Robust Stereo-Match Algorithm for Infrared Markers in Image-Guided Optical Tracking System.

Qinyong Lin,Rongqian Yang,Zhesi Zhang,Ken Cai,Zhigang Wang,Meiping Huang,Jinhua Huang,Yinwei Zhan,Jian Zhuang

DOI: https://doi.org/10.1109/ACCESS.2018.2869433

IF: 3.9

2018-01-01

IEEE Access

Abstract:The stereo-match to find the corresponding points in the left and right images from an optical tracking system is a crucial step in 3-D point reconstruction. Unfortunately, the presence of ghost markers remains an issue which is caused by indistinct correspondence when several markers are coplanar with two optical centers. Such ghost markers can lead to erroneous tracking of medical instruments and patients. Some methods have been proposed for this issue, but they are rarely applied in tracking systems. To address the issue, we propose a robust stereo-match method for combining the maximum silhouette reprojection consistency with the epipolar constraint into an objective function. For minimization of the objective function, we reconstruct a virtual sphere through optimal triangulation and construct the 3-D distance errors between the inverse projective rays of the marker contour and the reprojection contour on the virtual sphere surface. A contour-detecting method is proposed to obtain the accurate contours of marker blobs in images. Simulative and real experiments are performed to evaluate the performance of the proposed stereo-match method. We compare the proposed method with other methods performed on several real data sets which are acquired from a human phantom. The experimental results demonstrate that the proposed method is more robust and efficient in searching the correct correspondences of the markers.