In this paper, we present close-range vision navigation and guidance for rotary autonomous landing. To consider the case which the entire shape of the marker cannot be shown in close range, we suggest a marker design consists of `H' shape and concentric circles. To recognize circles using only portion of the marker, ellipse fitting method by direct least square is used. Thus, the pixel coordinates for the center of circles are calculated from the coefficients of general conic equation. By using them, the relative position can be obtained. Moreover, we propose a close-range vision-based guidance technique to make a rotary UAV land on the center of the landing marker. The simulation results show that the obtained relative position is close to the true relative position. Consequently, the vision-based autonomous landing is accomplished.