摘要
为了在通用硬件平台上实现光学遥感图像海面舰船检测,提出一套适合实际系统应用的算法。首先,在海陆分割阶段设计能快速粗分图像内容的自适应双门限阈值分割方法;其次,根据海面舰船分布稀疏的特点对分块后的图像进行目标存在性初判;接着,只对初判为有目标的分块提取目标候选区域,提高了检测效率,其中引入杂波率系数自适应地控制Top-Hat算子中结构元素的尺寸,使其更好地抑制背景;最后,采用基于特征的模式识别方法剔除虚警。测试结果验证了算法的性能,并在硬件平台上进行了实现。
A ship detection algorithm was developed for general hardware platforms for optical remote sensing images.The system uses an adaptive threshold histogram-based segmentation technique for the land masking.Then,the sparseness of ships on the sea is used to develop an effective prescreening approach based on the tile statistical features with only the target-labeled tiles processed in the next step.A clutter rate coefficient is used to control the structural elements of the morphology Top-Hat operator for better background suppression.Finally,a pattern recognition method is used for discrimination.Tests show the effectiveness of this algorithm implemented on a hardware platform.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2011年第1期105-110,共6页
Journal of Tsinghua University(Science and Technology)
基金
国家"八六三"高技术项目
