基于侧扫声纳的海底沉积物分类研究
文献类型：学位论文
| 作者 | 邓朝晔 |
| 学位类别 | 博士 |
| 答辩日期 | 2005 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 中国科学院声学研究所 |
| 关键词 | 海底沉积物 纹理分析 分类 K均值分类器 |
| 其他题名 | The Studies on Seabed Sediment Classification Based on Sidescan Sonar |
| 中文摘要 | 海底沉积物是海洋测绘的基本要素之一，海底沉积物的探测与分类对国防及国民经济建设都具有十分重要的意义。国内外已经对海底沉积物的声学分类进行了大量的研究，从理论和方法上证明了海底声学遥测的可行性。本文通过研究侧扫声纳图像的纹理特征量的提取和分类识别方法，找到了几个具有不错分类效果的特征量组合。在论文中，我们分析了表征图像纹理特性的各种特征量，它们是游程长度分析、灰度共生矩阵、纹理能量和分形维。利用大量己知类型的海底沉积物样本，使用K均值分类器对特征量的分类效果进行检验，结果说明单个特征量的分类效果是很有限的；所以我们对特征量进行了两两组合，并且也对它们的分类效果进行检验。实验证明：游程长度矩阵的游程长度分布和基于LSES模板的纹理能量特征量的组合、灰度共生矩阵的逆差矩和分形维特征量的组合、基于LSES模板的纹理能量和分形维的组合，这些特征量组合的分类效果无论在正确率上还是抗噪性上都是令人满意的。这些特征量的选取为后面的分类器的设计奠定了基础。 |
| 英文摘要 | Seabed sediment is one of the basic factors in sea mapping. It is important for national defence and economy to explore and classify the seabed sediments. A lot of studies have been done in seabed sediments classification by using acoustic methods, it proves the feasibility of seabed remote sensing by using acoustic methods in theory and fact. This dissertation finds several feature's combinations, which give correct classification purpose, by studying the methods of extracting of texture features and classifying in sidescan sonar images. In this dissertation, some kinds of features are analyzed, which express the texture characterization of images. They are run length matrix analysis, gray level co-occurence matrix, texture energy measure and fractal dimension. These features have been tested by using K-mean classifier in a great deal of seabed sediment samples of known types, the results show the effect of single feature is limited. So the features are combined, and their classification efftct are tested. Experiments prove: the combination of the run length nonuniformity of the run length matrix and the texture energy measure based on L5E5 mask, the combination of the the angular inverse difference monent of the gray level co-occurrence matrix and fractcl dimension, the combination of the texture energy measure based on L5E5 mask and fractcl dimension, these feature combinations' classification effects are satisfaction both in exactitude rate and robust. These features' selection is the base of the design of the classifiers. |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 54 |
| 源URL | [http://159.226.59.140/handle/311008/962]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 邓朝晔. 基于侧扫声纳的海底沉积物分类研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2005. |
入库方式： OAI收割
来源：声学研究所
浏览0
下载0
收藏0
其他版本
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。