摘要
针对水面舰船在复杂背景下存在大量虚警问题,提出了一种基于几何特征的水面舰船快速检测方法。首先,通过计算原灰度图的下采样序列,获得高斯降维图;其次,利用改进Sobel算法在粗分辨率下增强舰船的对比度,突显目标像素区域,用于区分降维图中目标与背景区域;再次,由于舰船目标与非舰船目标几何形状的不同,计算了长宽比、面积、区域矩形度等因子,结合先验信息鉴别大的舰船目标,同时剔除由于噪声、亮斑等引起的小虚警目标;最后,利用连通区域检测法标记出目标区域,实现海面目标的快速检测。实验结果表明:该方法具有一定的鲁棒性,能够快速、准确地实现海面舰船的自动检测。
To deal with the false alarms in the rapid detection of water vessels,a fast algorithm for detection of sea surface ships based on the geometrical features is proposed.Firstly,by calculating the under sampling sequence of the original gray image,the graph of Gaussian reduce dimension is obtained.Then,by using the improved Sobel edge detection methods in low resolution image,the contrast of ships is enhanced and the target pixel area is highlighted so that the target distinguishes itself from background in the reduce dimension image.Because of the difference between the ship target and non-ship target in geometrical features,the false alarm is eliminated by calculating the ratio of length and width,area and the region rectangle etc.with the help of the prior information.Finally,the target region is labeled by connected region methods and the sea surface object regions are extracted from the final label results.The rapid detection of sea surface objects is realized.The experimental results show that this method has certain robustness and achieves the automatic detection of sea surface ships quickly and accurately.
作者
王方超
张旻
宫丽美
WANG Fang-chao ZHANG Min GONG Li-mei(Dept.of Network System, Electronic Engineering Inst., Hefei 230037, China Key Laboratory Electronic Restricting Technique, Hefei 230037, China)
出处
《海军工程大学学报》
CAS
北大核心
2016年第5期57-63,共7页
Journal of Naval University of Engineering
基金
国家自然科学基金资助项目(61171170)
安徽省自然科学基金资助项目(1408085QF115)
关键词
降维图
SOBEL算法
几何特征
连通域标记
reduce dimension image
Sobel algorithm
geometric features
connected component labeling
