Genetic Algorithm Based Feature Selection for Fracture Surface Images Classification

Feature extraction and feature selection of fracture surface images provided by scanning electron microscopy (SEM) are two main challenges in classification of metal fracture surface images. In extracting features, the statistical characteristics of gray-level co-occurrence matrix and the fractal dimension of fracture surface images are computed as feature sets; and then , a genetic algorithm (GA) approach is presented to select a subset of features to discriminate different classes fracture surface images. A new fitness function based on minimum description length and maximum class separability is proposed to drive GA and it is compared with other feature selection method. Experimental results show that the GA driven by it selected a good subset of features to discriminate fracture surface images effectively.