SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
A method of mid-level pattern recognition provides for a pose invariant Hough Transform by parametrizing pairs of points in a pattern with respect to at least two reference points, thereby providing a parameter table that is scale- or rotation-invariant. A corresponding inverse transform may be applied to test hypothesized matches in an image and a distance transform utilized to quantify the level of match.
Image Pattern Recognition attempts to infer properties of the world from image data. Such capabilities are crucial for making measurements from satellite or telescope images related to Earth and space science problems. Such measurements can be the required product itself, or the measurements can be used as input to a computer graphics system for visualization purposes. At present, the field of image pattern recognition lacks a unified scientific structure for developing and evaluating image pattern recognition applications. The overall goal of this project is to begin developing such a structure. This report summarizes results of a 3-year research effort in image pattern recognition addressing the following three principal aims: (1) to create a software foundation for the research and identify image pattern recognition problems in Earth and space science; (2) to develop image measurement operations based on Artificial Visual Systems; and (3) to develop multiscale image descriptions for use in interactive image analysis.
New foundational ideas are used to define a novel approach to generic visual pattern recognition. These ideas proceed from the starting point of the intrinsic equivalence of noise reduction and pattern recognition when noise reduction is taken to its theoretical limit of explicit matched filtering. This led us to think of the logical extension of sparse coding using basis function transforms for both de-noising and pattern recognition to the full pattern specificity of a lexicon of matched filter pattern templates. A key hypothesis is that such a lexicon can be constructed and is, in fact, a generic visual alphabet of spatial vision. Hence it provides a tractable solution for the design of a generic pattern recognition engine. Here we present the key scientific ideas, the basic design principles which emerge from these ideas, and a preliminary design of the Spatial Vision Tree (SVT). The latter is based upon a cryptographic approach whereby we measure a large aggregate estimate of the frequency of occurrence (FOO) for each pattern. These distributions are employed together with Hamming distance criteria to design a two-tier tree. Then using information theory, these same FOO distributions are used to define a precise method for pattern representation. Finally the experimental performance of the preliminary SVT on computer generated test images and complex natural images is assessed.
Many recent theories and methods proposed for visual preprocessing and shape representation are summarized. The survey brings together research from the fields of biology, psychology, computer science, electrical engineering, and most recently, neural networks. It was motivated by the need to preprocess images for a sparse distributed memory (SDM), but the techniques presented may also prove useful for applying other associative memories to visual pattern recognition. The material of this survey is divided into three sections: an overview of biological visual processing; methods of preprocessing (extracting parts of shape, texture, motion, and depth); and shape representation and recognition (form invariance, primitives and structural descriptions, and theories of attention).
Topics related to photoreceptors are considered, giving attention to visual pattern recognition and directional orientation in insects, the sensory basis of orientation in amphibians, and the aerial and underwater visual acuity in the California sea lion as a function of luminance. Other subjects explored are in the fields of phonoreceptors, chemoreception, vestibular receptors, and electrical and magnetic sensitivity. Questions of the development and evolution of orientation are also investigated, taking into account field studies of mass emigration and orientation in the spiny lobster and investigations concerning the jumping behavior in the Gobiid fish. Individual items are announced in this issue.
A top-down image processing scheme is described. A three-dimensional model of a robotic working environment, with robot manipulators, workpieces, cameras, and on-the-scene visual enhancements is employed to control and direct the image processing, so that rapid, robust algorithms act in an efficient manner to continually update the model. Only the model parameters are communicated, so that savings in bandwidth are achieved. This image compression by modeling is especially important for control of space telerobotics. The background for this scheme lies in an hypothesis of human vision put forward by the senior author and colleagues almost 20 years ago - the Scanpath Theory. Evidence was obtained that repetitive sequences of saccadic eye movements, the scanpath, acted as the checking phase of visual pattern recognition. Further evidence was obtained that the scanpaths were apparently generated by a cognitive model and not directly by the visual image. This top-down theory of human vision was generalized in some sense to the frame in artificial intelligence. Another source of the concept arose from bioengineering instrumentation for measuring the pupil and eye movements with infrared video cameras and special-purpose hardware.
A fundamental element of future generic pattern recognition technology is the ability to extract similar patterns for the same scene despite wide ranging extraneous variables, including lighting, turbidity, sensor exposure variations, and signal noise. In the process of demonstrating pattern constancy of this kind for retinex/visual servo (RVS) image enhancement processing, we found that the pattern constancy performance depended somewhat on scene content. Most notably, the scene topography and, in particular, the scale and extent of the topography in an image, affects the pattern constancy the most. This paper will explore these effects in more depth and present experimental data from several time series tests. These results further quantify the impact of topography on pattern constancy. Despite this residual inconstancy, the results of overall pattern constancy testing support the idea that RVS image processing can be a universal front-end for generic visual pattern recognition. While the effects on pattern constancy were significant, the RVS processing still does achieve a high degree of pattern constancy over a wide spectrum of scene content diversity, and wide ranging extraneousness variations in lighting, turbidity, and sensor exposure.
Volcanoes appearing in SAR images of the surface of Venus are reliably detected and located automatically by a system designed to handle general visual pattern recognition problems.
We propose a sequential information maximization model as a general strategy for programming eye movements. The model reconstructs high-resolution visual information from a sequence of fixations, taking into account the fall-off in resolution from the fovea to the periphery. From this framework we get a simple rule for predicting fixation sequences: after each fixation, fixate next at the location that minimizes uncertainty (maximizes information) about the stimulus. By comparing our model performance to human eye movement data and to predictions from a saliency and random model, we demonstrate that our model is best at predicting fixation locations. Modeling additional biological constraints will improve the prediction of fixation sequences. Our results suggest that information maximization is a useful principle for programming eye movements.
