8 resultados para 3D vision
em CaltechTHESIS
Resumo:
This thesis presents a novel framework for state estimation in the context of robotic grasping and manipulation. The overall estimation approach is based on fusing various visual cues for manipulator tracking, namely appearance and feature-based, shape-based, and silhouette-based visual cues. Similarly, a framework is developed to fuse the above visual cues, but also kinesthetic cues such as force-torque and tactile measurements, for in-hand object pose estimation. The cues are extracted from multiple sensor modalities and are fused in a variety of Kalman filters.
A hybrid estimator is developed to estimate both a continuous state (robot and object states) and discrete states, called contact modes, which specify how each finger contacts a particular object surface. A static multiple model estimator is used to compute and maintain this mode probability. The thesis also develops an estimation framework for estimating model parameters associated with object grasping. Dual and joint state-parameter estimation is explored for parameter estimation of a grasped object's mass and center of mass. Experimental results demonstrate simultaneous object localization and center of mass estimation.
Dual-arm estimation is developed for two arm robotic manipulation tasks. Two types of filters are explored; the first is an augmented filter that contains both arms in the state vector while the second runs two filters in parallel, one for each arm. These two frameworks and their performance is compared in a dual-arm task of removing a wheel from a hub.
This thesis also presents a new method for action selection involving touch. This next best touch method selects an available action for interacting with an object that will gain the most information. The algorithm employs information theory to compute an information gain metric that is based on a probabilistic belief suitable for the task. An estimation framework is used to maintain this belief over time. Kinesthetic measurements such as contact and tactile measurements are used to update the state belief after every interactive action. Simulation and experimental results are demonstrated using next best touch for object localization, specifically a door handle on a door. The next best touch theory is extended for model parameter determination. Since many objects within a particular object category share the same rough shape, principle component analysis may be used to parametrize the object mesh models. These parameters can be estimated using the action selection technique that selects the touching action which best both localizes and estimates these parameters. Simulation results are then presented involving localizing and determining a parameter of a screwdriver.
Lastly, the next best touch theory is further extended to model classes. Instead of estimating parameters, object class determination is incorporated into the information gain metric calculation. The best touching action is selected in order to best discern between the possible model classes. Simulation results are presented to validate the theory.
Resumo:
This thesis is concerned with spatial filtering. What is its utility in tone reproduction? Does it exist in vision, and if so, what constraints does it impose on the nervous system?
Tone reproduction is just the art and science of taking a picture and then displaying it. The sensors available to capture an image have a greater dynamic range than the media that may be used to display it. Conventionally, spatial filtering is used to boost contrast; it ameliorates the loss of contrast that results when the sensor signal range is scaled down to fit the display range. In this thesis, a type of nonlinear spatial filtering is discussed that results in direct range reduction without range scaling. This filtering process is instantiated in a real-time image processor built using analog CMOS VLSI.
Spatial filtering must be applied with care in both artificial and natural vision systems. It is argued that the nervous system does not simply filter linearly across an image. Rather, the way that we see things implies that the nervous system filters nonlinearly. Further, many models for color vision include a high-pass filtering step in which the DC information is lost. A real-time study of filtering in color space leads to the conclusion that the nervous system is not that simple, and that it maintains DC information by referencing to white.
Resumo:
Waking up from a dreamless sleep, I open my eyes, recognize my wife’s face and am filled with joy. In this thesis, I used functional Magnetic Resonance Imaging (fMRI) to gain insights into the mechanisms involved in this seemingly simple daily occurrence, which poses at least three great challenges to neuroscience: how does conscious experience arise from the activity of the brain? How does the brain process visual input to the point of recognizing individual faces? How does the brain store semantic knowledge about people that we know? To start tackling the first question, I studied the neural correlates of unconscious processing of invisible faces. I was unable to image significant activations related to the processing of completely invisible faces, despite existing reports in the literature. I thus moved on to the next question and studied how recognition of a familiar person was achieved in the brain; I focused on finding invariant representations of person identity – representations that would be activated any time we think of a familiar person, read their name, see their picture, hear them talk, etc. There again, I could not find significant evidence for such representations with fMRI, even in regions where they had previously been found with single unit recordings in human patients (the Jennifer Aniston neurons). Faced with these null outcomes, the scope of my investigations eventually turned back towards the technique that I had been using, fMRI, and the recently praised analytical tools that I had been trusting, Multivariate Pattern Analysis. After a mostly disappointing attempt at replicating a strong single unit finding of a categorical response to animals in the right human amygdala with fMRI, I put fMRI decoding to an ultimate test with a unique dataset acquired in the macaque monkey. There I showed a dissociation between the ability of fMRI to pick up face viewpoint information and its inability to pick up face identity information, which I mostly traced back to the poor clustering of identity selective units. Though fMRI decoding is a powerful new analytical tool, it does not rid fMRI of its inherent limitations as a hemodynamics-based measure.
Resumo:
This thesis addresses a series of topics related to the question of how people find the foreground objects from complex scenes. With both computer vision modeling, as well as psychophysical analyses, we explore the computational principles for low- and mid-level vision.
