Expert System Application to Constant False Alarm Rate (cfar) Processor

An artificial intelligence system improves radar signal processor performance by increasing target probability of detection and reducing probability of false alarm in a severe radar clutter environment. This utilizes advances in artificial intelligence and expert systems technology for the development of data analysis and information (signal) processors used in conjunction with conventional (deterministic) data analysis algorithms to combine radar measurement data (including observed target tracks and radar clutter returns from terrain, sea, atmospheric effects, etc.) with topographic data, weather information, and similar information to formulate optimum filter coefficients and threshold tests. Present fielded radar systems use one CFAR algorithm for signal processing over the entire surveillance volume. However, radar experiments have shown that certain CFAR algorithms outperform others in different environments. The system intelligently senses the clutter environment, and selects and combines the most appropriate CFAR algorithm(s) to produce detection decisions that will outperform a processor using a single algorithm. The invention provides for improved performance through the application of rule-based and data-based expert system computer software technology to CFAR signal processors, thereby improving target detection by reducing processing losses which result from a mismatch between the single, fixed CFAR processor and dynamically changing environment in which a radar must operate. EXPERT SYSTEM CONSTANT FALSE ALARM RATE (CFAR) PROCESSOR STATEMENT OF GOVERNMENT INTEREST The invention described herein may be manufactured and used by or for the Government for governmental purposes without the payment of any royalty thereon. BACKGROUND OF THE INVENTION The present invention generally relates to constant false alarm rate (CFAR) signal processors, and in particular, to a CFAR which improves radar signal processor performance by increasing target probability of detection and reducing probability of false alarms in a severe radar clutter environment. False alarms are a significant problem in wide area surveillance radar. The conflicting requirements for a high probability of detection and a low probability of false alarm are rarely met due to the dynamically changing environment. The typical assumption of a homogeneous, Gaussian, thermal noise like background is routinely violated due to the spatial variation in clutter characteristics and effects of clutter edges, discretes, and multiple targets. Many different CFAR algorithms have been developed to effectively deal with the various types of backgrounds that are encountered. However, any single algorithm is likely to be inadequate in a dynamically changing environment as described Paper presented at the RTO SET Lecture Series on “Knowledge-Based Radar Signal and Data Processing”, held in Stockholm, Sweden, 3-4 November 2003; Rome, Italy, 6-7 November 2003; Budapest, Hungary, 10-11 November 2003; Madrid, Spain, 28-29 October 2004; Gdansk, Poland, 4-5 November 2004, and published in RTO-EN-SET-063. Expert System Application to Constant False Alarm Rate (CFAR) Processor 4 2 RTO-EN-SET-063 above. The approach suggested here is to intelligently select the CFAR algorithm or algorithms being executed at any given time, based upon the observed characteristics of the environment. This approach requires sensing the environment, employing the most suitable CFAR algorithm(s), and applying an appropriate multiple algorithm fusion scheme or consensus algorithm to produce a global detection decision. Adaptive threshold techniques are usually employed to control false alarm rates in varying background environments. The most common of these techniques is Constant False Alarm Rate (CFAR) processing. CFAR processors are designed to maintain a constant false alarm rate by adjusting the threshold for a cell under test by estimating the interference in the vicinity of the test cell. A “cell” is a sample in the domains of interest (eg: range, Doppler, angle, polarization). In general the data operated on by the CFAR processor may be pre-filtered to improve detection performance. This pre-filtering may include Doppler filtering, adaptive space-time processing, pre-whitening, and channel equalization. Constant False Alarm Rate (CFAR) signal processing for automatic detection radar is an important part of the system design problem. The classical theory is developed under the assumption that detection is to be performed for the targets in the presence of stationary, Gaussian noise with known statistics, i.e. receiver noise. For groundbased radar systems looking high above the nearrange clutter this was a valid assumption. For the case of modern, long-range, airborne surveillance radars, the situation is more complicated. The steep grazing angles associated with a look down radar produce far stronger clutter returns than those observed in any ground based radar, effectively masking targets flying above these clutter regions. Clutter changes dynamically as the platform moves such that the processor must effectively deal with clutter edges, discretes, multiple targets and non-Gaussian interference. The general theory for optimum detection in nonstationary, non-Gaussian clutter of interference is not well developed, even when the statistical properties of the environment are known. In an attempt to solve this problem, presently fielded radar systems employ canceller-based signal processing techniques which exploit anticipated differences between target returns and clutter. The classical techniques for automatic target detection in receiver noise are then employed. Non-zero clutter residues at the output of these canceller-based systems degrade the performance of classical detectors, and improved CFAR signal processing is required to achieve the desired probability of detection and probability of false alarm. Historically, the design of filters for clutter cancellation is performed separately from the design or CFAR signal processors. Consequently, detection performance will be suboptimum. Also, the selection of a single filter, and a single CFAR processor, to perform in all environments, will surely be mismatched to the ever changing radar returns, and will result in further degraded performance. The Expert System CFAR (ESCFAR) Processor solves this problem by intelligently sensing the environment, employing one or more CFAR algorithms for data analysis, and combining results to make detection decisions. The task of providing a CFAR processor that improves radar signal processor performance by increasing target probability of detection and reducing probability of false alarms in a severe radar clutter environment, is alleviated to some extent by the systems disclosed in the following U.S. patents, the disclosures of which are incorporated herein by reference: U.S. Pat. No. 5,075,856 issued to Kneizys et al; U.S. Pat. No. 5,093,665 issued to Wieler; U.S. Pat. No.5,063,607 issued to FitzHenry el al; U.S. Pat. No. 4,970,660 issued to Marchant; and U.S. Pat. No. 4,749,994 issued to Taylor. The patent to Marchant discloses an accumulated statistics CFAR method and device using integrated data to maximize the probability of target detection for a given false alarm rate. The remaining patents are of interest, but do not disclose improving performance through the application of rule-based and data-based expert system computer software technology to CFAR signal processors, thereby improving target detection by reducing processing losses which result from a mismatch between the single fixed CFAR processor and the dynamically changing