Geodesic Active Contours and Level Sets for the Detection and Tracking of Moving Objects
IEEE Transactions on Pattern Analysis and Machine Intelligence
Shape Priors for Level Set Representations
ECCV '02 Proceedings of the 7th European Conference on Computer Vision-Part II
Real-Time Tracking Using Level Sets
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 2 - Volume 02
International Journal of Robotics Research
Real-Time Imaging - Special issue on multi-dimensional image processing
A Target Dependent Colorspace for Robust Tracking
ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 03
Modeling Magnetic Torque and Force for Controlled Manipulation of Soft-Magnetic Bodies
IEEE Transactions on Robotics
Wide-angle localization of intraocular devices from focus
IROS'09 Proceedings of the 2009 IEEE/RSJ international conference on Intelligent robots and systems
Hi-index | 0.00 |
Successful ophthalmic surgeries using intraocular untethered microrobots or tethered robotic microtools require methods to robustly track the microdevices in the posterior of the human eye. The dimensions and specularities of the microdevices are major obstacles for accurate tracking. In addition, the optical structure of the human eye makes it challenging to keep the objects of interest constantly in focus, resulting in blurred images. In this paper, the advantages of using different colorspacaes for intraocular tracking are examined. After selection of the appropriate colorspace, thresholds that ensure maximum separation of the device from the background are calculated. Based on trained color histograms, level sets are used to track in real time, and the use of statistical shape information is incorporated in the existing tracking framework. The efficacy of the algorithm is demonstrated by tracking a microrobot in a model eye, using a custom made ophthalmoscope and off-the-shelf opthalmoscopy lenses. With the appropriate colorspace and threshold selection, tracking errors are minimized and are further diminished using shape information.