摘要
用图像处理和计算机视觉技术研究了脑外科立体定向手术中标记点的自动检测与定位,提出了下述检测与定位方法:首先为提高检测准确率设计A、B两类具有特殊纹理的标记,然后,检测B类和A类标记点并基于极线约束搜索缺失的A类标记点,最后进行标记点的双目匹配和三维重建。通过实验测试了该方法的检测准确率和定位精度。结果表明,该方法的标记点误检率为0.08%,漏检率为零,每对图像的平均处理时间小于200ms,标记点的平均定位误差小于0.55mm。该方法简化了脑外科立体定向手术的图像注册过程,而且通过跟踪头部标记点可以在线更新注册。
epipolar constraint. Finally, match the detected markers in the binocular camera images and calculate the coordi- nate of markers. The detection rate and localization accuracy of the approach were tested by the experiment. The results showed that the markers' false rate was 0.08%, the markers' lost rate was zero, the average processing time per pair of images was less than 200ms, and the markers' average localization error was tess than 0.55mm. This approach can simplify the registration procedure of stereotactic neurosurgery and continuously update the registration data by tracking the markers attached to the head.
出处
《高技术通讯》
CAS
CSCD
北大核心
2013年第3期270-276,共7页
Chinese High Technology Letters
基金
863计划(2009AA045300-4)资助项目
关键词
人工标记点
图像注册
极线约束
脑外科立体定向手术
artificial markers, image registration, epipolar constraint, stereotactic neurosurgery