A new aircraft attitude estimation algorithm by use of line features for multi-camera images was proposed based on the direction vector method of angle bisector line. First, the angle bisector method was used to get direction vectors of the central and coplanar vertical axes of the aircraft, thus the rotation matrix R from the aircraft coordinate system to the theodolite camera reference coordinate system was obtained. Then, all line edges of the wings on theodolite images were extracted, and the total spacial coplanar error equation functions of line characteristics were derived with the pitch, azimuth and other parameters of theodolite. Moreover, an improved orthogonal iteration method was presented to minimize the total error of rotation matrix R. Finally, the optimized rotation matrix R was decomposed to get the aircraft's attitude. By a simulation experiment, the accuracy of four-camera system achieves 0.17° and the speed is 32 frame/s, which shows the accuracy is much better than that of the angle bisector method and the processing speed reaches real-time level. These results prove the effectiveness and superiority of the proposed algorithm.