The deployment of Floating Offshore Wind Turbines (FOWT) in offshore sea conditions is accelerating rapidly. These turbines face ad- verse metocean conditions, necessitating continuous monitoring to ensure operational safety. Vibration-based monitoring is a promising approach for assessing structural health, but the complexity of FOWTs and their motion requires dense instrumentation. Extreme operating conditions further complicate the installation of traditional sensors. To overcome these challenges, recent research has focused on computer vision-based techniques for non-intrusive vibration sensing to infer structural health. Non-intrusive sensing offers significant advantages over intrusive meth- ods, as it avoids installation complications and minimizes interference issues. However, the scarcity of datasets for computer vision techniques poses a challenge. Physics-based Rendering (PBR) software, such as Blender, can simulate physical responses from 3D models in realistic environments, yet remains underexplored in structural health monitoring (SHM). In this study, the work is divided into two parts. Firstly, YOLOv8 (You Only Look Once) is employed to detect the motion of the entire structure and the turbine blades. In the second part, both YOLOv8 and BoT-SORT algorithms are used to non-intrusively track and detect the relative response of turbine blades to vibrations and to detect and track damage due to deformations. This second part of the work is currently underway, and the data has already been generated for further analysis. A dataset was generated using Blender and the algorithms successfully detected the overall motion of the structure and tracked the relative vibration response with promising accuracy. The proposed scheme effectively tracked both the overall motion of the structure and the relative response of the turbine blades, demonstrating excellent results.