Computer aided surgery (CAS) often involves the use of instrument, implants and other interactive device whose pose is tracked in real time so as to provide navigated positioning of the instrument to guide the surgeon carrying out the operation. The infrared tracking system has been proved to be a valuable alternative to tracking system based on other technologies, such as magnetic, acoustic, gyroscopic and mechanical due to its precision and robustness. Often, markers attached to instruments for infrared tracking system are spherical and therefore not distinguishable, making identification targets from the points cloud necessary. Owing to the wear, tear or damage of the instrument in service, the relative position between the work point and the markers on instruments will change, which makes the instrument calibration before the application obbligato. This paper presents the effective algorithms for detecting the targets from the reconstructed 3D point cloud and estimating the relative position of work point to markers placed on the instrument. Both of them are essential in the procedure of pose estimation of surgical instrument.