Video Clip Assignment Collusion Detection

Document Sample
Video Clip Assignment Collusion Detection Powered By Docstoc
School of Electronic, Communication & Electrical Engineering
Learning and Teaching Development Fund BLU-SKY Award Report

Video Clip Assignment Collusion Detection

Aims of the Project:
Plagiarism and collusion phenomena are increasing in the recent years. On
the market, there is a wide range of software for detecting collusion. However,
they all are text based detection. With the growing need to assess students’
multimedia assignments, and the growing number of students, it becomes
essential to have a tool capable of detecting collusion/plagiarism in
multimedia coursework.

The main aims of the project are summarised below:
   - Develop a user-friendly software, which detect collusion of video clip
   - The batch submission of all video documents to be cross-checked.
   - The developed software will allow a cross comparison of all the images
     embedded in the submitted video clips
   - The result will graphically indicate the clips which are colluded and the
     location inside the video clip.

Work undertaken and outcomes achieved
Text-based detection process carries out electronic comparison of students'
work against electronic text sources from the internet and other students. With
the growing demand of multimedia courses, it becomes necessary to detect
plagiarism on the students’ multimedia coursework. This project develops the
software tool for identifying collusion on assessed video clips. The detection
method which is described here addresses only collusion, in which student A
submits an assessed video clip wholly or partly copied by student B, and
submits it as his/her own. For this reason, the underlying assumption is that
all of the source material that need be examined is directly available in the
form of the students' submissions.

The software project has been divided into three distinctive parts: file
preparation, processing of the detection algorithm and display.

Due to the large amount of data contained in a video clip, it is essential to
retain a fraction of the information for each frame also called as features.
Instead of performing frame comparison between frames from different
sources, a block preparation process takes place by dividing a frame into a
set of B  B blocks. The initial feature extraction approach is to only retain the
average for each block for all the frames in the video clip. This process is
replicated for all the other frames of the other clips. Then, the absolute values
of the difference between corresponding features from two video clips are
extracted. If the difference is below is pre-defined threshold then the frames
are identical. It means that collusion can be detected. Throughout the
computing, all collusion offences are recorded in a collusion log table. The
School of Electronic, Communication & Electrical Engineering
user will be able to visualise the video collusion offences on two separated
video monitors.

Transfer to other areas
The outcome of the project is an user–friendly software, which can be used by
lecturers for detecting collusion of multimedia pieces of assessment. The
software can be targeted to disciplines across the faculties:
- Studio Practice and Content Creation              School of Electronic,
- Multimedia System Design & Implementation Communication and
- Advanced Digital Broadcast Systems                Electrical Engineering
- Multimedia Authoring
- Advanced Multimedia & Interactive Systems          School of Computer Sciences
- Digital and Lens Media Video And Sound
                                                     School of Film Music and Media
- Moving Images

This tool can also be used to deter collusion and be presented to students in
the above programmes.

So far, the video clip collusion detection process and the file preparation
process have been implemented. The project is on schedule to be submitted
to the BLU team in August 2007. Because the video collusion detection tool is
still being developed, no thorough evaluation analysis can be provided in this

Dissemination of the project outcomes
The collusion detection algorithm of the project has been presented at the 2nd
International Blended Learning Conference held at the University of
Hertfordshire in June 2007. The complete tool will be presented at the
Engineering and Information Science Faculty Learning and Teaching
Conference this September.


Herbland, A. & Siau, J. “Student’s Video Clip Collusion Detection: the next
generation of collusion detection”, 2nd International Blended Learning
Conference, University of Hertfordshire, June 2007.

Herbland, A. & Siau, J. “Offline Student’s Video Clip Collusion Detection
Tool”, EIS Faculty Learning and Teaching Conference, University of
Hertfordshire, September 2007. [to be published]
rtfordshire, September 2007. [to be published]

Shared By: