Security thread detection in passport using improved template matching

In recent years, passport has been paid more and more attention. Passport is not only a certificate of the passport holders, but also involves the international anti-terrorism situation. Passport security thread, as a security feature which is the most direct and easy to identify, is generally used by national passports. However, in the passport manufacturer, the current inspection method of the passport security thread is manual inspection. Computer vision can be applied in this aspect for automatic inspection through some systems or machines. This paper proposes that a custom-built computer vision system can utilize the reflected light to collect images, and detect the buried security thread with the relative high accuracy. After analysis, the detection of the security thread can be considered as a class of the object detection. On account of the gorgeous page’s pattern around the security thread in passport, the most of object detection algorithms are failed to complete this task. Taking both accuracy and detection speed into account, in this work, we develop an improved algorithm based on the traditional SURF operator to achieve security line detection in passport. After verification on a sample set containing 134 samples, this approach has been a certain ability to detect the security thread with the accuracy of 84.33%.