We first explore the computational methods of generating saliency maps from images and image sequences. We propose an extremely fast algorithm called Image Signature that detects the locations in the image that attract human eye gazes. With a series of experimental validations based on human behavioral data collected from various psychophysical experiments, we conclude that the Image Signature and its spatial-temporal extension, the Phase Discrepancy, are among the most accurate algorithms for saliency detection under various conditions.
In the second part, we bridge the gap between fixation prediction and salient object segmentation with two efforts. First, we propose a new dataset that contains both fixation and object segmentation information. By simultaneously presenting the two types of human data in the same dataset, we are able to analyze their intrinsic connection, as well as understanding the drawbacks of today’s “standard” but inappropriately labeled salient object segmentation dataset. Second, we also propose an algorithm of salient object segmentation. Based on our novel discoveries on the connections of fixation data and salient object segmentation data, our model significantly outperforms all existing models on all 3 datasets with large margins.
In the third part of the thesis, we discuss topics around the human factors of boundary analysis. Closely related to salient object segmentation, boundary analysis focuses on delimiting the local contours of an object. We identify the potential pitfalls of algorithm evaluation for the problem of boundary detection. Our analysis indicates that today’s popular boundary detection datasets contain significant level of noise, which may severely influence the benchmarking results. To give further insights on the labeling process, we propose a model to characterize the principles of the human factors during the labeling process.
The analyses reported in this thesis offer new perspectives to a series of interrelating issues in low- and mid-level vision. It gives warning signs to some of today’s “standard” procedures, while proposing new directions to encourage future research.
Resumo:
The visual system is a remarkable platform that evolved to solve difficult computational problems such as detection, recognition, and classification of objects. Of great interest is the face-processing network, a sub-system buried deep in the temporal lobe, dedicated for analyzing specific type of objects (faces). In this thesis, I focus on the problem of face detection by the face-processing network. Insights obtained from years of developing computer-vision algorithms to solve this task have suggested that it may be efficiently and effectively solved by detection and integration of local contrast features. Does the brain use a similar strategy? To answer this question, I embark on a journey that takes me through the development and optimization of dedicated tools for targeting and perturbing deep brain structures. Data collected using MR-guided electrophysiology in early face-processing regions was found to have strong selectivity for contrast features, similar to ones used by artificial systems. While individual cells were tuned for only a small subset of features, the population as a whole encoded the full spectrum of features that are predictive to the presence of a face in an image. Together with additional evidence, my results suggest a possible computational mechanism for face detection in early face processing regions. To move from correlation to causation, I focus on adopting an emergent technology for perturbing brain activity using light: optogenetics. While this technique has the potential to overcome problems associated with the de-facto way of brain stimulation (electrical microstimulation), many open questions remain about its applicability and effectiveness for perturbing the non-human primate (NHP) brain. In a set of experiments, I use viral vectors to deliver genetically encoded optogenetic constructs to the frontal eye field and faceselective regions in NHP and examine their effects side-by-side with electrical microstimulation to assess their effectiveness in perturbing neural activity as well as behavior. Results suggest that cells are robustly and strongly modulated upon light delivery and that such perturbation can modulate and even initiate motor behavior, thus, paving the way for future explorations that may apply these tools to study connectivity and information flow in the face processing network.
Resumo:
As the worldwide prevalence of diabetes mellitus continues to increase, diabetic retinopathy remains the leading cause of visual impairment and blindness in many developed countries. Between 32 to 40 percent of about 246 million people with diabetes develop diabetic retinopathy. Approximately 4.1 million American adults 40 years and older are affected by diabetic retinopathy. This glucose-induced microvascular disease progressively damages the tiny blood vessels that nourish the retina, the light-sensitive tissue at the back of the eye, leading to retinal ischemia (i.e., inadequate blood flow), retinal hypoxia (i.e., oxygen deprivation), and retinal nerve cell degeneration or death. It is a most serious sight-threatening complication of diabetes, resulting in significant irreversible vision loss, and even total blindness.
Unfortunately, although current treatments of diabetic retinopathy (i.e., laser therapy, vitrectomy surgery and anti-VEGF therapy) can reduce vision loss, they only slow down but cannot stop the degradation of the retina. Patients require repeated treatment to protect their sight. The current treatments also have significant drawbacks. Laser therapy is focused on preserving the macula, the area of the retina that is responsible for sharp, clear, central vision, by sacrificing the peripheral retina since there is only limited oxygen supply. Therefore, laser therapy results in a constricted peripheral visual field, reduced color vision, delayed dark adaptation, and weakened night vision. Vitrectomy surgery increases the risk of neovascular glaucoma, another devastating ocular disease, characterized by the proliferation of fibrovascular tissue in the anterior chamber angle. Anti-VEGF agents have potential adverse effects, and currently there is insufficient evidence to recommend their routine use.
