传统的部分Ｈａｕｓｄｏｒｆｆ 距离，由于采用固定的分位数，在图像匹配应用中具有相当的局限性。运用非参数和参数统计手段，从模板距离图计算距离变量的经验分布函数，充分挖掘模板中有关像元分布结构的信息，设计出一种稳健的自适应Ｈａｕｓｄｏｒｆｆ 距离计算法。分析表明，新的距离计算式对分位数初值的敏感程度较传统的部分Ｈａｕｓｄｏｒｆｆ 距离小得多，在强噪声和复杂背景下具有很好的抗差性能。 The traditional partial Hausdorff distance, due to the fixed quantile, has considerable limitations in image matching applications. Using nonparametric and parametric statistical means, the empirical distribution function of distance variables is calculated from the template distance map, and the information about the pixel distribution structure in the template is fully tapped to design a robust adaptive Hausdorff distance algorithm. The analysis shows that the new distance calculation formula is more sensitive to the initial quantile than the traditional part Hausdorff distance, and has good robustness against strong noise and complex background.