Donald E. Waagen

Image super-resolution for improved automatic target recognition

Infrared imagers used to acquire data for automatic target recognition are inherently limited by the physical properties of their components. Fortunately, image super-resolution techniques can be applied to overcome the limits of these imaging systems. This increase in resolution can have potentially dramatic consequences for improved automatic target recognition (ATR) on the resultant higher-resolution images. We will discuss superresolution techniques in general and specifically review the details of one such algorithm from the literature suited to real-time application on forward-looking infrared (FLIR) images. Following this tutorial, a numerical analysis of the algorithm applied to synthetic IR data will be presented, and we will conclude by discussing the implications of the analysis for improved ATR accuracy.

Sparse Manifold Learning with Applications to SAR Image Classification

Nonlinear data-driven dimensionality reduction techniques have recently gained popularity due to the emergence of high dimensional data sets. The algorithmic complexity and storage requirements of these techniques, however, can make them prohibitive in resource-limited applications. It is therefore beneficial to reduce the number of exemplar samples required for performing an out-of-sample extension to a test point. In this paper, we propose a novel method for selecting a minimal set of exemplars and performing the out-of-sample extension. In the case of two-class target recognition with synthetic aperture radar (SAR) data, we compare the efficacy of the proposed approach with other approaches for selecting a subset of the available training samples. We show that the proposed algorithm outperforms the existing methods by providing low-dimensional embeddings that maintain interclass separability using fewer retained exemplars.

A combined particle/Kalman filter for improved tracking of beam aspect targets

Track continuity is difficult to maintain when tracking beam aspect targets. The loss of Doppler discrimination allows clutter to mask the target return, making it nearly impossible to detect. In order to improve tracking performance, a combination particle/Kalman filter has been developed. The tracking filters obviate each other as appropriate. When a target enters a Doppler blind zone, the particle filter replaces the Kalman filter as the tracking algorithm until the target exits the zone. The particle filter expands over the uncertainty region so that when the target is once again visible, it can immediately resume track via re-initialization of the Kalman filter. This paper discusses the design and simulation of this algorithm and shows the resulting improvement in track continuity. We briefly discuss how our combined particle/Kalman filter approach can be used to address the problem of targets obscured in altitude return.

Activities in integrated sensing and processing

Advances in sensor technologies, computation devices and algorithms have created enormous opportunities for intelligent, automated threat identification, target acquisition, and surveillance. Unfortunately, as information requirements grow, conventional network processing techniques require ever-increasing bandwidth between sensors and processors, as well as potentially exponentially complex data reduction methods. Practical approaches require that the sensing and computation be jointly engineered, to raise the quality of data and classification and to minimize computation, power consumption and cost. This paper summarizes several activities at Raytheon investigating the integration of sensing and processing and presents some preliminary results.

A statistical analysis of 3D structure tensor features generated from LADAR imagery

Extraction and efficient representation of informative structure from data is the goal of pattern recognition. Efficient and effective parametric and nonparametric representations for capturing the geometry of three-dimensional objects are an area of current research. Tang and Medioni have proposed tensor representations for characterization and reconstruction of surfaces. 3-D structure tensors are extracted by mapping surface geometries using a rank-2 covariant tensor. Distributional differences between representations of objects of interest can (theoretically) be used for target matching and identification. This paper analyzes the statistical distributions of tensor representation extracted from 3-D LADAR imagery and quantifies a measure of divergence between images of three vehicles as a function of tensor feature support size.

Statistical analysis of spin images for differentiation of wheeled and tracked objects

Spin images originated within the robotics group at Carnegie Mellon University and are representations of 3-space surface regions. This representation provides a means for surface matching that is invariant to rigid body rotations and translations while being robust in the presence of 3D image noise, clutter, and surface occlusion. Of particular interest is the viability of using spin images to differentiate between two object classes in 3D imagery where there is significant intra-class diversity, e.g. to differentiate between wheeled and tracked vehicles. The specificity of spin map representations in differentiation of wheeled and tracked vehicles is statistically characterized. Using synthetic imagery of various wheeled and tracked vehicles, the class separability of wheeled vs. tracked vehicle spin image sets is nonparametrically quantified via entropic characterization as well as the Friedman-Rafsky two-sample test statistic. Additionally, class separability is analyzed in lower dimensional feature spaces generated via the Hotelling transform as well as a random projection method, comparing and contrasting the spin map class differentiation in the original and transformed data sets.

Multiscale target manifold characterization for 3D imaging ladar

Manifold extraction techniques, such as ISOMAP, are capable of projecting nonlinear, high-dimensional data to a lower-dimensional subspace while retaining discriminatory information. In this investigation, ISOMAP is applied to 3D LADAR range imagery. Selected man-made objects are reduced to sets of spin-image feature vectors that describe object surface geometries. At various spin-image support scales, we use the distribution-free Henze-Penrose statistic test to quantify differences between man-made objects in both the high-dimensional spin-image vector representation and in the low-dimensional spin-image manifold extracted using ISOMAP.

Integrated sensing and processing for distributed sensor networks

Advances in sensor technologies, computation devices, and algorithms have created enormous opportunities for intelligent, automated threat identification, target acquisition, and surveillance. Unfortunately, as information requirements grow, conventional network processing techniques require ever increasing bandwidth between sensors and processors, as well as potentially exponentially complex data reduction methods. Practical approaches requires that the sensing and computation be jointly engineered, to raise the quality of data and classification, and minimize computation, power consumption, and cost.

Exploration of high-dimensional data manifolds for object classification

This investigation discusses the challenge of target classification in terms of intrinsic dimensionality estimation and selection of appropriate feature manifolds with object-specific classifier optimization. The feature selection process will be developed via nonlinear characterization and extraction of the target-conditional manifolds derived from the training data. We investigate defining the feature space used for classification as a class-conditioned nonlinear embedding, i.e., each training and test image is embedded in a target-specific embedding and the resultant embeddings are used for statistical characterization. We compare and contrast this novel embedding technique with Principal Component Analysis. The α-Jensen Entropy Difference measure is used to quantify the object-conditioned separation between the target distributions in the feature spaces. We discuss and demonstrate the effect of feature space extraction on classification efficacy.

A comparison of particle filters and multiple-hypothesis extended Kalman filters for bearings-only tracking

Bearings-only tracking is widely used in the defense arena. Its value can be exploited in systems using optical sensors and sonar, among others. Non-linearity and non-Gaussian prior statistics are among the complications of bearings-only tracking. Several filters have been used to overcome these obstacles, including particle filters and multiple hypothesis extended Kalman filters (MHEKF). Particle filters can accommodate a wide range of distributions and do not need to be linearized. Because of this they seem ideally suited for this problem. A MHEKF can only approximate the prior distribution of a bearings-only tracking scenario and needs to be linearized. However, the likelihood distribution maintained for each MHEKF hypothesis demonstrates significant memory and lends stability to the algorithm, potentially enhancing tracking convergence. Also, the MHEKF is insensitive to outliers. For the scenarios under investigation, the sensor platform is tracking a moving and a stationary target. The sensor is allowed to maneuver in an attempt to maximize tracking performance. For these scenarios, we compare and contrast the acquisition time and mean-squared tracking error performance characteristics of particle filters and MHEKF via Monte Carlo simulation.