Three-Dimensional Pose Estimation of Optically Transparent Microrobots

摘要

The use of microrobots for cell manipulation has a range of applications in biomedical research. Direct sensing of microrobot three-dimensional (3D) position and orientation, however, is practically challenging due to the small scale involved. This letter proposes a vision-based method for estimating the 3D pose of optically transparent microrobots based on Optical Tweezers (OT) manipulation by using Convolutional Neural Networks (CNNs).& x00A0;A model-based approach is used to generate the large training set required for CNNs. Detailed validation is performed to demonstrate the experimental use of the technique.