Neural network-based system for automatic passport stamp classification

Abstract

The international tourism growth forces governments to make a big effort to improve the security of national borders. Protecting the borders from illegal immigrants and simplifying border checkpoints for law-abiding citizens and visitors is a delicate compromise. In the era of speed, it is indispensable to analyze passport pages by an intel­ligent application that recognize and classify stamps of travel documents in order to ensure faster, safer and more efficient stamp controlling. This paper proposes a model of such a system based on artificial neural network. The major contribution of this paper is about its topic related to passport stamps not yet addressed in the literature of object detection and recognition. As main aim, we proposed a framework that performs detection and classi­fication to assist the border control. To the best of our knowledge, this is the first classification method for passport stamps. The originality of the proposed system based on low-cost neural network concerns several axes; the robustness in unconventional contexts, the high speed compared to other techniques such as the convolutional neural network, the low computational complexity with the help of using a classic classifier, the simplicity using intelligent tools guaranteeing the efficiency explained by promising accuracies with maximum accuracy of 0.945, and the high reliability explained by other classification metrics such as precision, recall and F1-score.