Search

Considerable attention has recently been focused on laser radar (ladar) systems for surveillance applications, because of the richness inherent in the three-dimensional data they collect. Several research groups have looked into the ability to exploit ladar data in automatic or assisted target recognition systems. Designers of practical ladar recognition systems must have answers to very fundamental questions related to the quality and quantity of data required, the fidelity of target models used by the algorithm, and the effects of incorporating prior knowledge. This paper presents implementation guidelines derived from a simulation-based analysis of many such factors that can affect classification accuracy, including measurement noise and corresponding noise model accuracy, number of measured points on target, target model accuracy, target pose error, prior information, and target occlusion. The study includes data from eight vehicles, including seven civilian automobiles of similar size and shape, chosen specifically to result in a 'hard' classification problem. Data are simulated under three hypothetical scenarios: 1) a pole-mounted ladar system for monitoring traffic through an intersection; 2) a low-flying helicopter-born ladar system for law enforcement applications; and 3) a low-altitude unmanned aerial vehicle (UAV) for military surveillance applications.

We derive approximate expressions for the probability of error in a two-class hypothesis testing problem in which the two hypotheses are characterized by zero-mean complex Gaussian distributions. These error expressions are given in terms of the moments of the test statistic employed and we derive these moments for both the likelihood ratio test, appropriate when class densities are known, and the generalized likelihood ratio test, appropriate when class densities must be estimated from training data. These moments are functions of class distribution parameters which are generally unknown so we develop unbiased moment estimators in terms of the training data. With these, accurate estimates of probability of error can be calculated quickly for both the optimal and plug-in rules from available training data. We present a detailed example of the behavior of these estimators and demonstrate their application to common pattern recognition problems, which include quantifying the incremental value of larger training data collections, evaluating relative geometry in data fusion from multiple sensors, and selecting a good subset of available features. Index Terms--Complex Gaussian channels, probability of error, model inaccuracy, moment estimators, Johnson's systems of distributions.

Methods to quantitatively assess a large number of small samples and the comparison of three models for synthetic aperture radar (SAR) images of 3-D objects with varying pose are presented. The relation of the model to the Gaussian distribution and its quantitative statistical assessment in terms of goodness-of-fit are discussed.

A family of conditionally Gaussian signal models for synthetic aperture radar (SAR) imagery is presented, extending a related class of models developed for high resolution radar range profiles. This signal model is robust with respect to the variations of the complex-valued radar signals due to the coherent combination of returns from scatterers as those scatterers move through relative distances on the order of a wavelength of the transmitted signal (target speckle). The target type and the relative orientations of the sensor, target, and ground plane parameterize the conditionally Gaussian model. Based upon this model, algorithms to jointly estimate both the target type and pose are developed. Performance results for both target pose estimation and target recognition are presented for publicly released data from the MSTAR program.

Considerable attention has recently been focused on laser radar (ladar) systems for surveillance applications, because of the richness inherent in the three-dimensional data they collect. Several research groups have looked into the ability to exploit ladar data in automatic or assisted target recognition systems. Designers of practical ladar recognition systems must have answers to very fundamental questions related to the quality and quantity of data required, the fidelity of target models used by the algorithm, and the effects of incorporating prior knowledge. This paper presents implementation guidelines derived from a simulation-based analysis of many such factors that can affect classification accuracy, including measurement noise and corresponding noise model accuracy, number of measured points on target, target model accuracy, target pose error, prior information, and target occlusion. The study includes data from eight vehicles, including seven civilian automobiles of similar size and shape, chosen specifically to result in a "hard" classification problem. Data are simulated under three hypothetical scenarios: 1) a pole-mounted ladar system for monitoring traffic through an intersection; 2) a low-flying helicopter-born ladar system for law enforcement applications; and 3) a low-altitude unmanned aerial vehicle (UAV) for military surveillance applications. [ABSTRACT FROM AUTHOR]

Describes a framework which allows for the direct comparison of alternate approaches to automatic target recognition from synthetic aperture radar images. Models for synthetic aperture radar data; Details of the conditionally Gaussian model; Estimation and recognition algorithms.