Cross-Sensor Micro-Texture Material Classification and Smartphone Acquisition do not go well together

Abstract

Intrinsic, non-invasive product authentication is still an important topic as it does not generate additional costs during the production process. This topic is of specific interest for medical products as non-genuine products can directly effect the patients’ health. This work investigates micro-texture classification as a mean of proving the authenticity of zircon oxide blocks (for dental implants). Samples of three different manufacturers were acquired using four smartphone devices with a clip-on macro lens. In addition, an existing drug packaging material database was utilized. While the intra-sensor microtexture classification worked well, the cross-sensor classification results were less promising. In an attempt to track down the limiting factors, intrinsic sensor features usually used in device identification were investigated as well.