Quantifying abrasion and micro-pits in polymer wear using image processing techniques

Identifying wear mechanisms from worn surfaces is a complex and tedious process involving high expertise. Wear scars from the contact surfaces can act as a potential indicator of wear process undergone by the material during the course of surface interaction. Currently, wear mechanism analysis is performed subjectively and usually only by the author(s) of the study. The aim of this paper is to develop image processing techniques to quantify two wear mechanisms: abrasion and micro-pitting. To characterize wear mechanisms, the required surface morphology was produced using wear testing of polymer in a twin-disc test rig. The micrographs of polymer contact surfaces were acquired using a conventional optical microscope. Subjective quality scores were collected by a structured human observer study to quantify the severity of micro pitting and abrasion. These scores were used to validate objective scores obtained using image processing techniques. Several image processing techniques are proposed to detect and quantify micro-pitting and abrasion mechanisms. These techniques consist of local and global thresholding segmentation with and without uneven illumination compensation, granulometry by binary opening and granulometry by gray-scale closing. The proposed image processing analysis reveals the severity of abrasion and pitting mechanisms (when they are dominant) which agree well with the mean opinion scores (MOS) given by observers. For abrasion, the correlation coefficients between objective scores obtained by different image processing techniques and subjective values show a high linear correlation. For pitting, we claim that objective values obtained from the new methods, distinguish the pitting mechanism. This study is a proof of concept towards automated identification and quantification of different wear mechanisms as a replacement for human observers.