Multi-View Reconstruction for Modeling Knee Rehabilitation Support Equipment

Knee osteoarthritis (KOA), one of the joints most commonly involved in osteoarthritis disease, has become a leading cause of physical disability in the elderly population, demonstrating a clear and growing public health need. Three-dimensional reconstruction of knee anatomy plays an important role in many biomedical fields such as ergonomics and sports rehabilitation. The real-time acquisition of human body parts is still a complex task due to the non-static and texture-free characteristics of the scanned objects. In this study, a binocular speckle optical acquisition device is designed for fast data acquisition. After capturing 8 images from various angles, a single-frame reconstruction technique based on the digital speckle correlation method was applied to generate a dense point cloud of the knee joint surface. The ICP point cloud registration algorithm was then employed to accurately stitch the 8 knee joint point clouds together, establishing a reliable model for the manufacturing of rehabilitation support equipment. This method can achieve the accuracy of plane reconstruction of 0.02mm and surface reconstruction of 0.09mm, which meets the application requirements.