No description available.

In this thesis we present a method for detecting planar targets in real-time on power-constrained, or low-powered, hand-held devices such as mobile phones. We adopt the feature recognition (also referred to as feature matching) approach and employ fast-to-compute local feature descriptors to establish point correspondences. To obtain a satisfactory localization accuracy, most local feature descriptors seek a transformation of the input intensity patch that is invariant to various geometric and photometric deformations. Generally, such transformations are computationally intensive, hence are not ideal for real-time applications on limited hardware platforms. On the other hand, descriptors which are fast to compute are typically limited in their ability to provide invariance to a vast range of deformations. To address these shortcomings, we have developed a learning-based approach which can be applied to any local feature descriptor to increase the system’s robustness to both affine and perspective deformations. The motivation behind applying a learning-based approach is to transfer as much of the computational burden (as possible) onto an offline training phase, allowing a reduction in cost during online matching. The approach comprises of identifying keypoints which remain stable under artificially induced perspective transformations, extracting the corresponding feature vectors, and finally aggregating the feature vectors of coincident keypoints to obtain the final descriptors. We strictly focus on objects which are planar, thus allowing us to synthesize images of the object in order to capture the appearance of keypoint patches under several perspectives.

In this thesis we present a method for detecting planar targets in real-time on power-constrained, or low-powered, hand-held devices such as mobile phones. We adopt the feature recognition (also referred to as feature matching) approach and employ fast-to-compute local feature descriptors to establish point correspondences. To obtain a satisfactory localization accuracy, most local feature descriptors seek a transformation of the input intensity patch that is invariant to various geometric and photometric deformations. Generally, such transformations are computationally intensive, hence are not ideal for real-time applications on limited hardware platforms. On the other hand, descriptors which are fast to compute are typically limited in their ability to provide invariance to a vast range of deformations. To address these shortcomings, we have developed a learning-based approach which can be applied to any local feature descriptor to increase the system’s robustness to both affine and perspective deformations. The motivation behind applying a learning-based approach is to transfer as much of the computational burden (as possible) onto an offline training phase, allowing a reduction in cost during online matching. The approach comprises of identifying keypoints which remain stable under artificially induced perspective transformations, extracting the corresponding feature vectors, and finally aggregating the feature vectors of coincident keypoints to obtain the final descriptors. We strictly focus on objects which are planar, thus allowing us to synthesize images of the object in order to capture the appearance of keypoint patches under several perspectives.

No description available.

No description available.

We discuss a variety of object recognition experiments in which human subjects were presented with realistically rendered images of computer-generated three-dimensional objects, with tight control over stimulus shape, surface properties, illumination, and viewpoint, as well as subjects' prior exposure to the stimulus objects. In all experiments recognition performance was: (1) consistently viewpoint dependent; (2) only partially aided by binocular stereo and other depth information, (3) specific to viewpoints that were familiar; (4) systematically disrupted by rotation in depth more than by deforming the two-dimensional images of the stimuli. These results are consistent with recently advanced computational theories of recognition based on view interpolation.

Injection molding is widely used to manufacture plastic parts with good surface finish, dimensional stability and low cost. The common examples of parts manufactured by injection molding include toys, utensils, and casings of various electronic products. The process of mold design to generate these complex shapes is iterative and time consuming, and requires great expertise in the field. As a result, a significant amount of the final product cost can be attributed to the expenses incurred during the product’s design. After designing the mold segments, it is necessary to machine these segments with minimum cost using an efficient tool-path. The tool-path planning process also adds to the overall mold cost. The process of injection molding can be simplified and made to be more cost effective if the processes of mold design and tool-path generation can be automated. This work focuses on the automation of mold design from a given part design and the automation of tool-path generation for manufacturing mold segments. The hypothesis examined in this thesis is that the automatic identification of mold features can reduce the human efforts required to design molds. It is further hypothesised that the human effort required in many downstream processes such as mold component machining can also be reduced with algorithmic automation of otherwise time consuming decisions. Automatic design of dies and molds begins with the part design being provided as a solid model. The solid model of a part is a database of its geometry and topology. The automatic mold design process uses this database to identify an undercut-free parting direction, for recognition of mold features and identification of parting lines for a given parting direction, and for generation of entities such as parting surfaces, core, cavity and side-cores. The methods presented in this work are analytical in nature and work with the extended set of part topologies and geometries unlike those found in the literature. Moreover, the methods do not require discretizing the part geometry to design its mold segments, unlike those found in the literature that result in losing the part definition. Once the mold features are recognized and parting lines are defined, core, cavity and side-cores are generated. This work presents algorithms that recognize the entities in the part solid model that contribute to the design of the core, cavity and side-cores, extract the entities, and use them in the design of these elements. The developed algorithms are demonstrated on a variety of parts that cover a wide range of features. The work also presents a method for automatic tool-path generation that takes the designed core/cavity and produces a multi-stage tool-path to machine it from raw stock. The tool-path generation process begins by determining tool-path profiles and tool positions for the rough machining of the part in layers. Typically roughing is done with large aggressive tools to reduce the machining time; and roughing leaves uncut material. After generating a roughing tool-path for each layer, the machining is simulated and the areas left uncut are identified to generate a clean-up tool-path for smaller sized tools. The tool-path planning is demonstrated using a part having obstacles within the machining region. The simulated machining is presented in this work. This work extends the accessibility analysis by retaining the topology information and using it to recognize a larger domain of features including intersecting features, filling a void in the literature regarding a method that could recognize complex intersecting features during an automated mold design process. Using this information, a larger variety of new mold intersecting features are classified and recognized in this approach. The second major contribution of the work was to demonstrate that the downstream operations can also benefit from algorithmic decision making. This is shown by automatically generating roughing and clean-up tool-paths, while reducing the machining time by machining only those areas that have uncut material. The algorithm can handle cavities with obstacles in them. The methodology has been tested on a number of parts.

