基于二维主分量分析的成趟足迹特征提取方法

METHODS FOR EXTRACTING CHARACTERS OF CLUSTER-FOOTPRINTS BASED ON 2DPCA

本文采用图象图形学的方法提取成趟足迹的特征值.首先,定义成趟足迹中左侧相邻单足足迹质心间的连线为左步行线,右侧相邻单足足迹质心间的连线为右步行线.第二,采用2D主分量分析(PCA)提取成趟足迹中每个单足足迹沿足长方向的特征向量.定义左步行线上足长特征向量与左步行线间在前进方向的夹角为左步角,右步行线上足长特征向量与右步行线间在前进方向的夹角为右步角.第三,设计了步行线,步长,步宽,步角的提取方法.对硬地地面成趟足迹样本所做的实验结果表明:对于经过同样预处理所得到的同一幅成趟足迹二值图象,提取的特征值是唯一的.本文提出的成趟足迹提取方法是可行的.

The image and graphic method is used to extract the cluster-footprint characters. First, in the cluster-footprints,the ligature linking one centroid and the other of two adjacent single-foot- footprints on the left is defined as the left-walk-line,and the ligature linking one centroid and the other of two adjacent single-foot-footprints on the right is defined as the right-walk-line. Second, 2D Principal Component Analysis （PCA） is made use of to extract eigenvectors along foot-length of every single-foot-footprint in the cluster-footprints. The angles made of the foot-length eigenvectors on left- walk-lines with the left-walk-lines are defined as the left-walk-angles, and the angles made of the foot-length eigenvectors on the right-walk-lines with the right-walk-lines are defined as the right- walk-angles. It should be noted that the walk-angles are in the forward direction of the walk-lines. Third, the methods for extracting walk-lines, walk-lengths, walk-widths, and walk-angles are designed. Results of experiments with the sample of the cluster-footprints on hard ground indicate that the characters to be extracted are unique for a boundary image which is obtained with use of the same reprocessing method. The methods for extracting the cluster footprint characters proposed in this paper are applicable.