Abstract: In order to reconstruct 3D medical data model more accurately，this paper converts the pixel on 2D contour line of medical image to 3D point cloud and exploits the Poisson surface reconstruction technique.Due to the reconstruction quality heavily relying on the accuracy of the normal，and according to gradients of the contour line data，the problem of assigning consistently oriented normal vectors to the points is addressed，and the quality of the normal is improved via an orientation-aware principle component analysis step.As a popular implicit reconstruction technique，Poisson surface reconstruction technique can handle the noise and the holes in medical image data very well.At the same time，the reconstruction results of the upper airway model demonstrate that this normal consistently oriented method improves the traditional methods and produces more accurate reconstruction of the medical model.

Key words: contour, poisson surface reconstruction, consistently oriented normal vector

摘要： 为了更准确地重建复杂的三维医学数据模型，把二维医学图像轮廓线上的像素点转化为三维点云，引入经典的泊松点云重建技术。由于泊松重建的效果依赖于点云法向的准确性，针对轮廓线数据的特点，结合图像二维梯度方向，给出了点云法向的一致定向及基于已知定向进一步精确估计法向的方法。泊松重建方法作为一种隐式重建技术，可以很好地处理医学图像数据中经常存在的噪声、拓扑复杂等问题。鼻咽喉模型重建的结果说明，一致定向方法改进了传统的一致定向方法，得到了更为准确的医学重建模型。

关键词: 轮廓线, 泊松重建, 法向一致定向