Physical object identification using micro-structure images

A micro-structure is the common name for all fine-surface details and material properties visible when a physical object is examined at close range or under magnification. In its most basic form, the micro-structure image serves as a unique, non-cloneable identifier for that object (see Figure 1). It is non-cloneable as the current level of material science technology cannot practically produce a physical object with the precision required to clone a specific micro-structure.

This protection scheme is attractive and highly competitive for large-scale, mass-market applications because of the non-invasive character of the protection and its easy, fast verification by non-experts using a mobile device. Applications include security documents, luxury items, spare aviation parts, and electronics. The non-cloneable character and uniqueness also mean that the deployed processing chain as well as the identification and authentication technologies share many elements with existing biometrics systems.

Figure 1. (a) A handheld mobile acquisition of a SPIE Certificate and (b-c) two extracted patches of an identical sample without any special equipment or lighting. Histogram equalization was used for visualization purposes.

We focused our micro-structure architecture elements on extracting the correct image patch, selecting robust or invariant features, dimensionality reduction, and quantization, resulting in a binary representation of the original image (see Figure 2). The acquired samples contain a printed mark used as a guide to extract the correct image patch from a fixed, determined position containing the micro-structure. This extraction needs to be vastly more precise than, for example, computer vision stitching applications. Extracted micro-patches without any geometrical distortions can be successfully modeled as Gaussian i.i.d realizations with additive white noise. This makes analytical analysis of the rest of the processing chain possible, including dimensionality reduction and quantization.¹ Identification systems based on micro-structure fingerprints are elegant and fast.

Figure 2. The architectures for feature-based authentication (left) and fingerprint-based identification (right).

If the printed mark is not used or is of insufficient quality, and the user wishes to use a handheld consumer mobile device, authentication is still possible using computer vision features. This processing chain is more intensive, as it now has to handle all kinds of geometrical and lighting distortions. The sensitivity of image features must be adapted for micro-structures, as these images are by far not as visually rich as natural scene photos. Feature detection is followed by several steps that detect outliers and verify geometric consistency between a query and a database item. Our tests showed that sample authentication was possible under these conditions if a more complex architecture is acceptable.^{2, 3}

We used and distributed three data sets that are publicly available for testing and development.⁴ We acquired the forensic authentication microstructure optical set (FAMOS) industrial data set with high-quality cameras and controlled lighting. The two other data sets, FAMOS-M-L and FAMOS-M-S, contain samples acquired with a handheld mobile phone without special lighting. In both cases, our originating objects were consumer paper packages such as an ordinary medicine box.

We judged our results using two criteria: P_m, the probability of a miss, and P_fa, the probability of a false alarm. A false alarm occurs when a false or non-enrolled item is wrongly accepted by the system as genuine, and a miss constitutes the rejection of a genuine enrolled item. Ideally, both should be zero.

Identification results based on an extracted and synchronized mark on the industrial FAMOS set are shown in Figure 3. Unsurprisingly, we obtained the best results when the enrollment and verification cameras are identical. This architecture is fast and allows for deployment of advanced digital fingerprinting methods when acquisition specifications are met.

Figure 3. Receiver operating characteristic curves for all deployed cameras in the forensic authentication microstructure optical set (FAMOS) industrial dataset comparing extracted synchronized patches using cross correlation. P_fa: The probability of authenticating a non-enrolled or counterfeited item as genuine. P_m: The probability of a miss of an enrolled item.

Our results using a consumer mobile device, shown in Table 1, demonstrated that feature package identification is possible while using relatively small patches and preventing false matches from non-discriminative weak descriptors. Going forward, our research will focus on further developing robust features specifically tailored for micro-structures.