Multi-objective Optimization on Pore Segmentation

In order to segment pores automatically without parameters set manually, it is necessary to design an adaptive algorithm which may be applied for different kinds of hardwood cross-section images. A novel adaptive method is proposed in this paper to evaluate the optimal threshold of closed region area for pore segmentation. Based on area histogram, this method classifies the regions into two classes with maximum between-class variance. Experiment shows that the method has more effective to diffuse porous wood and pore solitary, but many pores cannot be segmented for semi-diffuse porous wood, ring-porous wood or other pore combination except solitary pore. According to the domain knowledge of wood science, second objective function is used to improve the pore segmentation performance. Further experiment on genetic algorithm demonstrates that the task of pore segmentation can be completed successfully for all kinds of hardwood by multi-objective function.