How should we account for our ability to entertain simple, vision-based demonstrative thoughts about particular objects (that is, our ability to entertain thoughts about particular objects simply on the basis of seeing them)? In this thesis I propose an account of this ability that accords with the common-sense view that seeing an object puts one in a position to single it out by visually attending to it, and that this provides one with the ability to entertain demonstrative thoughts about it. An account of this type requires that we account for what it is to see a particular object and to visually attend to it without appealing to particular demonstrative abilities. However, it has been argued that a notion of seeing an object, and similarly a notion of attending to an object, which is accounted for in this way is unsuitable for accounting for demonstrative abilities. I argue that there is no real problem: what we need is a notion of experiential content which is concept-dependent only in a general manner. That is, the account of the relevant notion of experiential content requires appeal to the subject's conceptual abilities, but the account is not given in terms of specific conceptual abilities (especially, not specific demonstrative abilities). I then characterize a notion of attention to a seen object which can be accounted for without appeal to particular demonstrative abilities, and explain how attending to an object in the relevant sense provides the subject with the ability to think about the object demonstratively. It is widely agreed that spatial location plays a central role in an account of demonstratives. I explain this role in terms of the role played by location in visual attention to the object and the subject's grasp of the fact that he attends to the object.

Master of Science in Engineering - Electrical and Information Engineering / With the increase in complexity of modern software systems, there is a great demand for software engineering techniques. Calculation processes are becoming more and more complex, especially in the field of machine vision and computational intelligence. A suitable object oriented calculation process framework is developed in order to address this problem. To demonstrate the effectiveness of the framework, a simple shape classification system is implemented in C#. A suitable method for representing shapes of images is developed and it is used for classification by a neural network. Sets of real-world images of hands and automobiles are used to test the system. The performance of the object oriented system in C# is compared to a functional paradigm system in Matlab and it is found that object orientation is well suited to the later stages of machine vision while the functional approach is well suited to low level image processing tasks.

Real-time optical intensity correlation using a photorefractive BSO crystal and a liquid crystal television is implemented. The underlying physics basis is considered, some specific techniques to improve the operation are proposed, and several optical pattern recognition tasks are achieved. Photorefractive BSO is used as the holographic recording medium in the real-time intensity correlator. To improve the dynamic holographic recording, a moving grating technique is adopted. The nonlinear effects of moving gratings at large fringe modulation are experimentally investigated, and are compared with numerical predictions. Optical bias is adopted to overcome the difficulty of a large drop in the optimum fringe velocity with moving gratings. The effects of optical bias on the optimum fringe velocity and on the diffraction efficiency are studied. To overcome the inherent drawback of low discrimination of intensity correlation in optical pattern recognition, real-time edge-enhanced intensity correlation is achieved by means of nonlinear holographic recording in BSO. Real-time colour object recognition is achieved by using a commercially available and inexpensive colour liquid crystal television in the intensity correlator. Multi-class object recognition is achieved with a synthetic discriminant function filter displayed by the Epson liquid crystal display in the real-time intensity correlator. The phase and intensity modulation properties of the Epson liquid crystal display are studied. A further research topic which uses the Epson liquid crystal display to realize a newly designed spatial filter, the quantized amplitude-compensated matched filter, is proposed. The performance merits of the filter are investigated by means of computer simulations.

Object recognition and manipulation are critical for enabling robots to operate within a household environment. There are many grasp planners that can estimate grasps based on object shape, but these approaches often perform poorly because they miss key information about non-visual object characteristics, such as weight distribution, fragility of materials, and usability characteristics. Object model databases can account for this information, but existing methods for constructing 3D object recognition databases are time and resource intensive, often requiring specialized equipment, and are therefore difficult to apply to robots in the field. We present an easy-to-use system for constructing object models for 3D object recognition and manipulation made possible by advances in web robotics. The database consists of point clouds generated using a novel iterative point cloud registration algorithm, which includes the encoding of manipulation data and usability characteristics. The system requires no additional equipment other than the robot itself, and non-expert users can demonstrate grasps through an intuitive web interface with virtually no training required. We validate the system with data collected from both a crowdsourcing user study and a set of grasps demonstrated by an expert user. We show that the crowdsourced grasps can produce successful autonomous grasps, and furthermore the demonstration approach outperforms purely vision-based grasp planning approaches for a wide variety of object classes.

