高分辨率遥感影像飞机目标的双阶段检测方法

Two-stage aircraft detection for high-resolution remote sensing image

针对现有方法进行真实应用场景中大批量的遥感影像进行特定目标检测时耗时且困难的问题,该文根据由粗到精的思路提出了一种遥感影像飞机目标的双阶段检测框架。首先,对预训练模型进行迁移训练,在下采样影像中利用机场检测模型搜索机场区域,实现对目标区域的锁定。然后,对机场场景,利用第二次迁移的模型进行飞机目标检测。结果表明,所提方法可以提高遥感影像检测飞机目标的效率和精度。通过不同阶段的筛选可以去除大部分无效区域,有效避免了在非机场区域产生的误检,提高了准确率。

Aiming at the problem of time-consuming and difficult detection of specific targets in large batches of remote sensing images in real application scenarios by existing methods, this paper proposed a two-stage detection framework for aircraft targets in remote sensing images based on the idea of rough to precise. Firstly, transform and train the pre-trained model, and use the airport detection model to search for the airport area in the down-sampled image to achieve the lock on the target area. Then, for the airport scene, the second transferred model was used for aircraft object detection. The results showed that the proposed method could improve the efficiency and accuracy of remote sensing image detection of aircraft targets. Through different stages of screening, most of the invalid areas could be removed, which effectively avoided false detection in non-airport areas and improved the accuracy.