A hybrid evolutionary learning algorithm is presented that synthesizes a complete multiclass pattern recognition system. The approach uses a multifaceted representation that evolves layers of processing to perform feature extraction from raw input data, select cooperative sets of feature detectors, and assemble a linear classifier that uses the detectors' responses to label targets. The hybrid algorithm, called hybrid evolutionary learning for pattern recognition (HELPR), blends elements of evolutionary programming, genetic programming, and genetic algorithms to perform a search for an effective set of feature detectors. Individual detectors are represented as expressions composed of morphological and arithmetic operations. Starting with a few small random expressions, HELPR expands the number and complexity of the features to produce a recognition system that achieves high accuracy. Results are presented that demonstrate the performance of HELPR-generated recognition systems applied to the task of classification of high-range resolution radar signals.

This correspondence presents two soft morphological algorithms that process multiple images simultaneously. The first algorithm performs best when the structuring elements contain less than 19 points; whereas, the second algorithm should be used for larger structuring elements. Theoretical and experimental analyses show these algorithms are faster than the conventional algorithm.

This paper describes a system based on evolutionary learning, called MORPH, that semi-automates the generation of morphological programs. MORPH maintains a population of morphological programs that is continually enhanced. The first phase of each learning cycle synthesizes morphological sequences that extract novel features which increase the population's diversity. The second phase combines these newly formed operator sequences into larger programs that are better than the individual programs. A stochastic selection process eliminates the poor performers, while the survivors serve as the basis of another learning cycle. Experimental results are presented for binary and grayscale target recognition problems.

The camera calibration process relates camera system measurements (pixels) to known reference points in a three-dimensional world coordinate system. The calibration process is viewed as consisting of two independent phases: the first is removing geometrical camera distortion so that rectangular calibration grids are straightened in the image plane, and the second is using a linear affine transformation as a map between the rectified camera coordinates and the geometrically projected coordinates on the image plane of known reference points. Phase one is camera-dependent, and in some systems may be unnecessary. Phase two is concerned with a generic model that includes 12 extrinsic variables and up to five intrinsic parameters. General methods handling additional constraints on the intrinsic variables in a manner consistent with explicit satisfaction of all six constraints on the orthogonal rotation matrix are presented. The use of coplanar and noncoplanar calibration points is described.>