Texture features for bulk rock material grain boundary segmentation

Abstract

Determination of the grains size is one of the most common activities in the mineral processing industry. Application of machine vision to support the process has been the field of research for several years. Many trials start with detection of grains boundaries. The article deals with the problem of determining the boundaries of rock grains using texture analysis. The description of textures using Haralick features was subjected to research. The effect of the window size used to determine the features for a selected point on the effectiveness of grain edge discrimination was considered. The impact of different classification methods on the results obtained was assessed. The obtained results indicate that the most effective is the use of small window sizes (9–13).

Keywords

• Image processing
• Computer vision
• Machine learning
• Grain border determination