摘要: |
针对前视红外图像中的机场跑道检测问题,提出了基于深度学习的端到端的实时检测算法。算法首先利用深度学习在物体特征表达上的优点,采用当前主流端到端的YOLO(You Only Look Once)V2检测算法提取候选目标,寻找跑道所在位置;然后在已经获取跑道所在边框的基础上,在神经网络最后一层采用多尺度线段检测器(Line Segment Detector,LSD)进行精确的线段检测;最后对所检测的线段进行融合,提取轮廓线。真实实验数据表明,该算法基本上能满足机场轮廓提取实时性好、提取精度高、抗干扰性强等要求。 |
关键词: 机场跑道检测 前视红外(FLIR)图像 轮廓提取 深度学习 |
DOI: |
|
基金项目:航空工业技术创新基金项目(2014D6310R);航空科学基金项目(2015ZC31005) |
|
Contour extraction of runway′s forward looking infra-red(FLIR) images based on deep learning |
YUAN Lei,CHENG Yue,NIU Wensheng,LUO Wuyang |
(1.Xi′an Aeronautics Computing Technique Research Institute,AVIC,Xi′an 710068,China;2.Aviation Key Laboratory of Science and Technology on Airborne and Missile-borne Computer,Xi′an 710068,China) |
Abstract: |
A real-time detection algorithm based on end-to-end deep learning(DL) is proposed for the forward looking infra-red(FLIR) runway detection.The algorithm firstly uses the advantage of DL in the expression of object features,adopts current mainstream end-to-end You Only Look Once(YOLO)V2 to extract the candidate targets and find the position of the runway precisely.Then,based on the border of the runway,it applys the multiscale merging method of line segment detector(LSD) to extract the contour of the runway at the last layer of the neural network.Finally,it fits the extracted line segment.Real experimental data shows that the algorithm can meet the requirements of airport profile extraction such as high real time,high contour integrity and stronger anti-jamming and so on. |
Key words: airport runway detection forward looking infra-red(FLIR) image contour extraction deep learning |