Solid of revolution, whose shape can be obtained by rotating a plane curve around a straight line (the axis) lying on the same plane, is one of the most important parts of many spacecrafts. Some important parameters of these spacecrafts can be readily determined through the elliptical images of the boundaries of the revolution solids. An approach for ellipse extraction from spacecraft images based on the characters of the spacecraft image is presented. First, we detect the edges of an image, from which we remove all extraneous edge pixels that interfere with ellipse extraction by our preprocessing method. Second, the elliptical arcs are selected from the edge image under the consideration of the global shapes of edge segments. Then, the exact geometric distances of points to a conic are computed without any iteration to measure the similarity between any two of all selected arcs by a new criterion. Finally, we integrate those arcs which belong to the same ellipse together and then fit one ellipse into them. Both synthetic and real experiments are conducted to assess the efficiency and robustness of our method. The result shows that our method has better performance for extracting ellipses from spacecraft images with relatively short time than other ellipse extraction methods. This method will be significant for some space applications based on conic information.