没有合适的资源?快使用搜索试试~ 我知道了~
人工智能-目标检测-基于词包模型的高分辨率SAR图像舰船目标检测.pdf
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 92 浏览量
2022-06-25
22:55:44
上传
评论
收藏 1.76MB PDF 举报
温馨提示
试读
59页
人工智能-目标检测-基于词包模型的高分辨率SAR图像舰船目标检测.pdf
资源推荐
资源详情
资源评论
摘要
摘要
合成孔径雷达(SAR)技术的发展日新月异,研究如何在 SAR 图像上检测舰船
目标对我国民用和军事领域的海域监控都具有重大意义。本文围绕 SAR 图像舰船
目标检测问题开展了相关研究,主要研究了基于词包模型的高分辨率 SAR 图像目
标检测以及精确分割定位方法。
在舰船目标检测预处理过程中,本文分别研究了 SAR 图像去噪问题以及
SAR 图像海陆分割问题。针对 SAR 图像去噪问题,研究了 NL-means 去噪方法,
实验结果表明该去噪方法可以较好的达到去噪要求。针对 SAR 图像海陆分割问题,
通过总结现有的陆地掩模方法,综合考虑检测精度和算法复杂度问题,研究了基
于最大类间方差法(OSTU)结合形态学处理方法实现了海陆分割。以上两步的预处
理过程为后续的目标检测打下了良好的基础。
在目标检测以及精确定位方面,针对 SAR 图像分辨率的提高带来的细节信息
增加这一问题,本文通过分析高分辨率图像上舰船目标和海洋背景的特征区别,
提出了基于词包模型的舰船目标精细定位算法。首先在特征提取方面,本文提出
了基于图像块的特征提取方法,该特征提取方法可以很好的同时保留平滑区域和
纹理区域的特征。为了提高检测效率,本文在分类决策过程中提出了两级检测器
实现目标检测定位。第一级检测器是基于图像块灰度直方图构造方法,该方法通
过构造图像块灰度直方图,通过支持向量机(SVM)分类器实现目标检测,对检测
图像实现 1616 块分割。第二级检测器是基于局部图像块组合直方图构造方法,
通过构造局部图像块组合直方图,在第一级检测器的结果上继续进行 44 大小的
块分割。检测结果中当虚警率为 1%时,目标的检出率可以达到 93%以上,定位
精度误差大约为1 个像素点左右。实验结果表明,本文的算法很好的实现了舰船
目标精确定位。
关键字:舰船目标检测 词包模型 高分辨率 SAR 图像 多级检测器
Abstract
Abstract
Synthetic Aperture Radar (SAR) technology is developing rapidly. Implementing
automatic and exact ship target detection in SAR images is of great significance to the
defense of our country. The main contents of this paper are the bag-of-words based ship
detection method in high-resolution SAR images and the accurate segmentation method.
In the ship target detection preprocessing, we investigate the methods of denoising
and segmenting land and sea for SAR images respectively. The NL-means method is
selected for SAR images denoising. Experimental results show that NL-means can
achieve good results. For land and sea segmentation taking the detection accuracy and
the complexity of the algorithm into account, we summarize the existing land masking
method. Then, we combine the morphological method with OSTU to obtain the sea land
mask. The two-step preprocess has laid a good foundation for the subsequent detection
of the ship.
In order to accurately detect and segment ship in SAR images, in this paper a novel
detection algorithm is proposed, which is based on bag-of-words. In the feature
extraction step, we proposed an image block based feature extraction method, which can
well retain characteristics for both the smooth and textural regions. In order to improve
the detection efficiency, a two-stage detector proposed to detect the target step by step.
The first level detector adopts an image gray value block based histogram construction
method, which is achieved by support vector machine (SVM) and segments the
detecting image into1616 blocks. The second level detector adopts a local image block
based histogram construction method, which continue segments the result from the first
level detector into 4 blocks. Under the false rate of 1%, the ship detection rate is
more than 93%, and the positioning error is approximately 1 pixels. Experimental
results show that the proposed algorithm is highly precise for the ship target positioning.
Keywords:Ship detection Bag-of-words High-resolution SAR images Multi-
level detection
剩余58页未读,继续阅读
资源评论
programhh
- 粉丝: 8
- 资源: 3838
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功