【作者】 陈瑶；

【导师】 徐丹；

【作者基本信息】 云南大学 ， 计算机技术， 2011， 硕士

【摘要】 近年来,随着数字图像处理技术的发展,人们在数字图像处理技术的应用领域获得了喜人的成果。据文献记载,人们已经可以实现基于内容的图像检索以及对染色体图像的分类等等。论文设计与实现了一种活性污泥常见原后生动物图像识别的方法。对于一幅包含原后生动物和杂质的图片,首先要从中提取出原后生动物的主体,本文中介绍了两种提取方法,即全自动型目标提取算法与交互式目标提取算法。交互式目标提取是指输入一幅数字图像后,需要用鼠标标识出图像的目标与背景,而全自动目标提取方法则不需要。接着进行特征提取,本文从下面三个方面一共提取了16种特征：(1)形态特征,包括面积周长等,从这些特征中可以判断出这个原后生动物大致的形状,比如说类圆或类矩形等；(2)纹理特征,包括对比度、二阶矩等,从这些特征中可以得到这个原后生动物的表面结构组织排列信息以及它们与周围环境的联系；(3)其它特征,包括是否有尾巴、物体的均匀性等,从这些特征中可以得到这个原后生动物是否有尾巴以及它们的均匀情况。最后论文根据前面提取的特征,采用决策树的方法来进行分类。决策树算法的关键是给树的每个节点定义一个衡量标准。例如,在判断某个原后生动物的形状是否接近矩形时,可以用它的外接矩形与它自己本身的面积求商,如果这个值接近1表明它是接近矩形的。利用上述方法,论文针对活性污泥中常见的几种原后生动物,包括草覆虫、轮虫、线虫、豆虫、累枝虫和钟虫等进行了有效地识别。更多还原

【Abstract】 In recent years, with the development of the digital image processing, applications of digital image processing technology have been received satisfactory results. According to historical records, content-based image retrieval system has been invented and chromosome images can be classified and so on.In this paper, we introduce an image recognition system for common protozoan and metazoan in activated sludge. For a picture containing protozoan, metazoan and impurities, first of all we have to extract the main body of the protozoan and metazoan, in this paper, we introduced two extraction methods——automatic target extraction method and interactive target extraction method, interactive target extraction methods is after we enter a digital image, we need to use the mouse to identify the target and background, while we do not need to do it in automatic target extraction method. Then, feature extraction, in this paper, we extracted a total of 16 kinds of features according to the following three aspects:(1)Morphological characteristics, including area, perimeter, etc, from these features we can determine the approximate shape of the protozoan and metazoan. For example, round or rectangular, and so on. (2)Texture features, including contrast, second moment, etc, from these features, we can get the protozoan and metazoans’surface structure arrangement information and their links with the surrounding environment. (3)Other characteristics, including whether it has the tail, and its uniformity, etc. From these characteristics, we can know whether the protozoan and metazoan has the tail and its uniform situation. Finally, with the extracted features, we use Decision Tree approach to classification. Decision Tree algorithm is to focus on defining a measure to each node of the tree. For example, when we determine whether the protozoan and metazoan’shape close to rectangular, we can use an external rectangular area divided by its own area of its own, and if this value is close to 1 indicates that it is close to rectangular. Detail steps will be described in the following article. Using the above methods, this paper focused on several common protozoan and metazoan in activated sludge, including paramecium, rotifer and nematode and so on, conducted effectively identify.更多还原