一种新的合成孔径雷达图像船只几何特征提取方法

A Novel Ship Geometric Feature Extraction Method for SAR Imagery

海上船只监测在海洋交通、渔业管理等领域发挥着重要的作用。高分辨率合成孔径雷达卫星的发射,使船只类型识别成为可能,进一步提高了海洋监测的能力。几何特征是一种重要的船只类型识别特征,本文提出了一种新的合成孔径雷达图像船只几何特征提取方法。与传统方法不同,本文利用最大稳定极值区域算法,取代常用的恒虚警率算法,来检测定位船只。这种方法能够在同等检测率的情况下,有效的降低虚警率,并且具有更快的速度。在几何特征提取过程中,本文提出了改进的最小外接矩形提取方法,这种方法能够有效的抑制旁瓣对船只几何特征提取的影响。实验证明,本文提出的方法能够更快速、准确的提取船只的几何特征。

The monitoring of marine vessels plays an important role in the field of ocean transportation and fishery management. The launch of the high resolution synthetic aperture radar （SAR） makes it possi- ble to identify the type of vessel, and further enhance the capability of the ocean monitoring. Geometric feature is one type of the important characteristics of vessel type recognition. This paper presents a new method to extract geometric features of vessels from SAR image. Different from the traditional method, this paper uses the maximally stable extremal regions （MSER） algorithm, rather than widely used the constant false alarm rate （CFAR） algorithm, to detect the location of the vessel. This method can effec- tively reduce the false alarm rate in the case of the same detection rate, and has a faster speed. In the process of geometric feature extraction, this paper proposes an improved minimum bounding rectangle （MBR） extraction method. This method can effectively suppress the influence of the sidelobe on the geo- metric feature extraction of vessels. Experiments show that the proposed method can extract the geometric features of the shipmore quickly and accurately.