Arthur C. Kenton

Statistical parametric signature/sensor/detection model for multispectral mine target detection

A statistical parametric multispectral sensor performance model was developed by ERIM to support mine field detection studies, multispectral sensor design/performance trade-off studies, and target detection algorithm development. The overall model incorporates four components; a mission flight model, a multispectral target and background signature model, a multispectral sensor model, and a multispectral target detection model. Emphasis is placed on estimating the effects of mission multispectral target detection algorithms. Thus, the model ideally supports mission and multispectral sensor trade studies which require optimization of the systems overall target detection performance. The model and a typical example of performance prediction results are presented.

Detection of land mines with hyperspectral data

The objective of the US Army Hyperspectral Mine Detection Phenomenology program was to determine if spectral discriminants exist that are useful for the detection of land mines. Statistically significant mine signature data were collected over a wide spectral range and analyzed to identify robust spectral features that might serve as discriminants for new airborne sensor concepts. Detection metrics which characterize the deductibility of land miens and which predict the detection performance of a general class of hyperspectral detection algorithms were selected and applied. Detection performance of land mines was analyzed against background type, age of buried miens and possible sensor design parameters. This paper describes the result of this analysis and present EO/IR hyperspectral sensor and algorithm design concepts that could potentially be used to operationally detect buried land mines.

Laboratory characterization and field testing of the tunable filter multispectral camera

The Coastal Systems Station, in concert with Xybion Corp. has developed a tunable-filter multispectral imaging sensor for use in airborne reconnaissance. The sensor was completed in late 1999, and laboratory characterization and field- testing has been conducted since. The Tunable Filter Multispectral Camera (TFMC) is an intensified, gated, and tunable multispectral imaging camera that provides three simultaneous channels of 10-bit digital and 8-bit analog video from the near-UV to the near-IR. Exposure and gain can be automatically or manually controlled for each channel, and response has been linearized for approximate radiometric use. Additionally, each of the three channels as a separate programmable liquid-crystal tunable filter with a selectable center wavelength settings to which can be applied 100 different retardances for each of three channels. This paper will present setups, analysis methods, and preliminary results for both the laboratory characterization and field- testing of the TFMC. Laboratory objectives include measures of sensitivity, noise, and linearity. Field testing objectives include obtaining the camera response as the lighting conditions approached sunset of a clear day, signal-to-clutter ratios for a multiplicity of channel wavelength combinations and polarizations against several backgrounds, and resolution performance in field-conditions.

Vision Guidance Update: Synthetic Aperture Radar (SAR) Multiple Image Exploitation For Position And Velocity Determination

Vision Guidance Update (VGU) is a Synthetic Aperture Radar (SAR) multi-image exploitation technique which uses radar vision derived from range-angle imagery of stationary ground features to provide precision platform position and velocity estimates. The concept addresses applications to both navigational update and guidance functions for systems using SAR as an autonomous, long-range, standoff, all-weather, day/night sensing capability. The changes in perspectives of ground features detected in successive SAR spotlight images are analyzed and the coupled geometry problem solved to determine precision platform velocity and position estimates relative to any imaged point on the ground. A SAR VGU theory and algorithm were developed that utilizes two features in two images for deriving the platform kinematic estimates. Realistic SAR finite resolution effects lead to estimation errors. Errors in the knowledge of platform position and velocity, generated from navigation system errors, produce imaging geometry and image formation errors. Tolerance to reasonable platform kinematic errors is provided by iterating the SAR VGU algorithm to improve and provide precision estimates of all geometry, system, and kinematic parameters. The VGU theory and algorithm are presented with the initial modeling assumptions. Initial simulation results are qualitatively summarized.

Background adaptive multispectral band selection

AN initial automated band selection algorithm suitable for real-time application with tunable multispectral cameras is presented for multispectral target detection. The method and algorithm were developed from analyses of several background and target signatures collected from a field test using the prototype Tunable Filter Multispectral Camera (TFMC). Target and background data from TFMC imagery were analyzed to determine the detection performance of 32,768 unique 3-band channel combinations in the visible through and near IR spectral regions. This tuning knowledge base was analyzed to develop rules for an initial dynamic tuning algorithm. The performance data was sorted by conventional means to determine the best 3-band combinations. Methods were then developed to determine performance enhancing band sets for particular backgrounds and a variety of targets. This knowledge is then used in an algorithm to affect a real-time 3-band tuning capability. Additional band sets for real-time background categorization are chosen by both the ability to spectrally detect of one background from another. This work will illustrate an example of the performance results form the analysis for three targets on various backgrounds.

