基于声学分形特征的深海钴结壳识别研究

【作者】 周木荣；

【导师】 卜英勇；

【作者基本信息】 中南大学 ， 机械电子工程， 2009， 硕士

【摘要】 随着世界经济和科学技术的飞速发展,人类对矿产资源的需求量与日俱增,由于陆地资源的日趋枯竭,人们开始把目光转向海洋。深海钴结壳成为21世纪最具有商业开采价值的战略资源,世界主要发达国家已经开始了开采研究工作。面对国际社会对海洋资源争夺的形势,为维护我国的海洋权益,开辟我国新的矿产资源来源,本文在国家自然科学基金项目“深海钴结壳微地形监测技术与最佳采集深度建模研究”的资助下,对深海钴结壳的识别做了相关研究。根据分形理论在特征提取中的应用,本文提出了一种基于分形特征的声学回波识别深海钴结壳的方法。论文紧紧围绕特征提取、特征优化和分类器设计这三个问题,对钴结壳识别展开研究。根据超声波探测底质得到的回波信号具有分形特性,在讨论了利用单一维数不能完全刻画分形信号的基础上,提取了广义维数分形特征。由于原始特征的维数较高且存在非线性关系,利用核Fisher判别分析方法进行特征优化,得到了最佳非线性目标识别特征矢量。最后设计了概率神经网络分类器。在此理论基础上,对我国钴结壳调查区内的23种代表性海底底质,在实验室水池内进行钴结壳识别试验。试验结果表明:钴结壳正确识别率平均达到79.6%,非钴结壳被判为钴结壳的错判率平均仅为23%。由此可看出,本论文提出的钴结壳识别方法是有效可行的。本文的研究为深海钴结壳识别提供了良好的理论基础,为我国的深海采矿事业提供了有效的技术支持。更多还原

【Abstract】 Along with the fast development of economy and science in the world, the need of mineral resources is growing day by day. However, the land resources are becoming less and less. In this situation, people tend to mine the resources in the ocean. Oceanic cobalt crust resource has been a commercial foreground strategically resource in 21 century and the head developed countries in the world have been doing the work of investigati-on and exploitation. In order to mine the new resources by our country independently, the investigation is done of the identification and classifiction of the cobalt-crust in the dissertation, imbursed by the national natural science fund item "study on the Abyssalbenthic Cobalt-rich Crusts Tiny terrain Detecting Technology and the Best collection deepness model".A classifiction method of deep-sea cobalt-crust was proposed based on the application of fractal theory in feature extraction. The questions of feature extraction, feature optimization and classifier design were sutdy. As provide more information than individual dimension, the generalized dimensions of echo signals were used for cobalt-crust classifiction based on the assumption that the sonar echo signals have a fractal structure. To deal with the problem of original features with the characteristic of large dimension and nonlinear relationship among them, Kernel Fisher discriminant analysis is adopted to extract the optimal nonlinear discriminant features. At the end, the Probability Neural Network was used as the classifier. Based on the theory mentioned above, the classification experiments of 23 kinds of seabed materials were done. The experimental results indicated that the exactness identification rate of cobalt-crust was about 79.6 percents and the error rate of mistaking the non-cobalt-crust as cobalt-crust was only 23 percents. Look from the results, this method was effective and reliable.The research of the dissertation affords favorable theoretic value for the identification and classification of the abyssalbenthic cobalt-crust. The important is that the research furnishes effective technique sustain for the country abyssal mining.更多还原