LAMOST(大天区多目标光纤光谱天文望远镜)要求通过非接触测量快速准确地获得光纤端部的中心位置。在研究光斑中心提取技术的过程中，提出了一种改进的光重心方法，包括图像卷积预处理、用标准差迭代的方法取背景值和减背景后非线性加权获得光斑特征点。有效地抑制了边缘点对中心位置的影响，提高了在大视场多目标光纤条件下提取光斑特征点的精度，并通过数据模拟、实验室和现场实验进行了验证。

LAMOST(Large sky area multi-object fiber spectroscopy telescope) needs to get the center position of the fibers quickly and accurately through non-contact detecting. In the process of the research on detecting the position of the fibers，a new improved barycenter method is presented, which includes convolving with image、getting background value by iterating with standard deviation 、nonliner weighting after subtracting background value. This method restrains the edge points effectively and improves the precision under the condition of big view and large number of fibers . Also it is verified through the data simulation、laboratory and field experiments.