Feature extraction and automatic recognition of wear particles in ferro-graphic image based on Riesz transform

Wear Particle Analysis, as an effective method in mechanical equipment condition monitoring, has been widely and successfully applied to many fields, i.e. weapon equipment, maintenance and daily management. To avoid the influences such as complexity of tribo-system, scrambling and randomicity of the wear particle, an image analysis technique based on Riesz transforms has been proposed to extract features efficiently and recognize wear particles automatically. Local magnitude and local orientation is firstly estimated using Riesz transform. Then feature parameters have been extracted using the generalization of the traditional Canny edge detection procedure and the mean shift based color image segmentation. Finally, the principal component analysis (PCA) has been employed to automatically recognize wear particles.