Digitization and automatic processing of incoming paper mail is a crucial component of document management systems and is widely adopted in medium and large enterprises. Besides several advantages of this automated processing, it complicates the first line inspection of incoming documents which are often of vital financial or legal relevance. We have developed a number of different techniques to allow automatic detection of forged or manipulated documents over the last years. In this paper, we present an analysis of the application of our methods on a large real-world dataset of invoices to identify the weaknesses of our existing methods and propose some promising directions of future work in the field.
展开▼
