【作者】 鲍姗娟；

【导师】 姬光荣；

【作者基本信息】 中国海洋大学 ， 通信与信息系统， 2011， 硕士

【摘要】 近年来,随着海洋环境污染和海水富营养化程度的加剧,赤潮灾害越来越频繁,造成了严重的生态破坏和经济损失,引起了世界各国政府、公众和科学家的关注。我国从基础研究、高技术发展等不同方面研究赤潮发生的机制、预警、预报与防治方法,尝试建立赤潮监测体系等。其中快速有效的鉴定引发赤潮的主要藻种是赤潮自动监测技术中的一个重要组成部分。分形理论是现代非线性科学研究中一个十分活跃的数学分支,近年来它在图像处理领域中得到了广泛应用。其中分形编码因为具有高压缩比的优势而得到了迅速发展,同时根据分形编码具有吸引子不变性的特点,我们可以利用针对进行了分形编码的图像而定义的图像间相似度的测距方式——分形邻距来进行图像匹配,从而达到识别的目的。本文选题来源于国家高技术研究发展计划(863计划)“有害赤潮生物诊断系统技术研究”(编号:2006AA09Z178),主要针对我国沿海常见有害赤潮藻,通过采集藻种不同生长时期、不同角度的多视点图像,以分形理论为基础,对经过目标提取后的有害赤潮显微图像的识别进行了研究。主要工作包括:1、在图像采集的过程中,赤潮藻类细胞周围往往存在一些泥沙、残骸等干扰物,这使得图像的背景变得极为复杂,会对后续的自动识别产生较大的影响。针对这一缺陷,提出了一种基于大津法和轮廓提取技术相结合的目标提取方法,该方法对无角毛类藻种显微图像的目标细胞提取效果较好。此方法通过寻找二值图像的最大轮廓精确定位藻种细胞的位置并提取出主要细胞,从而去除藻种显微图像中的泥沙、噪声、杂质等干扰物。该工作为下一步的分类识别奠定了基础。2、分形编码算法耗时是导致其实时性较差的一个主要原因,有可能它在压缩速度上的局限已经抵消了分形编码算法的优越性。针对这一弊端,我们利用基于方差的改进分形编码方法来减少值域块寻找最佳匹配块的搜索量,此方法提高了编码速度并改善了解码图像质量。然后对使用此编码方法形成的分形编码库,进行基于分形邻距的显微图像识别。通过实验证实了本文识别方法的有效性。更多还原

【Abstract】 In recent years, following the aggravation of the ocean environment pollution and water eutrophication, harmful algal blooms occur more and more frequently, which bring serious disaster and loss. It caused more attention by the governments, the public and scientists. Our country arranges researches on formation mechanism, early warning, forecasting, preventing and controlling methods of red tides on different aspects, such as the fundamental research and high technology development, which aims to establish operational monitoring system of red tides. In automatic monitoring of the red tide, identifying the dominant species of red tide rapidly and effectively is an important component.Fractal theory is a very active mathematics branch in non-linear domain, and it is widely applied in image processing in recent years. Fractal encoding has been rapid developed because of its advantage of high compression, meanwhile according to the attractor invariant features of fractal encoding, we can use the fractal neighbor distance for image matching and then achieving the purpose of identification. The fractal neighbor distance is a similarity measure for the encoded images by fractal encoding.This topic comes from the National High Technology Research and Development Program (863 Program)“Research on the diagnosis system for biology of harmful algal blooms”(Grant No. 2006AA09Z178). According to the situation that harmful algal blooms appear in China’s coastal waters, acquiring the multi-viewpoints microscopic images of harmful algal blooms on different growing periods and different angles. Based on fractal theory, this paper studies the identification of the microscopic images of harmful algal blooms after object extraction. This paper mainly contains the following work: 1. In the process of image acquisition, the background of the microscopic images becomes extremely complex because of the algae cells are surrounding of the sand, noise and impurities. It will affect the following automatic identification. Aiming at the disadvantage, proposing an object extraction method for microscopic images based on the Otsu method and maximum contour extraction technology. This method can extract the object cell from the microscopic images well, which is useful for the species without setae. It can separate the object cell information in microscopic images. Besides, locating the cell in microscopic images accurately by extracting the maximum contour to remove the sand, noise and impurities. This work lays the foundation for further identification.2. Fractal encoding algorithm is time-consuming, this disadvantage lead it can’t be widely applied. It is possible that the limitation on the compression speed has been offset by the advantages of fractal encoding algorithm. Aiming at the disadvantage, we adopt the improved fractal encoding method based on the variance to reduce the searching numbers of the range blocks looking for the best matching blocks, thereby improving encoding speed and the quality of the decoded images. Therefore, forming fractal encoding library of the microscopic images by using the improved fractal encoding method, and then making the identification work based on fractal neighbor distance. And the experimental results prove the feasibility of this method.更多还原