【作者】 李楠；

【导师】 姚志红； 潘为平；

【作者基本信息】 上海交通大学 ， 电气工程及其自动化（专业学位）， 2013， 硕士

【摘要】 目前，我国对海洋赤潮及湖泊蓝绿藻的监视主要还是卫星遥感技术，这种监测手段的缺陷是不能进行预测、预警，而且实时性差、阴雨天气得不到图像。针对以上问题，及时的预报、预警、监测和控制从而减少环境灾害的影响，就是需要重要解决的问题，而藻类的种类和形态比例的图像识别是监测其中重要的一个环节。根据现场显微照片,利用图像识别软件来鉴定赤潮或水华类型、优势藻类和种群密度以及生长趋势。然而,由于藻类细胞复杂多样且显微图像受光线和藻类细胞颜色的影响,传统的图像分割算法对藻类轮廓的提取难以取得满意的效果。本文利用一种基于局部区域信息的C-V(LCV)模型，应用于藻类细胞显微图像的分割。首先，本文介绍了藻类研究的重要意义以及国内外图像分割技术的研究发展现状，主要对水平集方法及其在图像分割中的应用和进一步扩展进行深入的研究。对各种方法进行了分析和比较，总结了各方法的性能优缺点及存在的问题。针对几种常见的海洋赤潮生物及湖泊、水库蓝绿藻生物细胞本身具有复杂的结构特点，分析了传统的C-V模型所存在的问题，利用一种新的基于局部信息的Local Chan—Vese(LCV)模型，可以在较少的迭代次数内分割灰度不均匀图像，应用于藻类细胞显微图像的分割，取得了很好的效果。在MATLAB环境下，通过大量的图像处理分析，保留藻类的原有信息，最终返回识别出的只含藻类的位图。选出能最好区分藻类细胞的特征组合，供之后分析藻类特性时使用。通过实验对比，显示出LCV模型相对于传统分割方法可以分割灰度均匀或不均匀的藻类图像。更多还原

【Abstract】 At present, the monitoring of red tides on ocean and lake blue greenalgae in our country is mainly rely on satellite remote sensing technology,which can not be forecasting and early warning. In view of the above question,the forecasting, early warning, monitoring and control so as to reduce theinfluence of the environmental disasters is important. And the imagerecognition of algae species and form is the important link of the monitoring.According to the micrograph, identify the red tide or water bloom types,advantages algae, population density and growth trend with the imagerecognition software. However, due to the algae cell is complicated andmicroscopic image is susceptible to the light and the influence of the algalcells color, the traditional image segmentation algorithm on algae contourextraction is difficult to obtain satisfactory results. In this paper a methodbased on a local Chan—Vese(LCV)model, used in algae cell microscopicimage segmentation.First of all, this paper introduces the significance of algae research andthe current situation of the development of algae analysis technology both athome and abroad, mainly to the level set method and its application in imagesegmentation and further expand in-depth study. Various methods areanalyzed and compared, summarizes the advantages and disadvantages ofvarious methods, performance and problems. According to several commonMarine red tide organisms and lakes, reservoirs blue green algae biologicalcell structure characteristics, analyzed the traditional C-V model the existingproblems. Using a new based on Local information of the Local Chan-Vese(LCV) model, can be in less iterations internal division gray uneven image, applied to algae cell microscopic image segmentation, and good results havebeen achieved.In the MATLAB environment, through a large number of imageprocessing and analysis, retain the original information, algae eventuallyreturn to identify only contain algae bitmap. Choose the best distinguish algaecell characteristics combination, for later analysis using algae characteristic.Through the contrast, show LCV model compared with the traditionalsegmentation method can be partitioned gray uniform or non-uniform algaeimage.更多还原