Because the existing ellipse fitting methods have problems such as poor anti-interference ability and poor extraction accuracy when extracting the spot centroid under atmospheric turbulence disturbance, it is difficult to improve the servo tracking accuracy of the laser communication system. To this end, this paper proposes a fast spot centroid extraction method based on ellipse fitting. First, according to the degree of turbulence, a genetic algorithm is used to optimize the variable threshold segmentation and combine it with morphological processing, edge detection and other operations to extract the spot edges. Then, the set of edge feature points is divided into subsets of different sizes and the scatter matrices of different subsets are superimposed to obtain the scatter matrix of the entire set. Finally, the direct least squares method is used to calculate the fitting ellipse parameters and complete the spot centroid extraction. Experimental results show that when there is strong turbulence and the mask radius is 120 pixel, the mean gray intensity within the mask of this method is 1.32 pixel higher than that of the ellipse fitting method, and the horizontal center of mass offset is reduced by 4.59 pixel. This method can effectively achieve spot centroid extraction under varying degrees of turbulence disturbance, and can still maintain high accuracy when the spot is broken, providing technical support for spot centroid extraction in capture, pointing, and tracking systems for space laser communications.