Gregory A. Clark

A unified approach to time- and frequency-domain realization of FIR adaptive digital filters

Specific implementations of the finite impulse response (FIR) block adaptive filter in the frequency domain are presented and some of their important properties are discussed. The time-domain block adaptive filter implemented in the frequency domain is shown to be equivalent to the frequency-domain adaptive filter (derived in the frequency domain), provided data sectioning is done properly. All of the known time- and frequency-domain adaptive filters [1]-[12], [16]-[18] are contained in the set of possible block adaptive filter structures. Thus, the block adaptive filter is generic and its formulation unifies the current theory of time- and frequency-domain FIR adaptive filter structures. A detailed analysis of overlap-save and overlap-add implementations shows that the former is to be preferred for adaptive applications because it requires less computation, a fact that is not true for fixed coefficient filters.

Texture segmentation by clustering of Gabor feature vectors

Approaches the texture segmentation problem by clustering feature vectors created from a Gabor transform data block. Given an N*N image, the authors compute 24 Gabor transforms using Gabor kernels with six orientations and four sizes. This results in a Gabor data block composed of N/sup 2/ feature vectors of length 24. The feature vectors are then grouped based on their distribution in the high-dimensional feature space. The authors hypothesize that the pixels in a given group have similar characteristics, and thus are part of the same texture. Experimental results for segmenting a synthetic railroad track image were encouraging; a clear-cut segmentation of the image was obtained. By clustering Gabor features, the authors were able to segment an image into regions of uniform texture without prior knowledge of the types of texture, or the frequency and orientation characteristics of these textures. The clustering algorithm is a modified Kohonen self-organizing feature map.<<ETX>>

Detecting buried objects by fusing dual-band infrared images

The authors have conducted experiments to demonstrate the enhanced detectability of buried land mines using sensor fusion techniques. Multiple sensors, including visible imagery, infrared imagery, and ground penetrating radar (GPR), have been used to acquire data on a number of buried mines and mine surrogates. Because the visible wavelength and GPR data are currently incomplete, the paper focuses on the fusion of two-band infrared images. The authors use feature-level fusion and supervised learning with the probabilistic neural network (PNN) to evaluate detection performance. The novelty of the work lies in the application of advanced target recognition algorithms, the fusion of dual-band infrared images and evaluation of the techniques using two real data sets.<<ETX>>

Sensor feature fusion for detecting buried objects

Given multiple registered images of the earths surface from dual-band infrared sensors, our system fuses information from the sensors to reduce the effects of clutter and improve the ability to detect buried or surface target sites. The sensor suite currently includes two infrared sensors (5 micron and 10 micron wavelengths) and one ground penetrating radar (GPR) of the wide-band pulsed synthetic aperture type. We use a supervised learning pattern recognition approach to detect metal and plastic land mines buried in soil. The overall process consists of four main parts: preprocessing, feature extraction, feature selection, and classification. We present results of experiments to detect buried land mines from real data, and evaluate the usefulness of fusing feature information from multiple sensor types, including dual-band infrared and ground penetrating radar. The novelty of the work lies mostly in the combination of the algorithms and their application to the very important and currently unsolved operational problem of detecting buried land mines from an airborne standoff platform.

Buried object remote detection technology for law enforcement

A precise airborne temperature-sensing technology to detect buried objects for use by law enforcement is developed. Demonstrations have imaged the sites of buried foundations, walls and trenches; mapped underground waterways and aquifers; and been used to locate underground military objects. The methodology is incorporated in a commercially available, high signal-to-noise, dual-band infrared scanner with real-time, 12-bit digital image processing software and display. The method creates color-coded images based on surface temperature variations of 0.2 degree(s)C. Unlike other less-sensitive methods, it maps true (corrected) temperatures by removing the (decoupled) surface emissivity mask equivalent to 1 degree(s)C or 2 degree(s)C; this mask hinders interpretation of apparent (blackbody) temperatures. Once removed, it is possible to identify surface temperature patterns from small diffusivity changes at buried object sites which heat and cool differently from their surroundings. Objects made of different materials and buried at different depths are identified by their unique spectral, spatial, thermal, temporal, emissivity and diffusivity signatures. The authors have successfully located the sites of buried (inert) simulated land mines 0.1 to 0.2 m deep; sod-covered rock pathways alongside dry ditches, deeper than 0.2 m; pavement covered burial trenches and cemetery structures as deep as 0.8 m; and aquifers more than 6 m and less than 60 m deep. The technology could be adapted for drug interdiction and pollution control. For the former, buried tunnels, underground structures built beneath typical surface structures, roof-tops disguised by jungle canopies, and covered containers used for contraband would be located. For the latter, buried waste containers, sludge migration pathways from faulty containers, and the juxtaposition of groundwater channels, if present, nearby, would be depicted. The precise airborne temperature-sensing technology has a promising potential to detect underground epicenters of smuggling and pollution.

