基于轻量化网络的光学遥感图像飞机目标检测

光学遥感图像飞机检测是遥感分析的重要研究方向。现有检测方法难以达到满意的效果，传统检测方法由于手工特征建模困难，易受背景干扰，导致其鲁棒性普遍偏低；而以复杂度提升为代价来提高检测性能的深度学习目标检测方法无法在资源受限下的星载平台得到广泛应用。针对上述问题，本论文提出一种具有轻量化多尺度特点的深度学习飞机目标检测方法。在多尺度目标检测框架（SSD）基础上，利用密集连接结构和双卷积通道构成具有特征重复利用、计算效率高等特点的基础骨干网络，之后连接一个由残差模块和反卷积构成的多尺度特征融合检测模块，以提高飞机小目标的检测性能。实验结果表明，在多种复杂机场场景中，本文的方法与当前经典的深度学习目标方法相比，在保持较高目标检测精度的同时，又能具有较低的计算复杂度。 

Aircraft detection in Remote Sensing image based on Lightweight Network

Aircraft detection in optical remote sensing image is an important research direction in the field of remote sensing. The existing detection methods are difficult to achieve satisfactory results. Traditional detection methods are low robustness, due to manual feature modeling are difficult and subject to background interference; The deep learning target detection method, which improves the detection performance at the cost of complexity improvement, cannot be widely used in space-borne platforms limited resources. In view of the above problems, this paper proposes an aircraft target deep learning detection method with lightweight and multi-scale features. On the basis of the multi-scale target detection framework(SSD)，the method firstly uses the dense connection structure and the double convolution channel to form the basic backbone network with feature reuse and high computational efficiency. To improve the detection performance of the small aircraft target, the basis backbone network connects a residual module and deconvolution to compose the multi-scale feature fusion detection module. Compared with the current classical deep learning object detection methods, the experimental results show that the proposed method has the advantages of maintaining low computational complexity and achieving high detection accuracy.