Manuel F. Fernandez

Fusion of adaptive algorithms for the classification of sea mines using high resolution side scan sonar in very shallow water

A new sea mine computer-aided-detection/computer-aided-classification (CAD/CAC) processing string has been developed. The CAD/CAC processing string consists of preprocessing, adaptive clutter filtering (ACF), normalization, detection, feature extraction, classification and fusion processing blocks. The range-dimension ACF is an adaptive linear FIR filter, which is matched both to average highlight and shadow information, while also simultaneously suppressing background clutter. For each detected object, features are extracted and processed through an orthogonalization transformation, enabling an efficient application of the optimal log-likelihood-ratio-test (LLRT) classification rule, in the orthogonal feature space domain. The classified objects of 3 distinct processing strings, developed by 3 different researchers, are fused, using the classification confidence values as features and logic-based, M-out-of-N, or novel LLRT-based fusion rules. The utility of the overall processing strings and their fusion was demonstrated with new very shallow water high-resolution sonar imagery data. The processing string detection and classification parameters were tuned and the string classification performance was optimized, by appropriately selecting a subset of the original feature set. It was shown that LLRT-based fusion algorithms outperform the logic-based or the M-out-of-N ones. The LLRT-based fusion of the CAD/CAC processing strings resulted in a four-fold false alarm rate reduction, compared to the best single CAD/CAC processing string results, while maintaining a constant correct mine classification probability.

Processing string fusion approach investigation for automated sea mine classification in shallow water

A novel sea mine computer-aided-detection / computer-aided-classification (CAD/CAC) processing string has been developed. The overall CAD/CAC processing string consists of pre-processing, adaptive clutter filtering (ACF), normalization, detection, feature extraction, feature orthogonalization, optimal subset feature selection, classification and fusion processing blocks. The range-dimension ACF is matched both to average mine highlight and shadow information, while also adaptively suppressing background clutter. For each detected object, features are extracted and processed through an orthogonalization transformation, enabling an efficient application of the optimal log-likelihood-ratio-test (LLRT) classification rule, in the orthogonal feature space domain. The classified objects of 4 distinct processing strings are fused using the classification confidence values as features and M-out-of-N, or LLRT-based fusion rules. The utility of the overall processing strings and their fusion was demonstrated with new shallow water high-resolution sonar imagery data. The processing string detection and classification parameters were tuned and the string classification performance was optimized, by appropriately selecting a subset of the original feature set. Two significant improvements were made to the CAD/CAC processing string by employing sub-image adaptive clutter filtering (SACF) and utilizing a repeated application of the subset feature selection/LLRT classification blocks. It was shown that LLRT-based fusion of the CAD/CAC processing strings outperforms the M-out-of-N algorithms and results in up to an eight-fold false alarm rate reduction, compared to the best single CAD/CAC processing string results, while maintaining a constant correct mine classification probability.

Processing string fusion for automated sea mine classification in shallow water

A novel sea mine computer-aided-detection / computer-aided-classification (CAD/CAC) processing string has been developed. The overall CAD/CAC processing string consists of pre-processing, adaptive clutter filtering (ACF), normalization, detection, feature extraction, feature orthogonalization, optimal subset feature selection, classification and fusion processing blocks. The range-dimension ACF is matched both to average highlight and shadow information, while also adaptively suppressing background clutter. For each detected object, features are extracted and processed through an orthogonalization transformation, enabling an efficient application of the optimal log-likelihood-ratio-test (LLRT) classification rule, in the orthogonal feature space domain. The classified objects of 3 distinct processing strings are fused using the classification confidence values as features and logic-based, M-out-of-N, or LLRT-based fusion rules. The utility of the overall processing strings and their fusion was demonstrated with shallow water high-resolution sonar imagery data. The processing string detection and classification parameters were tuned and the string classification performance was optimized, by appropriately selecting a subset of the original feature set. A significant improvement was made to the CAD/CAC processing string by utilizing a repeated application of the subset feature selection / LLRT classification blocks. It was shown that LLRT-based fusion algorithms outperform the logic based or the M-out-of-N ones. The LLRT-based fusion of the CAD/CAC processing strings resulted in up to a eight-fold false alarm rate reduction, compared to the best single CAD/CAC processing string results, while maintaining a constant correct mine classification probability.

Cascaded Volterra fusion of processing strings for automated sea mine classification in shallow water

An improved sea mine computer-aided-detection/computer-aided-classification (CAD/CAC) processing string has been developed. The overall CAD/CAC processing string consists of pre-processing, adaptive clutter filtering (ACF), normalization, detection, feature extraction, optimal subset feature selection, feature orthogonalization, classification, and fusion processing blocks. The range-dimension ACF is matched both to average highlight and shadow information, while also adaptively suppressing background clutter. For each detected object, features are extracted and processed through an orthogonalization transformation, enabling an efficient application of the optimal log-likelihood-ratio-test (LLRT) classification rule, in the orthogonal feature space domain. The classified objects of 4 distinct processing strings are fused using the classification confidence values as features and either M-out-of-N or LLRT-based fusion rules. The utility of the overall processing strings and their fusion was demonstrated with new shallow water high-resolution sonar imagery data. The processing string detection and classification parameters were tuned and the string classification performance was optimized, by appropriately selecting a subset of the original feature set. Two significant fusion algorithm improvements were made. First, a new nonlinear (Volterra) feature LLRT fusion algorithm was developed. Second, a repeated application of the subset feature selection/feature orthogonalization/Volterra feature LLRT fusion block was utilized. It was shown that this cascaded Volterra feature LLRT fusion of the CAD/CAC processing strings outperforms the M-out-of-N and baseline LLRT algorithms, yielding significant improvements over the best single CAD/CAC processing string results, and providing the capability to correctly call all mine targets while maintaining a very low false alarm rate

