基于特征点的颅面复原技术

介绍了用人工神经网络算法确定待复原颅面的软组织厚度及颅面特征点的方法。在计算机辅助颅面复原过程中,通常只能根据颅骨上少数特征点及其所在部位的法向软组织厚度的统计值进行运算。由于现有统计值指标涵盖的年龄较宽泛,导致复原结果缺乏个性。在测定待复原颅骨骨龄的基础上,通过神经网络训练,确定软组织厚度随年龄的变化规律,较精确地计算出与拟复原对象的骨龄相适应的特征点软组织厚度值,并用最小二乘法及曲面法线原理建立数学模型,计算出颅面特征点。最后用基于Delaunay三角剖分的wavelet插值法复原出面貌雏形。

Neural network algorithm in application of facial reconstruction

Introducing a method to determine facial tissue thickness and facial landmarks using the algorithm of artificial neural network.In the process of computer aided facial reconstruction,the facial landmarks usually can be calculated by a few of landmarks on a skull and their corresponding tissue thickness in the normal direction.The existing data of thickness are statistic by experts,and each index covers a wide range of age.This results to such a fact that the effect of the reproducing face is lacking of individual features to a certain skull.Based on this,a skull of bone age is measurable,through the training of a neural network,the regulation of tissue thickness changed with the variance of age can be confirmed;and the relevant tissue thickness can be relatively approximate.Finally,according to least-square method and the normal surface theory,a mathematics model is build up,and facial landmark is obtained.And last the facial surface is reconstructed using wavelet interpolation based on Delaunay triangulation.