3D COORDINATES DIGITIZATION OF MOTION CAPTURE SYSTEM ON BADMINTON ATHLETE
The increase of computation ability also pushes the analysis ability on many fields, one of them is sport science. The motion capture system is one of the many tools for quantifying motions of an athlete. With such ability, we can get deeper understanding of an athlete’s performance. Existing motion...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/43120 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The increase of computation ability also pushes the analysis ability on many fields, one of them is sport science. The motion capture system is one of the many tools for quantifying motions of an athlete. With such ability, we can get deeper understanding of an athlete’s performance. Existing motion capture system that are available in the market has a relatively high cost. Therefore, it is hard to be implemented in Indonesia. This led to a need of a system that provides the same ability, but with the use of consumer grade cameras and open source software. This final project will discuss methods of digitization function in a motion capture system which use five consumer grades cameras and open source software. To reconstruct the 3D point of a marker, the cameras used must be calibrated to get the extrinsic and intrinsic parameter of the camera. This system will be able to do such calibration and put the parameters in the form of DLT coefficients. Intrinsic calibration uses the grid pattern method. While the extrinsic calibration uses the wand and Bundle Adjustment method. Direct Linear Transformation (DLT) method is used to perform the linear algebra of the camera matrix. This system was implemented on two different time data acquisition with the total of 8 badminton athlete doing lunges movement. The intrinsic calibration returns the error value of 4 pixels, while the extrinsic calibration returns the error value of 5 to 90 pixels. The 3D reconstruction has error has the value of 7.29 cm and 6.9 cm. |
|---|