engleski

Near-optimal MST-based Shape Description using Genetic Algorithm

A new method for the selection of the optimal structuring element for shape description and matching based on the morphological signature transform (MST) is presented in this paper. In MST shape description a representation based on the areas of the input binary image successively eroded by multiple rotated structuring elements at different resolutions is used. The multiple structuring elements are derived by rotation from the initial structuring element. This paper proposes a new method for the optimal selection of the initial structuring element. For a given class of shapes the optimal structuring element for MST method is selected by means of a genetic algorithm. The optimization criteria is formulated to enable a robust shape matching. The MST-based shape description method using the resulting optimal structuring element is optimal in the sense that it enables best discrimination of object shapes. A genetic algorithm was used to create new generations of structuring elements. The genetic algorithm utilized unconventional two-dimensional chromosomes to represent structuring elements. A new two-dimensional crossover operator with two crossover points was used to generate new individuals. Population entropy was used to measure the population diversity. A variable crossover rate was utilized and controlled by population entropy. The variable crossover rate was used to avoid the problem of premature convergence. The fitness criteria was formulated to enable shape matching with good discrimination property. The result of the genetic algorithm was the near-optimal structuring element which was used for MST-based shape description and matching. The proposed optimal shape representation method was applied to the problem of shape matching which evolves in many object recognition applications. Here, an unknown object from the input image is matched to a set of known objects in order to classify it into one of finite number of possible classes. Experiments have been performed on a class of model shapes to demonstrate the usefulness of the method. The genetic algorithm was used to determine the optimal structuring element for predefined class of shapes. The shapes from the class were distorted by boundary noise, rotation, and scaling to create the class of noisy shapes. The noisy shapes were matched to the original shapes using MST shape description based on the obtained optimal structuring element. Experimental results are presented and discussed.

genetic algorithm ; mathematical morphology ; image analysis ; shape description

nije evidentirano