The objective of the FIVES project is to develop novel investigative
tools specifically tailored for investigations involving images and
videos of child sexual abuse. The FIVES tool set will allow police and
law-enforcement agencies to
a) speed up the process of handling very large amounts of evidence
material on seized computers, and separate previously known illegal
material from new, potentially illegal, material by efficient file and
file fragment matching.
b) efficiently evaluate large amounts of new material by employing
perceptual optimization techniques. This aims to minimize the human
effort needed when classifying new material.
c) improve the capability of linking new illegal images and video
to previously known material by using object matching and image
similarity techniques to allow details of crime scenes to be linked
between different image sets or videos. This facilitates the widening of
investigations with the aim of rescuing the victims of sexual abuse.
The tools will be based on already existing research and software
created by the academic partners that will be adapted, extended and
integrated into an easy to use tool set that fits the police
requirements. The project's technical work packages will create a
forensic engine that provides base functionality, and a number of
modules. One module will provide new file fragment matching
functionality, and others will provides specialized image and video
handling functionality to support police work. There are also work
packages for performing a user requirements study and end-user field
tests as well as for ensuring sustainability of project results after
the end of the project.
Fast Classification of Indecent Video by Low Complexity Repetitive Motion Detection
T. Endeshaw, J. Garcia, and A. Jakobsson (Karlstad University)
This paper proposes a fast method for detection of indecent video content using repetitive movement analysis.
Unlike skin detection, motion will provide invariant features irrespective of race and color.
The video material to be evaluated is divided into short fixed-length sections.
By filtering different combinations of B-frame motion vectors using adjacency in time and space,
one dominant motion vector is constructed for each frame. The power spectral density estimate of
this dominant motion vector is then computed using a periodogram. The resulting power spectrum is
then subjected to a selection window to restrict the spectrum to an limited frequency range typical
of indecent movement, as empirically derived by us. A threshold detector is then applied to detect
repetitive motion in video sections. However, there are many instances where repetitive motion occurs
in these shorter sections without the video as a whole being indecent. As a second step, an additional
detector is employed to determine if the sections over a longer period of time can be classified as as
having indecent material. The proposed method is resource efficient not requiring the IDCT step of the
video decoding. Evaluations performed using a restricted set of videos with different amounts of texture,
lighting conditions and complex backgrounds show very promising results.
|