In this work, for the first time, a paradigm shift in the treatment of diabetic retinopathy is proposed: providing localized, supplemental oxygen to the ischemic tissue via an implantable MEMS device. The retinal architecture (e.g., thickness, cell densities, layered structure, etc.) of the rabbit eye exposed to ischemic hypoxic injuries was well preserved after targeted oxygen delivery to the hypoxic tissue, showing that the use of an external source of oxygen could improve the retinal oxygenation and prevent the progression of the ischemic cascade.
The proposed MEMS device transports oxygen from an oxygen-rich space to the oxygen-deficient vitreous, the gel-like fluid that fills the inside of the eye, and then to the ischemic retina. This oxygen transport process is purely passive and completely driven by the gradient of oxygen partial pressure (pO2). Two types of devices were designed. For the first type, the oxygen-rich space is underneath the conjunctiva, a membrane covering the sclera (white part of the eye), beneath the eyelids and highly permeable to oxygen in the atmosphere when the eye is open. Therefore, sub-conjunctival pO2 is very high during the daytime. For the second type, the oxygen-rich space is inside the device since pure oxygen is needle-injected into the device on a regular basis.
To prevent too fast or too slow permeation of oxygen through the device that is made of parylene and silicone (two widely used biocompatible polymers in medical devices), the material properties of the hybrid parylene/silicone were investigated, including mechanical behaviors, permeation rates, and adhesive forces. Then the thicknesses of parylene and silicone became important design parameters that were fine-tuned to reach the optimal oxygen permeation rate.
The passive MEMS oxygen transporter devices were designed, built, and tested in both bench-top artificial eye models and in-vitro porcine cadaver eyes. The 3D unsteady saccade-induced laminar flow of water inside the eye model was modeled by computational fluid dynamics to study the convective transport of oxygen inside the eye induced by saccade (rapid eye movement). The saccade-enhanced transport effect was also demonstrated experimentally. Acute in-vivo animal experiments were performed in rabbits and dogs to verify the surgical procedure and the device functionality. Various hypotheses were confirmed both experimentally and computationally, suggesting that both the two types of devices are very promising to cure diabetic retinopathy. The chronic implantation of devices in ischemic dog eyes is still underway.
The proposed MEMS oxygen transporter devices can be also applied to treat other ocular and systemic diseases accompanied by retinal ischemia, such as central retinal artery occlusion, carotid artery disease, and some form of glaucoma.
Resumo:
Current technological advances in fabrication methods have provided pathways to creating architected structural meta-materials similar to those found in natural organisms that are structurally robust and lightweight, such as diatoms. Structural meta-materials are materials with mechanical properties that are determined by material properties at various length scales, which range from the material microstructure (nm) to the macro-scale architecture (μm – mm). It is now possible to exploit material size effect, which emerge at the nanometer length scale, as well as structural effects to tune the material properties and failure mechanisms of small-scale cellular solids, such as nanolattices. This work demonstrates the fabrication and mechanical properties of 3-dimensional hollow nanolattices in both tension and compression. Hollow gold nanolattices loaded in uniaxial compression demonstrate that strength and stiffness vary as a function of geometry and tube wall thickness. Structural effects were explored by increasing the unit cell angle from 30° to 60° while keeping all other parameters constant; material size effects were probed by varying the tube wall thickness, t, from 200nm to 635nm, at a constant relative density and grain size. In-situ uniaxial compression experiments reveal an order-of-magnitude increase in yield stress and modulus in nanolattices with greater lattice angles, and a 150% increase in the yield strength without a concomitant change in modulus in thicker-walled nanolattices for fixed lattice angles. These results imply that independent control of structural and material size effects enables tunability of mechanical properties of 3-dimensional architected meta-materials and highlight the importance of material, geometric, and microstructural effects in small-scale mechanics. This work also explores the flaw tolerance of 3D hollow-tube alumina kagome nanolattices with and without pre-fabricated notches, both in experiment and simulation. Experiments demonstrate that the hollow kagome nanolattices in uniaxial tension always fail at the same load when the ratio of notch length (a) to sample width (w) is no greater than 1/3, with no correlation between failure occurring at or away from the notch. For notches with (a/w) > 1/3, the samples fail at lower peak loads and this is attributed to the increased compliance as fewer unit cells span the un-notched region. Finite element simulations of the kagome tension samples show that the failure is governed by tensile loading for (a/w) < 1/3 but as (a/w) increases, bending begins to play a significant role in the failure. This work explores the flaw sensitivity of hollow alumina kagome nanolattices in tension, using experiments and simulations, and demonstrates that the discrete-continuum duality of architected structural meta-materials gives rise to their flaw insensitivity even when made entirely of intrinsically brittle materials.
Resumo:
Experiments are described using the random dot stereo patterns devised by Julesz, but substituting various colors and luminances for the usual black and white random squares. The ability to perceive the patterns in depth depends on a luminance difference between the colors used. If two colors are the same luminance, then depth is not perceived although each of the individual squares which make up the patterns is easily seen due to the color difference. This is true for any combination of different colors. If different colors are used for corresponding random squares between the left and right eye patterns, stereopsis is possible for all combinations of binocular rivalry in color, provided the luminance difference is large enough. Rivalry in luminance always precludes stereopsis, regardless of the colors involved.