Chan, Chi Wai Alan. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 121-126). / Abstracts in English and Chinese. / Abstract --- p.v / Acknowledgements --- p.viii / List of Figures --- p.ix / List of Tables --- p.xi / Chapter 1. --- Introduction --- p.12 / Chapter 1.1. --- Object-oriented databases --- p.12 / Chapter 1.2. --- Object-oriented Data Model --- p.14 / Chapter 1.3. --- Class and Object Instances --- p.15 / Chapter 1.4. --- Inheritance --- p.16 / Chapter 1.5. --- Constraint --- p.18 / Chapter 1.6. --- Physical Design for OODB Storage --- p.19 / Chapter 1.7. --- Problem Description --- p.20 / Chapter 1.8. --- Genetic Algorithm --- p.22 / Chapter 1.8.1. --- Constraint Handling Methods in GA --- p.25 / Chapter 1.9. --- Contributions of this work --- p.27 / Chapter 1.10. --- Outline of this work --- p.30 / Chapter 2. --- Literature Review --- p.32 / Chapter 2.1. --- Object-oriented database --- p.32 / Chapter 2.2. --- Object-Oriented Data model --- p.33 / Chapter 2.3. --- Physical Storage Model for OODBs --- p.35 / Chapter 2.3.1. --- Home Class (HC) Model --- p.36 / Chapter 2.3.2. --- Repeated Class (RC) Model --- p.38 / Chapter 2.3.3. --- Split Instance (SI) Model --- p.39 / Chapter 2.4. --- Solving physical storage design for OODBs --- p.40 / Chapter 2.5. --- Transaction-Based Approach --- p.41 / Chapter 2.6. --- Minimize database operational cost --- p.42 / Chapter 2.7. --- Combinational Optimization Method --- p.43 / Chapter 2.8. --- Research in Genetic Algorithm --- p.46 / Chapter 2.9. --- Implementation in GA --- p.47 / Chapter 2.10. --- Fitness function --- p.49 / Chapter 2.11. --- Crossover operation --- p.50 / Chapter 2.12. --- Encoding and Representation --- p.51 / Chapter 2.13. --- Parent Selection in Crossover Operation --- p.52 / Chapter 2.14. --- Reproductive selection --- p.53 / Chapter 2.14.1. --- Selection of Crossover Operator --- p.54 / Chapter 2.14.2. --- Replacement --- p.54 / Chapter 2.15. --- The Use of Constraint Handling Method --- p.55 / Chapter 2.15.1. --- Penalty function --- p.56 / Chapter 2.15.2. --- Decoder gives instruction to build feasible solution --- p.57 / Chapter 2.15.3. --- Adjustment method --- p.58 / Chapter 3. --- Solving Physical Storage Problem for OODB using GA --- p.60 / Chapter 3.1. --- Physical storage models for OODB --- p.61 / Chapter 3.2. --- Database operation for transactions --- p.62 / Chapter 3.3. --- Properly designed physical storage structure --- p.68 / Chapter 3.4. --- Fitness Evaluation --- p.69 / Chapter 3.5. --- Initial population --- p.72 / Chapter 3.6. --- Cross-breeding --- p.72 / Chapter 3.7. --- GA Operators --- p.74 / Chapter 3.8. --- Physical Design Problem Formulation for GA --- p.75 / Chapter 3.9. --- Representation and Encoding --- p.75 / Chapter 3.10. --- Solving Physical Storage Problem for OODB in GA --- p.76 / Chapter 3.10.1. --- Representation of design solution --- p.76 / Chapter 3.10.2. --- Encoding --- p.78 / Chapter 3.10.3. --- Initial population --- p.80 / Chapter 3.10.4. --- Parent Selection for breeding --- p.80 / Chapter 3.11. --- Traditional Constraint handling method --- p.83 / Chapter 3.11.1. --- Improve the Performance of Inheritance Constraint Handling methods --- p.85 / Chapter 3.12. --- Weakness in Gorla's GA approach --- p.87 / Chapter 4. --- Proposed Methodology --- p.88 / Chapter 4.1 --- Enhanced Crossover Operator --- p.90 / Chapter 4.2. --- Infeasible Solutions and Enhanced Adjustment Method --- p.93 / Chapter 4.3. --- Propagation Adjustment Method --- p.97 / Chapter 5. --- Computational Experiments --- p.99 / Chapter 5.1. --- Introduction --- p.99 / Chapter 5.2. --- Experiment Objective --- p.101 / Chapter 5.3. --- Tools and Setup --- p.102 / Chapter 5.4. --- Crossover Operator --- p.105 / Chapter 5.5. --- Mutation Operator --- p.105 / Chapter 5.6. --- Termination condition --- p.106 / Chapter 5.7. --- Computational Experiments --- p.107 / Chapter 5.7.1. --- An Illustrative Example ´ؤ UNIVERSITY database --- p.107 / Chapter 5.7.2. --- Simulation ´ؤ 9 classes and 25 classes --- p.115 / Chapter 5.7.3. --- Result --- p.116 / Chapter 6. --- Conclusions --- p.118 / Chapter 6.1. --- Summary of Achievements --- p.118 / Chapter 7. --- Bibliography --- p.121 / Chapter 8. --- Appendix --- p.127

