Background Segmentation Methods in Analysis of Live Sport Video Recordings
(2018) In Master's Theses in Mathematical Sciences FMAM05 20181Mathematics (Faculty of Engineering)
- Abstract
- A sports video analysis application was developed by Spiideo for mobile devices. It presents recordings from practices and competitive games for game-play analysis. Available tools, such as on-screen drawings, require a robust background foreground segmentation. The currently used segmentation method have difficulties to master the shifting conditions in weather, shadows and shirt colors.
The purpose of this project was to improve the background foreground segmentation in the application by evaluating alternative methods and examine possible improvements of the currently used method.
A data-set with recordings from Spiideo clients was created and used to evaluate and compare different segmentation methods with the current method.... (More) - A sports video analysis application was developed by Spiideo for mobile devices. It presents recordings from practices and competitive games for game-play analysis. Available tools, such as on-screen drawings, require a robust background foreground segmentation. The currently used segmentation method have difficulties to master the shifting conditions in weather, shadows and shirt colors.
The purpose of this project was to improve the background foreground segmentation in the application by evaluating alternative methods and examine possible improvements of the currently used method.
A data-set with recordings from Spiideo clients was created and used to evaluate and compare different segmentation methods with the current method. The evaluation included pixel classification scores, complexity measurements and a visual evaluation.
A median background model frame difference approach showed better performance with lower computational time compared to what is currently used. A Mixture of Gaussians method gave the best pixel classification result, but increased the calculation time. Suggested alterations of the currently used method also showed minor improvements in performance. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8948473
- author
- Hammar, Fredrik LU and Flinke, Johan LU
- supervisor
- organization
- course
- FMAM05 20181
- year
- 2018
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Image Analysis, Background Foreground Segmentation, Segmentation Evaluation, Live Sports Video Analysis
- publication/series
- Master's Theses in Mathematical Sciences
- report number
- LUTFMA-3358-2018
- ISSN
- 1404-6342
- other publication id
- 2018:E40
- language
- English
- id
- 8948473
- date added to LUP
- 2018-06-19 16:21:08
- date last changed
- 2018-06-19 16:21:08
@misc{8948473,
abstract = {{A sports video analysis application was developed by Spiideo for mobile devices. It presents recordings from practices and competitive games for game-play analysis. Available tools, such as on-screen drawings, require a robust background foreground segmentation. The currently used segmentation method have difficulties to master the shifting conditions in weather, shadows and shirt colors.
The purpose of this project was to improve the background foreground segmentation in the application by evaluating alternative methods and examine possible improvements of the currently used method.
A data-set with recordings from Spiideo clients was created and used to evaluate and compare different segmentation methods with the current method. The evaluation included pixel classification scores, complexity measurements and a visual evaluation.
A median background model frame difference approach showed better performance with lower computational time compared to what is currently used. A Mixture of Gaussians method gave the best pixel classification result, but increased the calculation time. Suggested alterations of the currently used method also showed minor improvements in performance.}},
author = {{Hammar, Fredrik and Flinke, Johan}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master's Theses in Mathematical Sciences}},
title = {{Background Segmentation Methods in Analysis of Live Sport Video Recordings}},
year = {{2018}},
}