Surface roughness analysis of side-polished fiber based on the importance of texture features

A detailed analysis regarding the surface roughness of Side-polished Fiber (SPF) has been presented in this work using the Gray Level Co-occurrence Matrix (GLCM) technique. The flat areas of the SPF surface exhibited texture characteristics with higher mean values of contrast and entropy, and correspondingly lower values for the angular second moment (ASM), homogeneity, and correlation parameters. Such attributes corresponded to the smaller residual thickness on the flat areas for the polished surface of the fiber optic. Similarly, a stronger intensity for the light transmission with uneven distribution of gray levels along with the appearance of fine texture was also observed with their rich local features primarily oriented in a direction parallel to the fiber-core. Employing the Random Forest (RF) method of feature importance ranking which was based on the Gini coefficient and out-of-bag error estimation, this study assessed the sensitivity of various GLCM texture parameters in classifying roughness levels of the SPF polished surfaces. A feature subset comprising variance, ASM, entropy, and contrast was identified as an optimal set. Utilizing this subset, a validation experiment for analyzing the roughness of the SPF polished surfaces has been conducted via RF classification. The results demonstrated an RF classification accuracy of 95.65%. This research explores the impact of surface roughness on the mechanism of light coupling in SPF optic sensors employing environmental materials and its influence on sensor sensitivity. It lays the foundation for reconnoitering ways to precisely identify the high-sensitivity areas on SPF optic surfaces, thereby aiming to focus on the development of a rapidly emerging technology of a micro-probe type "lab-on-fiber" for photonics applications.