Primates can generate accurate, smooth eye-movement responses to moving target objects of arbitrary shape and size, even in the presence of complex backgrounds and/or the extraneous motion of non-target objects. Most previous studies of pursuit have simply used a spot moving over a featureless background as the target and have thus neglected critical issues associated with the general problem of recovering object motion. Visual psychophysicists and theoreticians have shown that, for arbitrary objects with multiple features at multiple orientations, object-motion estimation for perception is a complex, multi-staged, time-consuming process. To examine the temporal evolution of the motion signal driving pursuit, we recorded the tracking eye movements of human observers to moving line-figure diamonds. We found that pursuit is initially biased in the direction of the vector average of the motions of the diamond's line segments and gradually converges to the true object-motion direction with a time constant of approximately 90 ms. Furthermore, transient blanking of the target during steady-state pursuit induces a decrease in tracking speed, which, unlike pursuit initiation, is subsequently corrected without an initial direction bias. These results are inconsistent with current models in which pursuit is driven by retinal-slip error correction. They demonstrate that pursuit models must be revised to include a more complete visual afferent pathway, which computes, and to some extent latches on to, an accurate estimate of object direction over the first hundred milliseconds or so of motion.
Response latency in visual pattern perception situation
Processing of digital images destined for visual consumption raises many interesting questions regarding human visual sensitivity. This talk will survey some of these questions, including some that have been answered and some that have not. There will be an emphasis upon visual masking, and a distinction will be drawn between masking due to contrast gain control processes, and due to processes such as hypothesis testing, pattern recognition, and visual search.
Several recent studies have shown performance decrements with increasing display size when background texture elements are present in a same-different feature discrimination task--a result that challenges the traditional notion that the identities of simple visual features are processed in parallel, preattentively. Four experiments are reported that explore the implications of these results. Experiment 1 replicates the recent studies but limits the generalizability of the results to small target numbers. Experiments 2 and 3 show that the observed performance decrements are not due to a serial or even limited-capacity, parallel process. Experiment 4 suggests that decision factors idiosyncratic to the use of texture elements in a same-different task are responsible for the effect. It is concluded that the identification of simple visual features proceeds in parallel, with unlimited capacity (i.e., preattentively).
Multichannel optical correlator implements shift-invariant, high-discrimination pattern-recognizing neural network based on paradigm of neocognitron. Selected as basic building block of this neural network because invariance under shifts is inherent advantage of Fourier optics included in optical correlators in general. Neocognitron is conceptual electronic neural-network model for recognition of visual patterns. Multilayer processing achieved by iteratively feeding back output of feature correlator to input spatial light modulator and updating Fourier filters. Neural network trained by use of characteristic features extracted from target images. Multichannel implementation enables parallel processing of large number of selected features.
Recently, quantitative models based on signal detection theory have been successfully applied to the prediction of human accuracy in visual search for a target that differs from distractors along a single attribute (feature search). The present paper extends these models for visual search accuracy to multidimensional search displays in which the target differs from the distractors along more than one feature dimension (conjunction, disjunction, and triple conjunction displays). The model assumes that each element in the display elicits a noisy representation for each of the relevant feature dimensions. The observer combines the representations across feature dimensions to obtain a single decision variable, and the stimulus with the maximum value determines the response. The model accurately predicts human experimental data on visual search accuracy in conjunctions and disjunctions of contrast and orientation. The model accounts for performance degradation without resorting to a limited-capacity spatially localized and temporally serial mechanism by which to bind information across feature dimensions.
Investigations of working memory tend to focus on the retention of verbal information. The present experiments were designed to characterize the active maintenance rehearsal process used in the retention of visuospatial information. Rhesus monkeys (Macaca mulatta; N = 6) were tested as well as humans (total N = 90) because these nonhuman primates have excellent visual working memory but, unlike humans, cannot verbally recode the stimuli to employ verbal rehearsal mechanisms. A series of experiments was conducted using a distractor-task paradigm, a directed forgetting procedure, and a dual-task paradigm. No evidence was found for an active maintenance process for either species. Rather, it appears that information is maintained in the visuospatial sketchpad without active rehearsal.
The ModelFest Phase One dataset is a collection of luminance contrast thresholds for 43 two-dimensional monochromatic spatial patterns confined to an area of approximately two by two degrees. These data were collected by a collaboration among twelve laboratories, and were designed to provide a common database for calibration and testing of spatial vision models. Here I report fits of the ModelFest data with five models: Peak Contrast, Contrast Energy, Generalized Energy, a Gabor Channels model, and a Discrete Cosine Transform model. The Gabor Channels model provides the best fit, though the other, simpler models, with the exception of Peak Contrast, provide remarkably good fits as well. Though there are clear individual differences, regularities in the data suggest the possibility of constructing a standard observer for spatial vision. c2000 Optical Society of America.
The binding of identity and location information in disjunctive feature search was studied. Ss searched a heterogeneous display for a color or a form target, and reported both target identity and location. To avoid better than chance guessing of target identity (by choosing the target less likely to have been seen), the difficulty of the two targets was equalized adaptively; a mathematical model was used to quantify residual effects. A spatial layout was used that minimized postperceptual errors in reporting location. Results showed strong binding of identity and location perception. After correction for guessing, no perception of identity without location was found. A weak trend was found for accurate perception of target location without identity. We propose that activated features generate attention-calling "interrupt" signals, specifying only location; attention then retrieves the properties at that location.