An algorithm called gradient clustering based area proposal method (APM) is proposed to solve the problem that the existing methods are slow to detect objects, which is based on a large number of edges of artificial objects in aerial images. Then the extracted regions of interest are detected by the object detection method. The real-time performance and precision rate of this method are evaluated on the DOTA (Dataset for Object Detection in Aerial Images). The research results show that the proposed method greatly improves the detection speed of large-size, target-dense aerial images by the object detection algorithm, and still keeps a high recall rate.