DEVELOPMENT OF A SIMPLE COLOR MARKER-BASED MOTION CAPTURE SYSTEM USING DETECTION AND TRACKING

• Main Author: Yahya Muhammad, Syihabuddin
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75494
• Institution: Institut Teknologi Bandung

Recently, Artificial Intelligence (AI) technology has gained popularity in motion capture systems. AI-based motion capture systems offer fast processing times and do not require additional equipment, making them highly sought after. However, many people believe that AI-based motion capture systems have lower accuracy compared to traditional motion capture systems. To address this issue, a simple AI- based motion capture system and a traditional simple motion capture system will be developed to compare their accuracy. The development of the traditional simple motion capture system requires precise object tracking and object detection processes to achieve maximum accuracy. The constructed motion capture system will utilize multiple cameras, and from each camera, videos will be captured to track the 2D positions of color-coded markers on the actor's body. These positions will then be further processed to obtain 3D coordinates in the world coordinate system. The focus of this study is to build accurate object tracking and object detection processes to achieve maximum accuracy in the motion capture system. Based on the conducted experiments, the optimal distance between the actor and the camera is found to be between 3 and 5 meters. Additionally, the detection and tracking errors of the 2D marker positions cannot be used as performance indicators for the system since consistent errors in each video result in correctly estimated 3D marker positions.