Parallel Computer Vision: Person Data Extraction

Lang Christian
Fachhochschule Nordwestschweiz
Fachhochschule Nordwestschweiz, 2013


   title={Parallel Computer Vision: Person Data Extraction},

   author={Christian, Lang},



Download Download (PDF)   View View   Source Source   



Face recognition has been established in many environments these days. It is used in security systems, social media platforms or in digital cameras to support the user. In addition, the rapidly rising number of CPU cores in modern PCs or handhelds let us do more complex work on a single machine. The central question of this work is: Is it possible to create a system that can detect and recognises people in a video stream in real time and only with the resources of one PC, but with at least one GPGPU capable graphics adapter? To answer this question, such an application is developed with the use of C++, the computer vision library OpenCV and the GPGPU language CUDA from nVidia. To optimize the application for real time usage, the Concurrency Visualizer of Microsoft Visual Studio 2012 has been used. It is shown how to use difference images to calculate motion in videos and how to stabilize such motion areas with the use of a self-designed sweep line algorithm. In the first part of this master project, the technologies to create such software are evaluated and the first steps of video processing, motion detection and speed optimization are done.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2021 hgpu.org

All rights belong to the respective authors

Contact us: