Exposing forgeries in soccer images using geometric clues

摘要

In this study, new algorithms are proposed for exposing forgeries in soccer images. We propose a new and automatic algorithm to extract the soccer field, field side and the lines of field in order to generate an image of real lines for forensic analysis. By comparing the image of real lines and the lines in the input image, the forensic analyzer can easily detect line displacements of the soccer field. To expose forgery in the location of a player, we measure the height of the player using the geometric information in the soccer image and use the inconsistency of the measured height with the true height of the player as a clue for detecting the displacement of the player. In this study, two novel approaches are proposed to measure the height of a player. In the first approach, the intersections of white lines in the soccer field are employed for automatic calibration of the camera. We derive a closed-form solution to calculate different camera parameters. Then the calculated parameters of the camera are used to measure the height of a player using an interactive approach. In the second approach, the geometry of vanishing lines and the dimensions of soccer gate are used to measure a player height. Various experiments using real and synthetic soccer images show the efficiency of the proposed algorithms.