基于Ecogniton的光学遥感图像舰船目标检测

Ecognition-based Ship Detection on Optical Remote Sensing Images

利用卫星遥感技术对海上舰船目标进行识别定位具有重要的意义。提出了一种基于面向对象的图像分割和分类软件Ecognition的技术方法来提取高分辨率遥感卫星图像中的海上舰船目标。利用多尺度分割算法和光谱差异分割算法对高分辨率卫星图像进行初步分割;利用归一化差异水体指数(NDWI)或灰度共生矩阵(GLCM)同质性纹理特征对陆地和海洋进行辨别;利用对象的相邻关系特征和形状特征来获取潜在的舰船目标;利用数学形态学算子实现舰船目标的连续化,从而获得最终的海上舰船目标。利用高分辨率卫星图像进行了实验验证,实验结果说明了所提出的技术可以有效地实现海上舰船目标的识别与定位。

Ship detection and ship location on the satellite remote sensing images has great significance for civilian and military usage. This paper proposes a practical ship detection method based on an object-oriented image segmentation and classification software named Ecognition on remote sensing images of high spatial resolution. Firstly multi-resolution segmentation and spectral difference segmentation are involved for high-resolution satellite images,and normalized difference water index（ NDWI） or Grey Level Concurrence Matrix（ GLCM） homogeneity texture features calculated from segment units are used to distinguish the land and the sea. Then,the underlying regions embedding ships can be detected by object features and class related features. Finally,the ship detection results can be obtained after ships are joined by mathematical morphology operations. Based on two cases of experiments on the high-resolution satellite images,it is shown that the proposed method can efficiently realize the ship detection and ship location.