Main-beam multi-target monopulse super-resolution

Current monopulse radar does not resolve multiple sources within the main beam, a scenario that arises in a number of radar applications such as: ballistic missile defense, where the incoming missile complex consists of a large number of objects; air defense, where a detection may correspond to one or multiple targets; cruise missile defense for low angle target tracking in multipath; etc. This paper describes a super-resolution procedure designed for radar and EW applications requiring resolving multiple targets within the main beam given a single snapshot of array data. Our procedure enables the practical implementation of such super-resolution process by first forming multiple receive beams to provide data and degree-of-freedom reductions without significant information loss. Super-resolution is then achieved by applying matrix techniques to the beamspace data. The overall approach can be interpreted in terms of projecting sensor data into a compressed “information domain”. In our particular application this results in a generalization of the monopulse processing scheme to the main-beam multi-target case.

Dual-sided subspace mappings for main-beam multi-target super-resolution in clutter and interference

The need for localizing multiple main-beam targets in the presence of interference, clutter and jammers arises in scenarios such as: air defense, where a radar detection may correspond to one or multiple targets (e.g., a missile launched from an airborne platform); ballistic missile defense, where the incoming missile complex involves a large number of objects; cruise missile defense for low angle target tracking in multipath; etc. This paper presents a super-resolution approach for the sub-beamwidth localization of multiple main-beam targets in the presence of clutter and interference, doing so with a single snapshot of sensor array data while placing nulls at specified locations. Such process first maps array data into compact subspaces containing the information of interest (e.g., a cluster of receive beams with imbedded nulls). Super-resolution is then achieved via small-matrix operations on the subspace data. The ensuing result is a mapping of array data into a compressed “information domain,” yielding an effective, practical Sensor-to-Information process that is more accurate, robust and versatile than current super-resolution methods.

Analog beamspace super-resolution radar processing

Current monopulse radar does not resolve multiple sources within the main beam, a scenario that arises in a number of radar applications such as ballistic missile defense, where the incoming missile complex consists of a large number of objects; air defense, where a detection may correspond to a single plane or multiple planes; cruise missile defense for low angle target tracking in multipath; etc. This paper describes a superresolution technique designed for radar applications requiring resolving multiple targets within the main beam given a single snapshot of multiple beam data. The technique enables the practical implementation of such super-resolution algorithm by first forming multiple beams in the analog domain to provide data and degree-of-freedom (DOF) reductions without significant loss of information. Super-resolution is then achieved through the use of matrix processing techniques operating on the digitized beamspace data. The overall procedure can be considered a generalization of the monopulse processing scheme to the multi-target case.

Blocking-matrix and quasimatrix techniques for extended-null insertion in antenna pattern synthesis

Inserting extended nulls while preserving a desired beam pattern is a topic of interest for sensor array pattern synthesis in radar, sonar, communications, acoustics, etc., where nulls are needed to combat interference from known directions. Blocking-matrices can address this problem by modifying the weights forming a quiescent antenna pattern, but they require trading the degrees of freedom used for null insertion with those needed for pattern preservation. This paper uses quasi- and mixed-matrix techniques to obtain the minimal set of blocking-matrix basis vectors that best define an extended continuous null; it then presents efficient practical implementation procedures for inserting single as well as multiple extended and/or discrete nulls.

Digital beamforming of sub-aperture cluster beams with enhanced angle estimation capabilities

We are motivated to exploit advances in digital radar technology and super-resolution techniques to improve current radar search-and-track functions. In this paper, we focus on the benefits of using digital beamforming for enhancing angle estimation capabilities. These capabilities include flexibility in monopulse processing, using multiple simultaneous beams to extend angle coverage and resolving multiple main-beam targets using clusters of sub-aperture beams. The super-resolution capability is a salient feature to be included in next generation radars that may need to cope with stressful scenarios of closely-spaced targets, or it can be used to relax the resolution requirement in radar design.

Automatic target recognition algorithm for high resolution multi-band sonar imagery

An improved automatic target recognition processing string has been developed. The overall processing string consists of pre-processing, subimage adaptive clutter filtering, normalization, detection, data regularization, feature extraction, optimal subset feature selection, feature orthogonalization and classification processing blocks. The classified objects of 3 distinct strings are fused using the classification confidence values and their expansions as features, and using ldquosummingrdquo or log-likelihood-ratio-test (LLRT) based fusion rules. The utility of the overall processing strings and their fusion was demonstrated with new high-resolution threefrequency band sonar imagery. The ATR processing strings were individually tuned to the corresponding three-frequency band data. Two significant fusion algorithm improvements were made. First, a nonlinear 2nd order (Volterra) feature LLRT fusion algorithm was developed. Second, a repeated application of a subset Volterra feature selection / feature orthogonalization / LLRT fusion block was utilized. It was shown that cascaded Volterra feature LLRT fusion of the ATR processing strings outperforms baseline ldquosummingrdquo and single-stage Volterra feature LLRT algorithms, yielding significant improvements over the best single ATR processing string results, and providing the capability to correctly call the majority of targets while maintaining a very low false alarm rate.