Research on an automated system for precise capture of aerobics movement trajectories using video analysis technology

With the rapid advancement of artificial intelligence and computer vision technologies, the digitalization and intelligentization of sports education and training have been continuously enhanced. This paper designs and implements an automated system for precise capture of aerobic exercise trajectories based on video analysis technology. The system uses monocular video as input and achieves end-to-end processing from raw video to accurate trajectory through multi-module collaboration: human detection and tracking algorithms integrating spatiotemporal information complete target localization, improved OpenPose combined with Gaussian softening strategy ensures stable extraction of skeletal key points, TRAM framework core technology enables 3D trajectory reconstruction, and RoMo hybrid solver optimizes trajectory accuracy. Experiments on the aerobic subset of MultiSports dataset demonstrate that the system achieves correlation coefficients (CMC) of 0.93±0.05 for hip, 0.97±0.02 for knee, and 0.86±0.08 for ankle joints, with an average error of only 6.8% compared to standard trajectories—significantly outperforming traditional methods. This system meets the precision requirements for aerobic training evaluation, movement optimization, and competition analysis, providing technical support for teaching, training feedback, and intelligent assessment in aerobic education.