RFI ON ALGORITHMS AND ANALYSIS OF HYPERSPECTRAL DATA REQUEST FOR INFORMATION (RFI) FROM ALL SOURCES ON Algorithms, Analysis, and Processing of Hyperspectral Data. Ball Systems Engineering Operation (Ball) is gathering information regarding the algorithms, analysis, and processing of hyperspectral data for the government. The information will be used in developing plans for processing and exploitation systems architectures for the 1999 -- 2002 time period. Ball is issuing this Request For Information (RFI) on behalf of the government. Commercial, government, and academic sources involved with any of the components of the spectral data industry that may be affected by algorithm development and data processing are invited to respond. If this RFI is only partially applicable to your business, partial answers would be appropriate and appreciated. Our current plan is to collect, review, and identify the white papers with possible government interests applicability. This information could be used for research funding, utility benefit analysis, investment decisions and future requirements definition. The government has been gathering information regarding algorithm development and hyperspectral data processing in preparation for developing a government Spectral Processing Architecture. Selected subcontract opportunities may result from this RFI. Ball is willing to sign non-disclosure agreements with program participants. For the purpose of this RFI, hyperspectral data is defined as data from a sensor which measures the spectral properties of natural and man-made features/objects using hundreds of contiguous narrow band (bandwidth ranging from 0.1 to 10 nm) discrete channels allowing fine feature identification and discrimination. Nonliteral exploitation is defined as the process of extracting information from digital data, automatically or semi-automatically, using nontraditional, advanced processing techniques, which may employ models, measurements, signatures, or other information, to perform one or more of the following functions at the requisite levels of detail: detect, geolocate, classify, discriminate, characterize, identify, quantify, point, track, predict, target, or map and chart objects, emissions, events, or activities of interest represented in the data. The focus of this RFI is to learn what industry has been doing with algorithms, analysis, and processing of hyperspectral data. This information is important in light of the government's increasing dependence on remotely sensed data and the need for efficient processing. In order to develop a more robust spectral processing architecture for future applications, information is requested in the following areas: 1) anomaly detection with high emphasis on unsupervised automated techniques, 2) terrain classification, 3) target or material identification, 4) analysis and processing, 5) gas detection, 6) analysis of gas mixtures, 7) material quantification (i. e. target size) and 8) multi-sensor data fusion [all types of data available including, but not limited to, panchromatic, spectral, thermal, and radar sensors]. Major focus should concentrate on results with high probability of detection and low false alarm rates, confidence estimates in results, and factors that degrade algorithm results due to collection platform, parameters, and geometry or environment. 1. These questions are intended to determine the maturity and robustness of the areas of interest: a. The following questions are related to Anomaly Detection. * How are shadows and anomalies in shadows treated? * How do your algorithms handle mixed pixels? * How are the default or final thresholds calculated? * If you do not have an automated anomaly detection algorithm, then describe the supervised anomaly detection algorithm. b. The following questions are related to Terrain Categorization (TERCAT). * How are shadows and anomalies in shadows treated? * How do your algorithms handle mixed pixels? * How are the default or final thresholds calculated? * If you do not have an automated anomaly detection algorithm, then describe supervised anomaly detection algorithm. c. The following questions are related to Target Identification. * What targets have you tested against? * How do your algorithms handle mixed pixels? * What level of confidence do you have with your results? * How robust are the signatures? d. The following questions are related to Gas Detection and Analysis. * What experience and algorithms do you have for gas detection and analysis? * What level of confidence do you have with your results? * What data were used in the analyses and results? * What method do you use to correct for the atmospheric effects? * What algorithms do you use for material discrimination? e. The following questions are related to Material Quantification. * What hyperspectral data sources have you used? * What targets have you tested against? * What level of confidence do you have with your results? * What are you error estimates when calculating target size? f. The following questions are related to Data Fusion and All Source Data Products. * What other data sources haveyou fused with hyperspectral data? * What techniques do you use to fuse hyperspectral and other source data? * What fused products are produced ? * How have your all-source fusion technique(s) (with or without spectral data) helped answer questions? * Which problems of the government's interests have you addressed? g. The following questions are related to Testing & Validating your results. * How are you testing & validating your algorithms or spectral data products? * What data sources and data sets were used during your algorithm development and testing? * Were the results generated using a full data set or a small subset? * If the hyperspectral bands aggregated together, then describe the aggregation process. * What "truth" data was available? What and how was the truth data used? * What are the optimal conditions for maximum performance with the algorithm? * If you present Probability of False Alarms and Probability of Detection results, then please describe how these values are generated. 2. Other: a) Provide any further thoughts you may have about the algorithms, analysis, and processing of spectral data and nonliteral exploitation. b) Identify government interests satisfied or partially satisfied with spectral data products. c) With whom would you recommend we pursue additional discussions on these topics? Responses: Written responses should contain the following information: a) executive summary, b) background and previous related work, c) examples of results, d) explanation of results; performance of algorithms quantified in a well defined way (i.e. were analyses and results presented for an entire data cube or over a small subset), e) data sets used in analysis or algorithm development, f) desired areas of participation, g) proposed work and requisite funding, h) optional softcopy submissions must be in Microsoft Word/Excel/PowerPoint formats, i) proprietary information must be explicitly identified Evaluation process: Submissions will be evaluated against already existing algorithms in aneffort to develop ideas that will significantly improve the performance of hyperspectral data exploitation technology currently available to the government. Preference will be given to well quantified results. You are free to answer some or all of the above questions. All concise written responses to this RFI must be received at Ball Systems Engineering Operation; Attn: Peter Telek -- RFI Material; 2875 Presidential Drive, Suite 180; Fairborn, OH 45324 no later than 30 days after date of publication of this RFI. If you choose to submit classified or proprietary data, please mark it accordingly to ensure proper handling. Ball is neither negotiating nor calling for proposals during this process. This RFI does not constitute a commitment on the part of Ball to purchase or acquire services. Ball will not reimburse travel and response preparation expenses for this effort. RFI Project Focal Point: Peter Telek, (413) 684-4478. E-MAIL: click here to contact the project focal point via, ptelek@ball.com. Posted 04/01/99 (W-SN315489).

Loren Data Corp. http://www.ld.com (SYN# 0445 19990405\SP-0001.MSC)