The aim of this project is to enable robots to recognise objects and object categories by combining vision and touch. In this thesis, a novel inexpensive tactile sensor design is presented, together with a complete, probabilistic sensor-fusion model. The potential of the model is demonstrated in four areas: (i) Shape Recognition, here the sensor outperforms its most similar rival, (ii) Single-touch Object Recognition, where state-of-the-art results are produced, (iii) Visuo-tactile object recognition, demonstrating the benefits of multi-sensory object representations, and (iv) Object Classification, which has not been reported in the literature to date. Both the sensor design and the novel database were made available. Tactile data collection is performed by a robot. An extensive analysis of data encodings, data processing, and classification methods is presented. The conclusions reached are: (i) the inexpensive tactile sensor can be used for basic shape and object recognition, (ii) object recognition combining vision and touch in a probabilistic manner provides an improvement in accuracy over either modality alone, (iii) when both vision and touch perform poorly independently, the sensor-fusion model proposed provides faster learning, i.e. fewer training samples are required to achieve similar accuracy, and (iv) such a sensor-fusion model is more accurate than either modality alone when attempting to classify unseen objects, as well as when attempting to recognise individual objects from amongst similar other objects of the same class. (v) The preliminary potential is identified for real-life applications: underwater object classification. (vi) The sensor fusion model providesimprovements in classification even for award-winning deep-learning basedcomputer vision models.

Object classes are central to computer vision and have been the focus of substantial research in the last fifteen years. This thesis addresses the tasks of localizing entire objects in images (object class detection) and localizing their semantic parts (part detection). We present four contributions, two for each task. The first two improve existing object class detection techniques by using context and calibration. The other two contributions explore semantic part detection in weakly-supervised settings. First, the thesis presents a technique for predicting properties of objects in an image based on its global appearance only. We demonstrate the method by predicting three properties: aspect of appearance, location in the image and class membership. Overall, the technique makes multi-component object detectors faster and improves their performance. The second contribution is a method for calibrating the popular Ensemble of Exemplar- SVM object detector. Unlike the standard approach, which calibrates each Exemplar- SVM independently, our technique optimizes their joint performance as an ensemble. We devise an efficient optimization algorithm to find the global optimal solution of the calibration problem. This leads to better object detection performance compared to using independent calibration. The third innovation is a technique to train part-based model of object classes using data sourced from the web. We learn rich models incrementally. Our models encompass the appearance of parts and their spatial arrangement on the object, specific to each viewpoint. Importantly, it does not require any part location annotation, which is one of the main limits to training many part detectors. Finally, the last contribution is a study on whether semantic object parts emerge in Convolutional Neural Networks trained for higher-level tasks, such as image classification. While previous efforts studied this matter by visual inspection only, we perform an extensive quantitative analysis based on ground-truth part location annotations. This provides a more conclusive answer to the question.

The development of orienting object features during hand to mouth transport was studied in a group of infants between 6-12 months of age (N=60). The task was to bring a bottle to the mouth. The bottle was presented in six different orientations (up, down, toward, away, left, and right). To measure head movement, a 3D motion capture system (Qualisys) was used. The results revealed that older but not younger infants are able to plan actions based on the functional end of an object. Older infants take less time, are more successful at directing the nipple to the mouth, initially grasp the bottle in an efficient manner more frequently, and move the head less than younger infants. More broadly, the results offer insights into how infants become more efficient in feeding tasks as they get older. / 1 / Wendy P. Jung

This dissertation presents a new approach to image interpretation which can produce hierarchical descriptions of visually sensed scenes based on an incrementally learnt hierarchical knowledge base. Multiple segmentation and labelling hypotheses are generated with local constraint satisfaction being achieved through a hierarchical form of relaxation labelling. The traditionally unidirectional segmentation-matching process is recast into a dynamic closed-loop system where the current interpretation state is used to drive the lower level image processing functions. The theory presented in this dissertation is applied to a new object recognition and scene understanding system called Cite which is described in detail.

Static analysis is significant since it provides the information about the run- time behaviour of an analysed program. Such information has many applications in compiler optimisations and software engineering tools. Interprocedural anal- ysis is a form of static analysis, which can exploit information available across procedure boundaries. The analysis is traditionally designed as whole-program analysis, which processes the entire program. However, whole-program analysis is problematic when parts of the analysed program are not available to partici- pate in analysis. In this case, a whole-program analysis has to make conservative assumptions to be able to produce safe analysis results at the expense of some possible precision loss. To improve analysis precision, an analysis can exploit the access control mechanism provided by the underlying program language. This thesis introduces a points-to analysis technique for incomplete object-oriented programs, called com- pleteness analysis, which exploits the access and modification properties of classes, methods and fields to enhance the analysis precision. Two variations of the tech- nique, compositional and sequential completeness analysis, are described. This thesis also presents a mutability analysis (MA) and MA-based side-effect analy- sis, which are based on the output of completeness analysis, to determine whether a variable is potentially modified by the execution of a program statement. The results of experiments carried out on a set of Java library packages are presented to demonstrate the improvement in analysis precision.