We propose a novel characterization of surfaces implemented using the macro-enabled cameras of modern smartphones and artificial intelligence algorithms. Each object is characterized and identified by a unique fin-gerprint of the surface's optical characteristics of the object. The technology is primarily designed for paper but suitable for all non-reflective surfaces such as cardboard, plastic and numerous metals. For the description of the surface characteristics, key points are first determined using local feature descriptors like SIFT, ORB or feature descriptors from neural networks. For each key point, feature vectors that are as invariant as possible are then determined, which together describe the fingerprint of the surface. A supervised classification ap-proach is used to identify the individual fingerprint from a database. This approach was evaluated in practice using a label-printing machine and an industrial camera with similar optical properties of macro cameras of current high-quality smartphones. It could be shown that a single label can be identified from a set of 100,000 identical printed labels using the presented approach.