Sensor point spread function effects on the statistics of multispectral target signatures

A statistical parametric multispectral sensor performance model was developed by ERIM to support mine field detection studies, multispectral sensor design/performance trade-off studies, and target detection algorithm development. A key element in this performance model is the influence of the background on the targets multispectral statistics due to the size and shape of the target under the sensors point spread function and pixel sampling function. The multispectral statistics of interest include the first-order (mean) and second-order moments (covariance) of the targets radiance signature. This paper presents a formulation which not only considers the effects of a multispectral sensor with a single point spread function, but also considers the joint effects of multiple, potentially misregistered, point spread functions on the targets covariance statistics. The model and an example of sensor point spread function and pixel sampling function effects on the targets spectral statistics are presented.

Littoral assessment of mine burial signatures (LAMBS): buried landmine/background spectral-signature analyses

The objective of the Office of Naval Research (ONR) Rapid Overt Reconnaissance (ROR) program and the Airborne Littoral Reconnaissance Technologies (ALRT) projects LAMBS effort is to determine if electro-optical spectral discriminants exist that are useful for the detection of land mines in littoral regions. Statistically significant buried mine overburden and background signature data were collected over a wide spectral range (0.35 to 14 µm) to identify robust spectral features that might serve as discriminants for new airborne sensor concepts. LAMBS has expanded previously collected databases to littoral areas - primarily dry and wet sandy soils - where tidal, surf, and wind conditions can severely modify spectral signatures. At AeroSense 2003, we reported completion of three buried mine collections at an inland bay, Atlantic and Gulf of Mexico beach sites. We now report LAMBS spectral database analyses results using metrics which characterize the detection performance of general types of spectral detection algorithms. These metrics include mean contrast, spectral signal-to-clutter, covariance, information content, and spectral matched filter analyses. Detection performance of the buried land mines was analyzed with regard to burial age, background type, and environmental conditions. These analyses considered features observed due to particle size differences, surface roughness, surface moisture, and compositional differences.

Real-time implementation of a multispectral mine target detection algorithm

Spatial-spectral anomaly detection (the “RX Algorithm”) has been exploited on the USMCs Coastal Battlefield Reconnaissance and Analysis (COBRA) Advanced Technology Demonstration (ATD) and several associated technology base studies, and has been found to be a useful method for the automated detection of surface-emplaced antitank land mines in airborne multispectral imagery. RX is a complex image processing algorithm that involves the direct spatial convolution of a target/background mask template over each multispectral image, coupled with a spatially variant background spectral covariance matrix estimation and inversion. The RX throughput on the ATD was about 38X real time using a single Sun UltraSparc system. A goal to demonstrate RX in real-time was begun in FY01. We now report the development and demonstration of a Field Programmable Gate Array (FPGA) solution that achieves a real-time implementation of the RX algorithm at video rates using COBRA ATD data. The approach uses an Annapolis Microsystems Firebird PMC card containing a Xilinx XCV2000E FPGA with over 2,500,000 logic gates and 18MBytes of memory. A prototype system was configured using a Tek Microsystems VME board with dual-PowerPC G4 processors and two PMC slots. The RX algorithm was translated from its C programming implementation into the VHDL language and synthesized into gates that were loaded into the FPGA. The VHDL/synthesizer approach allows key RX parameters to be quickly changed and a new implementation automatically generated. Reprogramming the FPGA is done rapidly and in-circuit. Implementation of the RX algorithm in a single FPGA is a major first step toward achieving real-time land mine detection.

Joint spectral region buried land mine discrimination performance

Statistically significant sets of buried anti-tank mine and background electro-optic spectral signatures were collected and analyzed by the Veridian ERIM International team under the US Armys Night Vision and Electronic Sensors Directorate Hyperspectral Mine Detection Phenomenology FY98/99 Program as reported last year. Those analyses established predicted buried mine spectral discrimination performance in key practical sensor spectral regions using typical multispectral sensor bandwidths spanning 20 to 200 nm. This year, we report further analyses of selected sets of HMDP data that quantitatively predict performance for two specific cases of joint spectral regions. This work exhibits these initial results and compares the predicted buried mine spectral discrimination performance determined from the joint and the single spectral regions.

COBRA ATD minefield detection model initial performance analysis

A statistical performance analysis of the USMC Coastal Battlefield Reconnaissance and Analysis (COBRA) Minefield Detection (MFD) Model has been performed in support of the COBRA ATD Program under execution by the Naval Surface Warfare Center/Dahlgren Division/Coastal Systems Station . This analysis uses the Veridian ERIM International MFD model from the COBRA Sensor Performance Evaluation and Computational Tools for Research Analysis modeling toolbox and a collection of multispectral mine detection algorithm response distributions for mines and minelike clutter objects. These mine detection response distributions were generated form actual COBRA ATD test missions over littoral zone minefields. This analysis serves to validate both the utility and effectiveness of the COBRA MFD Model as a predictive MFD performance too. COBRA ATD minefield detection model algorithm performance results based on a simulate baseline minefield detection scenario are presented, as well as result of a MFD model algorithm parametric sensitivity study.