large and producing/rendering even a single image is expensive. Furthermore, current literature suggests that the ideal number of views for 3D (multiview) displays can be in the hundreds, which compounds the costs. The accepted notion that many views are required for {3D} displays is challenged by carrying out a novel human factor trials study. The results suggest that humans are actually surprisingly insensitive to the number of viewpoints with regard to their task performance, when occlusion in the scene is not a dominant factor. Existing stereoscopic rendering algorithms can have high set-up costs which limits their use and none are tuned for uncorrelated {3D} point rendering. This thesis shows that it is possible to improve rendering speeds for a low number of views by perspective reprojection. The novelty in the approach described lies in delaying the reprojection and generation of the viewpoints until the fragment stage of the pipeline and streamlining the rendering pipeline for points only. Theoretical analysis suggests a fragment reprojection scheme will render at least 2.8 times faster than na\"{i}vely re-rendering the scene from multiple viewpoints. Building upon the fragment reprojection technique, further rendering performance is shown to be possible (at the cost of some rendering accuracy) by restricting the amount of reprojection required according to the stereoscopic resolution of the display. A significant benefit is that the scene depth can be mapped arbitrarily to the perceived depth range of the display at no extra cost than a single region mapping approach. Using an average case-study (rendering from a 500k points for a 9-view High Definition 3D display), theoretical analysis suggests that this new approach is capable of twice the performance gains than simply reprojecting every single fragment, and quantitative measures show the algorithm to be 5 times faster than a naïve rendering approach. Further detailed quantitative results, under varying scenarios, are provided and discussed.

Systems made of many interacting elements may display unanticipated emergent properties. A system for which the desired properties are the same as those which emerge will be inherently robust. Currently available techniques for designing emergent properties are prohibitively costly for all but the simplest systems. The self-assembly of biological cells into tissues and ultimately organisms is an example of a natural dynamic distributed system of which the primary emergent behaviour is a fully operational being. The distributed process that co-ordinates this self-assembly is morphogenesis. By analysing morphogenesis with a cellular automata model we deduce a means by which this self-organisation might be achieved. This mechanism is then adapted to the design of self-organising patterns, reliable electronic systems and self-assembling systems. The limitations of the design algorithm are analysed, as is a means to overcome them. The cost of this algorithm is discussed and finally demonstrated with the design of a reliable arithmetic logic unit and a self-assembling, self-repairing and metamorphosising robot made of 12,000 cells.

Metrics for automatically predicting the compression settings for stereoscopic images, to minimize file size, while still maintaining an acceptable level of image quality are investigated. This research evaluates whether symmetric or asymmetric compression produces a better quality of stereoscopic image. Initially, how Peak Signal to Noise Ratio (PSNR) measures the quality of varyingly compressed stereoscopic image pairs was investigated. Two trials with human subjects, following the ITU-R BT.500-11 Double Stimulus Continuous Quality Scale (DSCQS) were undertaken to measure the quality of symmetric and asymmetric stereoscopic image compression. Computational models of the Human Visual System (HVS) were then investigated and a new stereoscopic image quality metric designed and implemented. The metric point matches regions of high spatial frequency between the left and right views of the stereo pair and accounts for HVS sensitivity to contrast and luminance changes in these regions. The PSNR results show that symmetric, as opposed to asymmetric stereo image compression, produces significantly better results. The human factors trial suggested that in general, symmetric compression of stereoscopic images should be used. The new metric, Stereo Band Limited Contrast, has been demonstrated as a better predictor of human image quality preference than PSNR and can be used to predict a perceptual threshold level for stereoscopic image compression. The threshold is the maximum compression that can be applied without the perceived image quality being altered. Overall, it is concluded that, symmetric, as opposed to asymmetric stereo image encoding, should be used for stereoscopic image compression. As PSNR measures of image quality are correctly criticized for correlating poorly with perceived visual quality, the new HVS based metric was developed. This metric produces a useful threshold to provide a practical starting point to decide the level of compression to use.

In this thesis, a construction kit for the implementation of neuronal networks on Field Programmable Gate Arrays (FPGA) is presented. The utility of this technology for the implementation of neuronal networks becomes apparent when we show that is possible to perform hyper-real time operation, a feature which allows neuronal networks designers to analyse in great detail the dynamics of their networks through the means of a comprehensive behavioural analysis. Additionally, the construction kit presented here is used to implement a biologically inspired version of the mammalian olfactory bulb based on previous work by T. C. Pearce et al. (2005). The results we obtain show that tasks such as identification, classification and segmentation of odours are successfully performed by the olfactory bulb model. We illustrate the practical utility of the model by including experiments using real odour information. Finally, a comprehensive behavioural analysis is performed by means of a massive exploration of the characteristics of the response of the model as it was subjected to different conditions of certain parameters. Results show that as the stimulating input approaches the trained input (target odour), the response tends to reach an attractor. This comprehensive analysis, made it possible to observe the effects of the training on the model response.

No description available.

