Cesar Bandera

Foveal machine vision systems

A class of machine vision systems is proposed, called foveal vision systems. These systems, modeled after advanced biological vision, feature space-variant (variable-resolution) imager topologies and a closed-loop system architecture. The imager topology is characterized by resolution which is high at the center of the sampling lattice and which decreases with distance from the center. The central axis is controlled by feedback from higher-level algorithms, allowing the allocation of sampling resources to the region(s) of interest and resulting in greater relevant information from the imager yet permitting considerable reduction in data. Preliminary investigations have demonstrated reductions in data structure size and computations of several orders of magnitude relative to conventional implementations. The savings factors increase with field-of-view and resolution.<<ETX>>

Foveal automatic target recognition using a multiresolution neural network

This paper presents a method for detecting and classifying a target from its foveal (graded resolution) imagery using a multiresolution neural network. Target identification decisions are based on minimizing an energy function. This energy function is evaluated by comparing a candidate blob with a library of target models at several levels of resolution simultaneously available in the current foveal image. For this purpose, a concurrent (top-down-and-bottom-up) matching procedure is implemented via a novel multilayer Hopfield neural network. The associated energy function supports not only interactions between cells at the same resolution level, but also between sets of nodes at distinct resolution levels. This permits features at different resolution levels to corroborate or refute one another contributing to an efficient evaluation of potential matches. Gaze control, refoveation to more salient regions of the available image space, is implemented as a search for high resolution features which will disambiguate the candidate blob. Tests using real two-dimensional (2-D) objects and their simulated foveal imagery are provided.

An MIMD computing platform for a hierarchical foveal machine vision system

A multiple instruction multiple data (MIMD) parallel computing platform built upon a network of TMS320C40/44s (C40/C44) for real-time image processing of a hierarchical foveal machine vision (HFMV) system is described in this paper. The architecture of the system, the parallel algorithm development environment, and strategies to map tasks into the computing platform are described. The platform supports both static and dynamic computing resource allocation. The performance of the computing platform is illustrated by examples.

An all CMOS foveal image sensor chip

A foveal image sensor chip that uses a standard CMOS process is presented. A new photo charge normalization scheme and its associated circuitry for use with active pixel sensors has been developed. The pixels are assembled into a multi-resolution rectilinear foveal image sensor chip using a novel access scheme to reduce the number of analog RAM cells needed. Localized spatial resolution declines monotonically with offset from the imagers optical axis, analogous to biological foveal vision. A proof-of-concept chip has been fabricated and tested.

Hierarchical multiresolution data structures and algorithms for foveal vision systems

A system that integrates space-variant sampling in foveal vision and time-variant allocation of imaging resources in active perception is sought. It is argued that by folding the foveal data into an image pyramid, the use of uniform lattice operators may be restored. In addition, the hierarchical multiresolution approach to the foveal vision problem suggests an explicit gaze control strategy. Thus, the organizational principles for an active foveal vision system incorporating both space-variant and time-variant resolution resource allocation may be approached through study of the foveal vision problem without prior solution to the active perception problem. Various foveal sensor geometries and the embedding of an image polygon data structure in the image pyramid are discussed, and algorithms and a gaze control strategy consistent with the suggested approach are presented.<<ETX>>

Real-time reconfigurable foveal target acquisition and tracking system

This paper presents a target acquisition and tracking system based on the biomimetic concept of foveal vision. The system electronically reconfigures the resolution, sizes, shape, and focal plane position of visual acuity to meet time- varying operational requirements while maximizing the relevance of acquired video. A reconfigurable multiresolution active pixel CMOS imaging array is integrated in a closed-loop fashion with video processing and configuration control. Imager and algorithm configuration is updated frame-by-frame and reactively to target and scene conditions. By dynamically tailoring the visual acuity of the senor itself, the relevance and acquired visual information is maximized and a fast update rate is achieved with reduced communications bandwidth and processing requirements throughout the entire system. The system also features small size and less power consumption, and does not require a pointing mechanism. The distinguishing features of reconfigurable foveal machine vision are presented, and the hardware and software architecture of the target acquisition and tracking system is discussed. Real-time experimental result for automated target search, detection, interrogation, and tracking are then presented.

