MSc Diploma Project
The DMFCam – a dynamic mobile follow-cam
Goal of this study is the development of a supportive helmet-cam system for winter-sports environments. In a scenario of two persons riding down the slopes – one acting as camera man, the other as athlete – the system should be able to autonomously keep track of the athlete.
The project therefore consists of three main tasks. First is the buildup of a hardware system that enables auto-centring of the target person within the image area. This will be done by implementing a low-cost pan-tilt-unit and a webcam on top of it to be mounted on a sports-helmet. Second is the people-tracking algorithm. As speed is more important than precision in this scenario two algorithms will be tested on performance. One is “Color Blob Flow” from Daimler Benz’ traffic systems based on color blobs, the other is “FragTrack” based on the integral histogram data structure. Third the results of “Color Blob Flow” and “FragTrack” are taken into account to implement a decision and motion-estimation algorithm. This will be used to predict the target’s position in the upcoming frame to counter-act latencies of the pan-tilt-unit and the camera system.
The implementation will be done using C++ in a linux environment with Qt, OpenCV, OpenGL and sources from the robotic framework “Robbie 13” of the Active Vision Group, University of Koblenz-Landau.
- Heisele, B. & Ritter, W.: Obstacle Detection Based on Color Blob Flow, Proceedings of the Intelligent Vehicles Symposium, 1995
- Adam, A. & Rivlin, E. & Shimshoni, I.: Robust Fragments-based Tracking using the Integral Histogram, Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition – Volume 1, Pages 798-805
- Porkili, F.: Integral histogram: A fast way to extract histograms in cartesian spaces. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2005.