一种基于形状的近岸舰船检测

高分辨率光学遥感图像上近岸舰船的灰度、纹理特征往往和码头比较接近,其自动检测比较困难,但是舰船独特的形状特征却能将其很好地区分开;不变性广义霍夫变换是一种可以在未知位置、大小和朝向的条件下定位形状的非常有效的方法,但是它对局部形状变形的适应性比较差。改进了不变性广义霍夫变换,提出了一种基于鲁棒不变性广义霍夫变换的近岸舰船检测方法。实验结果表明该方法能有效地进行高分辨光学遥感图像中各类近岸舰船的自动检测。

Shape-based inshore ships detection

In the high resolution remote sensing image,the gray and texture characteristics of inshore ships are very similar to that of coast establishment.Thus it is hard to utilize the traditional area-based methods to detect them.However,the unique shape feature could distinguish the inshore ships from the docks.The invariant Generalized Hough Transform (IGHT) is an effective method to locate shapes with unknown position,size and orientation.But it has a bad adaptation to local shape deformation.We improved the IGHT method and proposed an inshore ships detection method which based on robust invariant Generalized Hough Transform.Experiment results demonstrate the effectiveness of our method used in the automatic detection of inshore ships.