摘要
船舶检测在民用和军用领域都具有广阔的前景。利用高分辨率SAR图像可以提高海上船舶检测精度,但同时也面临船舶旁瓣、海表漂浮物及人工设施等物体的干扰。针对这些不足,提出了一种基于最佳熵双阈值算法对SAR影像检测的基础上,利用多种几何属性特征组合优化初步检测结果的算法。首先,基于最佳熵双阈值算法对研究区域的Radarsat-2影像进行初步检测,得到潜在船只目标;其次,计算潜在目标的核密度、长宽比、潜在像元数目等3个特征;最后,利用客观权重分配方法——变异系数法,针对3个特征进行权重分配,降低虚警率,达到优化船只检测结果的目的,同时利用AIS数据、K-CFAR算法检测结果及黄河口的ALOS-2数据验证了该算法的有效性和实用性。
Ship detection is crucial both in civil and military areas.Using high-resolution SAR images can improve the accuracies of ship detection on the sea,but it also faces the interferences of the boats＇ side lobes,floats at sea surface,artificial facilities and other objects.based on the optimal entropy dual-threshold algorithm,several features of geometric properties were combined to optimize the result of the detection.Firstly,apreliminary detection on the Radarsat-2image of the study area was conducted with the dual-threshold algorithm,and potential ship targets were obtained.Secondly,three features of the potential targets,including the kernel density estimation（KDE）,length-width ration（LWR）and ship pixels（SP）were calculated.Thirdly,the variation coefficient method is used to distribute appropriate weights to the three features,lowering the false alarm and achieving a better accuracy.Finally,the method was validated with the AIS information,K-CFAR algorithm and the ALOS-2 image of Yellow River mouth,and the validation results approved the method＇s effectiveness and applicability.
出处
《遥感技术与应用》
CSCD
北大核心
2017年第2期305-314,共10页
Remote Sensing Technology and Application
基金
高分辨率对地观测系统重大专项(01-Y30A03-9001-12/13)
关键词
船舶检测
高分辨SAR
权重分配
变异系数法
Ship detection
High-resolution SAR
Feature combination
Variation coefficient method
