中文摘要：

提出了一种基于平面投影不变量的目标跟踪算法．算法从图像中提取直线边缘计算投影不变量，用于对目标建模并跟踪．为提取直线边缘，使用改进的序列细化算法将边缘细化为单像素宽，而后用一种快速曲率估计方法估算边缘点的曲率，并保留估算值很小（约等于零）的点拟合直线．在所得直线族中按照邻近规则或者窗口规则挑选直线计算投影不变量．图像处理实验给出了用文中提出的图像预处理算法获得的直线边缘效果，并通过使用所得直线计算不变量的值衡量了所得不变量的稳定性和视角不变性．跟踪实验检验了跟踪算法的鲁棒性和实用性．

英文摘要：

An object tracking algorithm based on planar projective invariants is proposed, in which the values of invariants for object modeling and tracking are calculated with lines extracted from images. In order to attain the line edges in images, all edges are thinned by improved sequential thinning algorithm to be one pixel wide at first. Then a fast computing method is presented to estimate the curvature of edge points and the points at which curvature is less than some threshold are collected to fit lines. Lines are selected by neighborhood-rule or window-rule to compute projective invariants. In the image processing experiments, the straight lines are extracted with the image preprocessing algorithm proposed in this paper, and the stability and viewpoint-invariance of the invariants calculated by those lines are assessed. The tracking experiments verify the robustness and practicability of the algorithm.