No description available.

Injection molding is widely used to manufacture plastic parts with good surface finish, dimensional stability and low cost. The common examples of parts manufactured by injection molding include toys, utensils, and casings of various electronic products. The process of mold design to generate these complex shapes is iterative and time consuming, and requires great expertise in the field. As a result, a significant amount of the final product cost can be attributed to the expenses incurred during the product’s design. After designing the mold segments, it is necessary to machine these segments with minimum cost using an efficient tool-path. The tool-path planning process also adds to the overall mold cost. The process of injection molding can be simplified and made to be more cost effective if the processes of mold design and tool-path generation can be automated. This work focuses on the automation of mold design from a given part design and the automation of tool-path generation for manufacturing mold segments. The hypothesis examined in this thesis is that the automatic identification of mold features can reduce the human efforts required to design molds. It is further hypothesised that the human effort required in many downstream processes such as mold component machining can also be reduced with algorithmic automation of otherwise time consuming decisions. Automatic design of dies and molds begins with the part design being provided as a solid model. The solid model of a part is a database of its geometry and topology. The automatic mold design process uses this database to identify an undercut-free parting direction, for recognition of mold features and identification of parting lines for a given parting direction, and for generation of entities such as parting surfaces, core, cavity and side-cores. The methods presented in this work are analytical in nature and work with the extended set of part topologies and geometries unlike those found in the literature. Moreover, the methods do not require discretizing the part geometry to design its mold segments, unlike those found in the literature that result in losing the part definition. Once the mold features are recognized and parting lines are defined, core, cavity and side-cores are generated. This work presents algorithms that recognize the entities in the part solid model that contribute to the design of the core, cavity and side-cores, extract the entities, and use them in the design of these elements. The developed algorithms are demonstrated on a variety of parts that cover a wide range of features. The work also presents a method for automatic tool-path generation that takes the designed core/cavity and produces a multi-stage tool-path to machine it from raw stock. The tool-path generation process begins by determining tool-path profiles and tool positions for the rough machining of the part in layers. Typically roughing is done with large aggressive tools to reduce the machining time; and roughing leaves uncut material. After generating a roughing tool-path for each layer, the machining is simulated and the areas left uncut are identified to generate a clean-up tool-path for smaller sized tools. The tool-path planning is demonstrated using a part having obstacles within the machining region. The simulated machining is presented in this work. This work extends the accessibility analysis by retaining the topology information and using it to recognize a larger domain of features including intersecting features, filling a void in the literature regarding a method that could recognize complex intersecting features during an automated mold design process. Using this information, a larger variety of new mold intersecting features are classified and recognized in this approach. The second major contribution of the work was to demonstrate that the downstream operations can also benefit from algorithmic decision making. This is shown by automatically generating roughing and clean-up tool-paths, while reducing the machining time by machining only those areas that have uncut material. The algorithm can handle cavities with obstacles in them. The methodology has been tested on a number of parts.

Categorization of predators is fundamental for birds? life. However, the mechanism is not well explained yet. Two basic theories of categorization were tested in the laboratory conditions, where we have the possibility to focus on the behavior of each individuum. In the experiments we focused on the relevance of the key features and colour patterns. The responses of the Great Tit to variously modified dummies of Sparrowhawk were observed. The results confirm that the bird does not recognize the predator only on the basis of the key features and the colour patterns have a essential influence on the categorization. Categorization of the predator is a complex process and it cannot be explained only on the basis of one psychological theory. A combination of the key features and a prototype of a particular category learned throughout its life is probably used.