Research in biologically-inspired optimisation has been fl<;lurishing over the past decades. This approach adopts a bott0!ll-up viewpoint to understand and mimic certain features of a biological system. It has been proved useful in developing nondeterministic algorithms, such as Evolutionary Algorithms (EAs) and Swarm Intelligence (SI). Bacteria, as the simplest creature in nature, are of particular interest in recent studies. In the past thousands of millions of years, bacteria have exhibited a self-organising behaviour to cope with the natural selection. For example, bacteria have developed a number of strategies to search for food sources with a very efficient manner. This thesis explores the potential of understanding of a biological system by modelling the' underlying mechanisms of bacterial foraging patterns and investigates their applicability to engineering optimisation problems. :rvlodelling plays a significant role in understanding bacterial foraging behaviour. Mathematical expressions and experimental observations have been utilised to represent biological systems. However, difficulties arise from the lack of systematic analysis of the developed models and experimental data. Recently, Systems Biology has be,en proposed to overcome this barrier, with the effort from a number of research fields, including Computer Science and Systems Engineering. At the same time, Individual-based Modelling (IbM) has emerged to assist the modelling of a biological system. Starting from a basic model of foraging and proliferation of bacteria, the development of an IbM of bacterial systems of this thesis focuses on a Varying Environment BActerial Model (VEBAM). Simulation results demonstrate that VEBAM is able to provide a new perspective to describe interactions between the bacteria and their food environment. Knowledge transfer from modelling of bacterial systems to solving optimisation problems also composes an important part of this study. Three Bacteriainspired Algorithms (BalAs) have been developed to bridge the gap between modelling and optimisation. These algorithms make use of the. self-adaptability of individual bacteria in the group searching activities described in VEBAM, while incorporating a variety of additional features. In particular, the new bacterial foraging algorithm with varying population (BFAVP) takes bacterial metabolism into consideration. The group behaviour in Particle Swarm Optimiser (PSO) is adopted in Bacterial Swarming Algorithm (BSA) to enhance searching ability. To reduce computational time, another algorithm, a Paired-bacteria Optimiser (PBO) is designed specifically to further explore the capability of BalAs. Simulation studies undertaken against a wide range of benchmark functions demonstrate a satisfying performance with a reasonable convergence speed. To explore the potential of bacterial searching ability in optimisation undertaken in a varying environment, a dynamic bacterial foraging algorithm (DBFA) is developed with the aim of solving optimisation in a time-varying environment. In this case, the balance between its convergence and exploration abilities is investigated, and a new scheme of reproduction is developed which is different froin that used for static optimisation problems. The simulation studies have been undertaken and the results show that the DBFA can adapt to various environmental changes rapidly. One of the challenging large-scale complex optimisation problems is optimal power flow (OPF) computation. BFAVP shows its advantage in solving this problem. A simulation study has been performed on an IEEE 30-bus system, and the results are compared with PSO algorithm and Fast Evolutionary Programming (FEP) algorithm, respectively. Furthermore, the OPF problem is extended for consideration in varying environments, on which DBFA has been evaluated. A simulation study has been undertaken on both the IEEE 30-bus system and the IEEE l1S-bus system, in compariso~ with a number of existing algorithms. The dynamic OPF problem has been tackled for the first time in the area of power systems, and the results obtained are encouraging, with a significant amount of energy could possibly being saved. Another application of BaIA in this thesis is concerned with estimating optimal parameters of a power transformer winding model using BSA. Compared with Genetic Algorithm (GA), BSA is able to obtain a more satisfying result in modelling the transformer winding, which could not be achieved using a theoretical transfer function model.

The main objective in automatic robot controller development is to devise mechanisms whereby robot controllers can be developed with less reliance on human developers. One such mechanism is the use of evolutionary algorithms (EAs) to automatically develop robot controllers and occasionally, robot morphology. This area of research is referred to as evolutionary robotics (ER). Through the use of evolutionary techniques such as genetic algorithms (GAs) and genetic programming (GP), ER has shown to be a promising approach through which robust robot controllers can be developed. The standard ER techniques use monolithic evolution to evolve robot behaviour: monolithic evolution involves the use of one chromosome to code for an entire target behaviour. In complex problems, monolithic evolution has been shown to suffer from bootstrap problems; that is, a lack of improvement in fitness due to randomness in the solution set [103, 105, 100, 90]. Thus, approaches to dividing the tasks, such that the main behaviours emerge from the interaction of these simple tasks with the robot environment have been devised. These techniques include the subsumption architecture in behaviour based robotics, incremental learning and more recently the layered learning approach [55, 103, 56, 105, 136, 95]. These new techniques enable ER to develop complex controllers for autonomous robot. Work presented in this thesis extends the field of evolutionary robotics by introducing Gene Expression Programming (GEP) to the ER field. GEP is a newly developed evolutionary algorithm akin to GA and GP, which has shown great promise in optimisation problems. The presented research shows through experimentation that the unique formulation of GEP genes is sufficient for robot controller representation and development. The obtained results show that GEP is a plausible technique for ER problems. Additionally, it is shown that controllers evolved using GEP algorithm are able to adapt when introduced to new environments. Further, the capabilities of GEP chromosomes to code for more than one gene have been utilised to show that GEP can be used to evolve manually sub-divided robot behaviours. Additionally, this thesis extends the GEP algorithm by proposing two new evolutionary techniques named multigenic GEP with Linker Evolution (mgGEP-LE) and multigenic GEP with a Regulator Gene (mgGEP-RG). The results obtained from the proposed algorithms show that the new techniques can be used to automatically evolve modularity in robot behaviour. This ability to automate the process of behaviour sub-division and optimisation in a modular chromosome is unique to the GEP formulations discussed, and is an important advance in the development of machines that are able to evolve stratified behavioural architectures with little human intervention.

No description available.

No description available.