【作者】 郑建冬；

【导师】 伍铁军；

【作者基本信息】 南京航空航天大学 ， 机械设计及理论， 2005， 硕士

【摘要】 目前,在烟叶复烤厂中烟叶叶片结构的分析仍然是采用人工的检测手段。针对这种传统工艺方法进行烟叶叶片结构检测和分类的缺点,介绍了图像处理技术在烟叶叶片结构自动分类系统中的研究与应用。讨论了在分类系统中烟叶图像采集关键硬件的选取和图像数据的传输,而后分析了烟叶图像增强、彩色图像分割、连通区域标记、面积的检测、识别和分类等主要的图像处理算法。提出了一种动态的烟叶叶片面积大小分类的算法——最大方差比算法,它是基于类的分离度思想而提出的。该算法对于不同品质的烟叶图像具有自适应性,且计算准确,速度快,能满足实时性要求。实验结果表明该算法能有效地应用于烟叶叶片结构自动分类系统,并取得了令人比较满意的效果。更多还原

【Abstract】 Today, analyzing the tobacco leaf structure still depends on manpower in the tobaccoindustry. To overcome the flaw of the conventional method of detecting and classifying thetobacco leaf structure, the study and application of image processing in the tobacco leafstructure automatic classification is introduced. Discuss the selection principle of the keyhardware for the image acquisition, which are used in the tobacco classification system.And the transmission of digital image data is presented. Then the details of some mainimage processing algorithms such as image enhancing, color image segmentation,connected component labeling, area detecting, recognizing and classifying are analyzed.This paper presents a algorithm—the maximum ratio of variance, for classifying thetobacco leaf area dynamically. It based on the concept of the scatter of classification. Thealgorithm is adaptive to various quality tobacco images. The algorithm is precious and theprocessing time is decreased. It also runs in real time. Experimental results show that itworks efficiently and stably with the tobacco leaf structure classifying automaticallysystem. The practical results show that it is feasible and satisfactory.更多还原