SIA OpenIR  > 光电信息技术研究室
声呐图像水雷目标识别方法研究
其他题名For mine recognition method research of sonar image
李娟娟1,2
导师郝颖明
分类号TJ61
关键词声呐图像 目标识别 C-v模型
索取号TJ61/L32/2014
页数69页
学位专业模式识别与智能系统
学位名称硕士
2014-05-28
学位授予单位中国科学院沈阳自动化研究所
学位授予地点沈阳
作者部门光电信息技术研究室
摘要水雷是海战中必不可少的武器,由于水雷强大的破坏力,反水雷战已经越来越多地得到世界各国的重视。鉴于水下环境的特殊性,声呐成为实现水雷探测的主要手段。然而,由于声呐图像具有噪声大,灰度分布不均匀、目标边缘模糊等特点,在声呐图像中自动识别水雷目标仍是当下尚未解决的问题。因此,研究声呐图像的水雷目标识别方法,不仅是对水雷目标识别方法有意义的理论探索,也是实际工程应用的迫切需求,具有重要的理论意义和应用价值。本文以水雷识别为背景,开展声呐图像目标识别方法研究,分别针对水雷目标的形状特征和灰度特征,采用Chan-Vese主动轮廓模型的演化思想,修改了Chan-Vese模型的水平集函数和能量函数,提出了基于改进Chan-Vese模型的声呐图像水雷目标识别方法,并设计开发了声呐图像水雷目标识别软件。水雷目标在声呐图像中接近矩形,根据这一形状特征,本文将椭圆形状约束引入Chan-Vese模型的水平集函数,得到基于椭圆约束的Chan-Vese模型。该模型具有形状保持的特点,解决了声呐图像中水雷目标发生形状畸变的问题,但还存在识别结果不准确和演化过程不稳定的缺点。经过分析,将更适合水雷形状的超椭圆约束引入Chan-Vese模型的水平集函数,提出了基于超椭圆约束的Chan-Vese模型,解决了椭圆约束Chan-Vese模型存在的问题。在此基础上,针对声呐图像的灰度特征,引入另一个水平集函数,同时改进了Chan-Vese模型的能量函数,得到了基于多相水平集的超椭圆约束Chan-Vese模型。该模型不仅具有形状保持的特点,而且还融合了声呐图像中水雷的声影区特征,能够排除虚假水雷目标,实现正确识别水雷的目的。最后,基于改进Chan-Vese模型的水雷目标识别方法,开展了声呐图像水雷目标识别软件设计与实现技术研究,开发了水雷目标识别软件。该软件实现了四种基于不同演化模型的识别方法,并且能够实时显示各个演化模型在图像中的演化过程,便于用户使用和观察水雷目标识别实验结果。实验结果表明,本文提出的基于多相水平集的超椭圆约束Chan-Vese模型能够很好地识别声呐图像中的水雷目标;该方法不仅能够识别噪声较大、目标边缘比较模糊的声呐图像的水雷目标,而且识别结果具有很好的抗变形性能。上述研究成果已经在实际科研项目中得到应用。
其他摘要Mine is an essential part of weapons in the naval battle. In view of the mine’s strong destructive power, the mine warfare has got more and more attention of the countries all over the world. Due to the particularity of underwater environment, Sonar is the main method to realize mine detection. However, due to the sonar image characteristics of the weak feature and poor target imaging condition, it is still an unsolved problem to automatically recognize mine targets in the sonar image. Therefore, the study of mine target identification methods in sonar image has important theoretical significance and application value. It not only carries significant theoretical exploration of the recognition methods of mine targets in sonar image, but also lays the foundation for the practical engineering application.Based on the background of mine recognition, the research of mine target identification methods in sonar image has been carried out. According to the mine target shape characteristic and the gray scale characteristics, adopting the evolution thought of C-V model, the level set function and the energy function of C-V model has been changed. Therefore the improved C-V model has been proposed in this paper, and the mine target recognition software of sonar image has been designed and developed.The shape of mine in the sonar image is close to the rectangle. According to this characteristic, the elliptical shape constraint is introduced to the level set function of C-V model. Therefore, the C-V model based on elliptic constraint has been put forward. This model has the characteristic of shape preserving, which solves the shape distortion problem of mine target in the sonar image. However, this model has the disadvantages of inaccurate recognition and unstable evolution process. Then the C-V model based on the hyper-elliptic constraint is put forward, which is more suitable for the shape of mine. This method has resolved the problems of the C-V model based on elliptic constraint.Then on the basis of hyper-elliptic constraint C-V model, according to the gray scale characteristic of sonar images, another level set function is introduced. At the same time, the energy function of C-V model has been changed. Therefore, the multiphase level set C-V model based on hyper-elliptic constraint is proposed. The model not only has the characteristics of shape preserving, but also has the ability of eliminating the false mine target because of combining with acoustic shadow characteristics of mines, which achieves the goal of correct recognition of mine. At last, based on the mine identification method in sonar image of improved C-V model, the sonar image target recognition software design and implementation of technology research are carried out, and the mine target recognition software has been developed. The software implements the recognition methods based on four different evolution models, and the evolution process of various evolution models of the image can be real-time displayed, which makes it easy for users to use the software and observe the mine recognition experiment results.The experimental results show that, the mine objects in sonar images could be identified well by using the multiphase level set C-V model based on hyper-elliptic constraint, which is put forward in this paper. The proposed method can not only identify the mine of sonar image which has more noise and blurred mine target edge, but also has the identification results that have good performance of resistance to deformation. The above research results have been applied in an actual scientific research project.
语种中文
产权排序1
文献类型学位论文
条目标识符http://ir.sia.cn/handle/173321/14798
专题光电信息技术研究室
作者单位1.中国科学院沈阳自动化研究所
2.中国科学院大学
推荐引用方式
GB/T 7714
李娟娟. 声呐图像水雷目标识别方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2014.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
声呐图像水雷目标识别方法研究.pdf(2964KB) 开放获取CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[李娟娟]的文章
百度学术
百度学术中相似的文章
[李娟娟]的文章
必应学术
必应学术中相似的文章
[李娟娟]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。