BACKGROUND Landmark-based approaches of two- or three-dimensional coordinates are the most widely used in geometric morphometrics (GM). As human face hosts the organs that act as the central interface for identification, more landmarks are needed to characterize biological shape variation. Because the use of few anatomical landmarks may not be sufficient for variability of some biological patterns and form, sliding semi-landmarks are required to quantify complex shape. RESULTS This study investigates the effect of iterations in sliding semi-landmarks and their results on the predictive ability in GM analyses of soft-tissue in 3D human face. Principal Component Analysis (PCA) is used for feature selection and the gender are predicted using Linear Discriminant Analysis (LDA) to test the effect of each relaxation state. The results show that the classification accuracy is affected by the number of iterations but not in progressive pattern. Also, there is stability at 12 relaxation state with highest accuracy of 96.43% and an unchanging decline after the 12 relaxation state. CONCLUSIONS The results indicate that there is a particular number of iteration or cycle where the sliding becomes optimally relaxed. This means the higher the number of iterations is not necessarily the higher the accuracy.

中文翻译：

---

对 3D 人脸图像滑动半地标迭代效果的新颖研究。


背景技术基于地标的二维或三维坐标方法在几何形态测量学（GM）中应用最广泛。由于人脸的器官充当识别的中心界面，因此需要更多的标志来表征生物形状的变化。由于使用少量解剖标志可能不足以满足某些生物模式和形式的变化，因此需要滑动半标志来量化复杂的形状。结果本研究研究了滑动半地标迭代的影响及其结果对 3D 人脸软组织 GM 分析预测能力的影响。使用主成分分析（PCA）进行特征选择，并使用线性判别分析（LDA）预测性别，以测试每种松弛状态的效果。结果表明，分类精度受迭代次数影响，但不呈递进模式。此外，在 12 次松弛状态下保持稳定，最高准确度为 96.43%，并且在 12 次松弛状态后下降不变。结论 结果表明，存在特定次数的迭代或循环，其中滑动变得最佳松弛。这意味着迭代次数越多并不一定精度越高。