基于港口匹配和海域分割的靠岸舰船检测方法

发布时间：2018-06-04 18:57

  本文选题：靠岸舰船检测 + 港口定位　； 参考：《华中科技大学学报(自然科学版)》2017年10期

【摘要】：针对可见光图像中靠岸舰船的灰度、纹理等特征的自动检测比较困难的问题,提出一种基于港口匹配和海域分割的靠岸舰船检测方法.根据飞行器实时返回的姿态信息,将港口模板变换到实时图同一视角下进行港口配准,确定实时图海域部分,再对海域进行分割,检测出舰船.采用基于边缘梯度矢量的港口配准方法,克服岸内边缘干扰,提高匹配精度.提出一种特征融合聚类结合生长的分割算法,解决灰度不均匀导致的分割后舰船断裂现象,优化海陆分割效果.试验结果证明了该方法的有效性和鲁棒性.
[Abstract]:Aiming at the difficulty of automatic detection of landing ship's gray and texture features in visible light image, a method based on port matching and sea area segmentation is proposed. According to the attitude information returned by the aircraft in real time, the port template is transformed to the same angle of view of the real-time image for port registration, and the real-time sea area is determined, then the sea area is segmented, and the ship is detected. The port registration method based on edge gradient vector is used to overcome the inshore edge interference and improve the matching accuracy. A feature fusion clustering combined with growing segmentation algorithm is proposed to solve the ship fracture phenomenon caused by uneven grayscale and to optimize the effect of sea-land segmentation. The experimental results show that the method is effective and robust.
【作者单位】： 华中科技大学自动化学院;华中科技大学多谱信息处理技术国家重点实验室;北京航天自动控制研究所;
【基金】：国家十三五科技预研资助项目(41415020402)
【分类号】：TP751;U675.79
，

本文编号：1978494