Abstract: This paper employs the least square registration method to achieve the high precision patient registration in the image guided surgery. First of all, a laser tracker and a probability statistic method are used to calibrate the electromagnetic positioning system, which can acquire the realistic coordinates of marked points. Then, the puncture needle calibration adapts a spatial multi-gesture method to collect the positions of needle points through the positioning system, and applies the theories of homogeneous coordinate and quaternion to solve the coordinate transformation problem between the real space and virtual space. Finally, an operation standard of the patient registration is proposed for surgeons. The experimental results tested with the surgical robot indicate that the precision of the registration has reached 0.0402 mm. Therefore, the registration algorithm can meet the clinical requirements of image guided surgery.