面向航空飞行安全的遥感图像小目标检测

有人机和无人机等各种新型航空飞行器的发展,给航空飞行安全带来了极大挑战,对影响飞行安全的小型目标进行检测是保障安全飞行的首要条件。本文针对现有基于深度学习的目标检测方法在遥感图像小目标检测时存在的不足,以及检测目标尺度过小、图像背景复杂、噪声干扰等问题,探讨了深度学习技术在遥感图像小目标检测方面的研究进展,重点分析了特征金字塔网络、注意力机制、倾斜框检测等相关技术在遥感图像小目标检测上的可行性,提出了一种具有较强泛化能力的目标检测模型。本文以高分二号遥感图像的高压电塔检测为例进行试验,结果表明,本文提出的模型在检测精度和计算开销上可达到更好的效果。

Small Target Detection in Remote Sensing Images Based on Aviation Security

The development of all kinds of new aviation aircraft,such as Manned aircraft and UAV,has brought great challenges to flight safety,and small targets detection is the primary condition to ensure safe flight.Aiming to the problems of small scale of target,complex background and noise disturbance in remote sensing images,as well as the deficiencies of existing deep learning models for small target detection,This paper presents the advances of deep learning used in small target detection of remote sensing image.,mainly analyzes the feasibility of feature pyramid network,attention mechanism,inclined box detection and other related technologies in small target detection of remote sensing image,and proposes a target detection model with strong generalization ability.It is taking high voltage tower detection in No.2 remote sensing image as an example to perform the experiment.The results show that the newly proposed model has better detection accuracy,generalization ability and computation cost.