Background textures analysis for counterfeit detection
Oriol Ramos Terrades – Computer Vision Center, Universitat Autònoma de Barcelona (Spain)
Abstract: Counterfeiting and piracy has a damaging effect on business, economy and population in general. To fight against them, governments and authorities cooperate and develop security features to protect their security documents. In this talk we will first review texture-based features on machine learning algorithms and their performance to detect banknote and identity counterfeit documents. We will briefly summarize the main algorithms used for this purpose and provide some results on benchmark datasets. Then, we will move to more recent architectures using deep learning architecture, and more specifically, using recurrent neural networks. We will outline the main components of the proposed architectures and their performance. We will conclude the talk with some open questions and challenges in real scenarios.
Dr. Oriol Ramos obtained his PhD in Computer Science in 2006. He is currently associated professor in the Computer Science Department at the UAB and an active researcher within the Document Analysis group in the Computer Vision Center. He mainly teaches database systems for undergraduate students and probabilistic graphical models in the Computer Vision master. His research interests are inference and learning of probabilistic graphical models (PGMs) and their application to document image analysis, document forensics and symbol retrieval, among others, co-authoring more than 70 papers published in international conferences and journals.