Model-based automatic target recognition using hierarchical foveal machine vision

This paper presents a target detection and interrogation techniques for a foveal automatic target recognition (ATR) system based on the hierarchical scale-space processing of imagery from a rectilinear tessellated multiacuity retinotopology. Conventional machine vision captures imagery and applies early vision techniques with uniform resolution throughout the field-of-view (FOV). In contrast, foveal active vision features graded acuity imagers and processing coupled with context sensitive gaze control, analogous to that prevalent throughout vertebrate vision. Foveal vision can operate more efficiently in dynamic scenarios with localized relevance than uniform acuity vision because resolution is treated as a dynamically allocable resource. Foveal ATR exploits the difference between detection and recognition resolution requirements and sacrifices peripheral acuity to achieve a wider FOV (e.g. faster search), greater localized resolution where needed (e.g., more confident recognition at the fovea), and faster frame rates (e.g., more reliable tracking and navigation) without increasing processing requirements. The rectilinearity of the retinotopology supports a data structure that is a subset of the image pyramid. This structure lends itself to multiresolution and conventional 2-D algorithms, and features a shift invariance of perceived target shape that tolerates sensor pointing errors and supports multiresolution model-based techniques. The detection technique described in this paper searches for regions-of- interest (ROIs) using the foveal sensors wide FOV peripheral vision. ROIs are initially detected using anisotropic diffusion filtering and expansion template matching to a multiscale Zernike polynomial-based target model. Each ROI is then interrogated to filter out false target ROIs by sequentially pointing a higher acuity region of the sensor at each ROI centroid and conducting a fractal dimension test that distinguishes targets from structured clutter.

A novel disjoint set division algorithm for joint scheduling and routing in wireless sensor networks

High network connectivity and low energy consumption are two major challenges in wireless sensor networks (WSNs). It is even more challenging to achieve both at the same time. To tackle the problem, this paper proposes a novel disjoint Set Division (SEDO) algorithm for joint scheduling and routing in WSNs. We finely divide sensors into different disjoint sets with guaranteed connectivity based on their geographical locations to monitor the interested area. We propose a class of scheduling and routing algorithms, which sequentially schedule each disjoint set to be on and off and balance the energy consumption during packet transmission. Simulation results show that SEDO outperforms existing schemes with lower packet delivery latency and longer network lifetime.

Rich media streaming for just-in-time training of first responders

The diversity of first responders and of asymmetric threats precludes the effectiveness of any single training syllabus. Just-in-time training (JITT) addresses this variability, but requires training content to be quickly tailored to the subject (the threat), the learner (the responder), and the infrastructure (the C2 chain from DHS to the responder’s equipment). We present a distributed system for personalized just-in-time training of first responders. The authoring and delivery of interactive rich media and simulations, and the integration of JITT with C2 centers, are demonstrated. Live and archived video, imagery, 2-D and 3-D models, and simulations are autonomously (1) aggregated from object-oriented databases into SCORM-compliant objects, (2) tailored to the individual learner’s training history, preferences, connectivity and computing platform (from workstations to wireless PDAs), (3) conveyed as secure and reliable MPEG-4 compliant streams with data rights management, and (4) rendered as interactive high-definition rich media that promotes knowledge retention and the refinement of learner skills without the need of special hardware. We review the object-oriented implications of SCORM and the higher level profiles of the MPEG-4 standard, and show how JITT can be integrated into - and improve the ROI of - existing training infrastructures, including COTS content authoring tools, LMS/CMS, man-in-the-loop simulators, and legacy content. Lastly, we compare the audiovisual quality of different streaming platforms under varying connectivity conditions.

Foveal automatic target recognition using a neural network

This paper proposes a method for identifying and classifying a target from its foveal imagery using a neural network. The methods criterion for identifying a target is based on finding the global minimum of an energy function. This energy function is characterized by matching the candidate target and a library of target models at several levels of resolution of nonuniformly sampled foveal image data. For this purpose, a top-down and bottom-up (concurrent) matching procedure is implemented via a multi-layer Hopfield neural network. The corresponding energy function supports not only connections between cells at the same resolution level, but also interconnections between two sets of nodes at two different resolution levels. The proposed method also utilizes a feature analysis at the higher resolution levels of the target to relocate the center of the fovea to a more salient region of the target (gaze control). The results of an experimental scenario for foveal target recognition are presented.