Hybrid vision-based motion capture for a human leg

Gargallo Vazquez, Miguel

4. term

Vision, Graphics and Interactive Systems, Master

2017

2017-09-08

30

Dette projekt befinder sig i krydsfeltet mellem computervision og computergrafik og fokuserer på motion capture (registrering af menneskelig bevægelse). To udbredte metoder er markørbaserede systemer, som sporer små markører fastgjort på kroppen, og markørløse systemer, som estimerer bevægelser ud fra kamerabilleder uden markører. Hver metode har fordele og ulemper: markørbaserede løsninger er ofte mere præcise, men kræver tidskrævende opsætning, mens markørløse løsninger er mere bekvemme, men mindre nøjagtige. Vi undersøger, hvordan en hybrid tilgang kan kombinere dem for at bevare en enkel opsætning og samtidig fastholde brugbar nøjagtighed. Vi præsenterer den nødvendige baggrund og en praktisk metode til at vælge, hvordan data fra begge systemer bedst kombineres. For at afprøve ideen gennemførte vi et forenklet forsøg, hvor et menneskeligt ben bevæger sig og går langs en lige linje. Bevægelsen blev optaget med et videokamera og et markørbaseret system med markører på benets segmenter. Derefter kombinerede vi segmentoplysningerne fra begge kilder på flere måder, beregnede knæ- og ankelvinkler og sammenlignede resultaterne med vinkler beregnet ud fra det fulde markersæt.

This project sits at the intersection of computer vision and computer graphics and focuses on motion capture (the process of recording human movement). Two common approaches are marker-based systems, which track small markers attached to the body, and markerless systems, which estimate motion from camera images without markers. Each has trade-offs: marker-based methods are often more precise but require time-consuming setup, while markerless methods are more convenient but less accurate. We explore how a hybrid system could combine them to keep the setup simple while preserving useful accuracy. We provide the background and a practical method for choosing how to mix data from both systems. To test the idea, we built a simplified experiment in which a person’s leg moves and walks along a straight line. We recorded the motion with a video camera and with a marker-based system using markers on the leg’s segments. We then combined the segment information from both sources in several ways, calculated knee and ankle angles, and compared these results to angles computed using the full set of markers.

[This abstract was generated with the help of AI]

motion capture ; hybrid ; tracking ; markerless ; computer vision

Download
View record in AAU Student Projects