Automatic image analysis for detecting and quantifying gamma-ray sources in coded-aperture images

We report the development of an automatic image analysis system that detects gamma-ray source regions in images obtained from a coded aperture, gamma-ray imager. The number of gamma sources in the image is not known prior to analysis. The system counts the number (K) of gamma sources detected in the image and estimates the lower bound for the probability that the number of sources in the image is K. The system consists of a two-stage pattern classification scheme in which the probabilistic neural network is used in the supervised learning mode. The algorithms were developed and tested using real gamma-ray images from controlled experiments in which the number and location of depleted uranium source disks in the scene are known. The novelty of the work lies in the creative combination of algorithms and the successful application of the algorithms to real images of gamma-ray sources.

Dual-band infrared capabilities for imaging buried object sites

We discuss dual-band infrared (DBIR) capabilities for imaging buried object sites. We identify physical features affecting thermal contrast needed to distinguish buried object sites from undisturbed sites or surface clutter. Apart from atmospheric transmission and system performance, these features include: object size, shape, and burial depth; ambient soil, disturbed soil and object site thermal diffusivity differences; surface temperature, emissivity, plant-cover, slope, albedo and roughness variations; weather conditions and measurement times. We use ground instrumentation to measure the time-varying temperature differences between buried object sites and undisturbed soil sites. We compared near surface soil temperature differences with radiometric infrared (IR) surface temperature differences recorded at 4.7 +/- 0.4 micrometers and at 10.6 +/- 1.0 micrometers . By producing selective DBIR image ratio maps, we distinguish temperature-difference patterns from surface emissivity effects. We discuss temperature differences between buried object sites, filled hole sites (without buried objects), cleared (undisturbed) soil sites, and grass-covered sites (with and without different types of surface clutter). We compared temperature, emissivity-ratio, visible and near-IR reflectance signatures of surface objects, leafy plants and sod. We discuss the physical aspects of environmental, surface and buried target features affecting interpretation of buried targets, surface objects and natural backgrounds.

Computer vision for locating buried objects

Given two registered images of the Earth, measured with aerial dual-band infrared (IR) sensors, the authors use advanced computer vision/automatic target recognition techniques to estimate the positions of buried land mines. The images are very difficult to interpret, because of large amounts of clutter. Conventional techniques use single-band imagery and simple correlations. They rely heavily on the judgment of the human doing the interpretation, and give unsatisfactory results with difficult data sets of the type analyzed here. The automatic algorithms used by the authors are able to eliminate most of the clutter and give greatly improved indications of regions in the image that could be interpreted as mines. The novelty of the present approach lies in the following aspects: (1) a patented data fusion technique using two IR images and physical principles based on Plancks law; (2) a new region-based texture segmentation algorithm using Gabor transform features and a clustering/thresholding algorithm based on a neural network (self-organizing feature map); (3) prior knowledge of measured feasible temperatures and emissivities; and (4) results with real data using buried surrogate mines.<<ETX>>

Computer vision for detecting and quantifying gamma-ray sources in coded-aperture images

We report the development of an automatic image analysis system that detects gamma-ray source regions in images obtained from a coded aperture, gamma-ray imager. The number of gamma sources in the image is plot known prior to analysis. The system counts the number (K) of gamma sources detected in the image and estimates the lower bound for the probability that the number of sources in the image is K. The system consists of a two-stage pattern classification scheme in which the probabilistic neural network is used in the supervised learning mode. The algorithms were developed and tested using real gamma-ray images from controlled experiments in which the number and location of depleted uranium source disks in the scene are known.<<ETX>>

Efficient realization of adaptive digital filters in the time and frequency domains

A unified approach to the efficient realization of FIR adaptive digital filters in both the time and frequency domains is presented. It is shown that previously published frequency domain implementations are contained within this unified approach. It is further shown that various combinations of implementations are possible, and that in working with data streams, care must be taken in using overlap-save and overlap-add sectioning procedures for proper frequency domain implementations.