A New Paradigm in Surface Reconstruction: LTP Features from Surface Images for Inverse Process Modeling

This study explores the adaptation of Local Ternary Patterns (LTP) for surface reconstruction, extending its use beyond traditional texture analysis. A systematic methodology, including patch optimization, parameter optimization, and feature selection, is proposed to extract surface-relevant features from images. Applied to spark-eroded surfaces, LTP enables differentiation of surface structures, with slightly reduced classification performance for surfaces with very fine or very coarse roughness. Despite this, the method demonstrates high potential for inverse process modeling, offering a cost-effective alternative for correlating manufacturing parameters with surface characteristics, contributing to improved process optimization and quality control in precision manufacturing.