To promote the accuracy of Automatic Target Recognition (ATR) system, a new method was proposed for detecting airplane on high resolution remote sensing imageries, which had another procedure of clutter rejecting after target detecting in ATR. Firstly, we divided the whole image into small blocks based on the strategy of average area with superposition. Then the local recursive image segmentation algorithm was used to segment interesting targets on every block. With respect to the processes performed before segmentation and the fact that the segmentation of airplane could not be performed perfectly, an airplane’s axes detecting method was put forward by using added line and radon transformation. Through these processes, the recognition efficiency and the performance of airplane ATR were expected to be promoted much better. Finally, an experiment shows the potential ability of this new method especially for the under-segmented binary image of airplane.