Non-facing ID card identification

Jul.2016~Dec.2016

This module is matching a hologram is positioned correctly. The hologram is extracted from an ID card captured by a user. I was in charge of all areas of development in this project. The user places his or her ID card on a white background and takes a video so that whole the ID card comes out with the camera flash turned on within one to three seconds. Before normalizing the ID card, Detecting the boundary of your ID card first. In case, failed to locate the boundary of the ID card, the system interpolates using some of the bunch boundaries to obtain the ID card boundary. Developed using applied PCA, this system corrects the card even when the card is tilted. Extract holograms separated from light sources by comparing dynamic saturation and intensity with surrounding pixels. Using the six template matching methods, compare the hologram obtained with the template image. So, compared to other hologram cards, the system has been able to detect ID cards with high probability.

 

System flow diagram: normalizing the location of an ID card with a video of an ID card taken for approximately 3 seconds, and extracting the hologram of an ID card. Pixel matching between extracted the hologram and the template image ensures identification is determined.