液晶与显示, 2019, 34 (8): 803, 网络出版: 2019-10-12   

基于谱残差和梯度纹理融合特征的舰船检测

Ship detection based on spectral residual and gradient texture fusion features
作者单位
1 中国科学院 长春光学精密机械与物理研究所, 吉林 长春 130033
2 中国科学院大学, 北京 100049
3 北京信息科技大学 机电工程学院,北京 100192
摘要
本文提出一种谱残差和梯度纹理融合特征来检测可见光遥感图像中复杂海面背景下的舰船目标。首先, 使用谱残差模型定位疑似舰船目标, 并采用自适应动态阈值法获取候选目标区域。然后, 根据舰船的形状特点, 对梯度方向直方图特征进行改进, 设计出表征舰船特性的梯度方向特征。同时, 将提取候选目标区域的统一化LBP特征的方差以及灰度共生矩阵特征相结合来描述舰船的纹理信息, 得到30维特征向量。最后, 通过训练好的AdaBoost分类器来完成舰船目标鉴别。本文的检测算法, 针对尺寸为1 024×1 024的可见光遥感图像, 检测时间为4.792 6 s, 检测精度为95.51%, 召回率为96.65%。实验结果表明: 本文算法能准确提取海面舰船目标, 获取舰船目标的数量和位置信息, 从检测时间和精度上来看, 可以作为实际工程参考。
Abstract
A spectral residual and gradient texture fusion feature is proposed to detect ship targets in complex sea backgrounds in visible-light-remote sensing image. Firstly, the spectral residual model is used to locate the suspected ship target, while the adaptive dynamic threshold method is used to obtain the candidate target region. Then, according to the shape characteristics of the ship, the gradient direction histogram features are improved to design the gradient direction characteristics that show the feature of the ship. Meanwhile, the variance of the uniform LBP features and the gray level co-occurrence matrix features of the candidate target regions are extracted to describe the ships texture information, and the 30-dimensional feature vector is obtained. Finally, the ships target identification is accomplished through the trained AdaBoost classifier. For visible- light-remote sensing image with size1 024×1 024 , the detection time is 4.792 6 s, the detection accuracy is 95.51% and the recall rate is 96.65% via using the detection algorithm of this paper. The experimental results show that the proposed algorithm can successfully extract the surface targets of ships, and obtain the number and position information of ship targets accurately. From the perspective of detection time and accuracy, it can be used as a reference for practical engineering.

李庆峰, 何斌, 王文胜, 苏畅, 韩玺钰, 梁怀丹. 基于谱残差和梯度纹理融合特征的舰船检测[J]. 液晶与显示, 2019, 34(8): 803. LI Qing-feng, HE Bin, WANG Wen-sheng, SU Chang, HAN Xi-yu, LIANG Huai-dan. Ship detection based on spectral residual and gradient texture fusion features[J]. Chinese Journal of Liquid Crystals and Displays, 2019, 34(8): 803.

本文已被 1 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!