基于机器视觉的寄生虫卵显微图像自动识别研究

发布时间：2019-02-15 20:41

【摘要】：随着计算机技术的发展及交叉学科在各领域的广泛应用,基于机器视觉的显微图像自动识别技术也在医学领域中得到迅猛发展,不仅为临床诊断提供了一个可靠高效的工具,同时也为医学科研和教学带来了方便。寄生虫卵检测是医学中重要检测项目之一,传统寄生虫卵检测大多采用人工镜检方式,不仅效率低,工作强度大,而且受操作人员主观影响较大,无法保存完整的病例资料,不利于临床动态观察。基于机器视觉的自动检测不仅效率高,无污染,而且适应现代医疗信息化发展的需求,因此本文在前人对寄生虫卵、细胞显微图像识别研究的基础上,着重研究了基于机器视觉的寄生虫卵显微图像采集到识别的过程。具体研究内容如下:1.着重介绍了基于机器视觉的寄生虫卵显微图像自动识别系统的硬件结构设计,包括虫卵标本的自动采集装置、显微镜的自动拍摄装置、照明系统、CCD相机、图像采集卡等。2.针对寄生虫卵显微图像高噪声的特点,研究图像预处理及分割方法,并提出了一种改进的二维最大熵阈值遗传算法结合形态学分割图像的优化算法,能准确快速的实现寄生虫卵显微图像的除噪分割。3.详细介绍了寄生虫卵显微图像的几种常见特征类型,并研究提取各类特征的方法,找到一组最佳的特征组合作为寄生虫卵显微图像的输入量来表征各类典型寄生虫卵显微图像的特点。4.研究几种智能识别算法,如BP(Back Propagation)神经网络,SVM(Support Vector Machine),针对提取的寄生虫卵显微图像特征,分别设计了基于BP神经网络的分类器和基于SVM分类器,并且统计了各自识别的精确度,然后将这两种分类器的识别率作对比,从中选出高精度高效率的分类器。通过本文的研究和试验,证明基于机器视觉的寄生虫卵显微图像自动识别技术是可行的、有效的,在医学上具有一定的实用价值。
[Abstract]:With the development of computer technology and the wide application of interdiscipline in various fields, the automatic recognition technology of microscopic image based on machine vision has been developed rapidly in the field of medicine, which not only provides a reliable and efficient tool for clinical diagnosis. At the same time, it also brings convenience to medical research and teaching. The detection of parasite eggs is one of the most important items in medicine. The traditional examination of parasite eggs is mostly by artificial microscope, which is not only low efficiency and high working intensity, but also can not preserve the complete case data due to the subjective influence of the operators. It is unfavorable to clinical dynamic observation. The automatic detection based on machine vision is not only efficient and pollution-free, but also adapt to the needs of modern medical information development. Therefore, this paper is based on the research of microscopic image recognition of parasite eggs and cells. The process of parasite egg microscopic image acquisition and recognition based on machine vision is studied. The specific research contents are as follows: 1. This paper mainly introduces the hardware structure design of the parasite egg microscopic image recognition system based on machine vision, including the automatic collection device of egg specimen, the automatic shooting device of microscope, the illumination system, the CCD camera, the image acquisition card and so on. In view of the high noise characteristics of parasite egg microscopic image, the image preprocessing and segmentation methods are studied, and an improved two-dimensional maximum entropy threshold genetic algorithm combined with morphological image segmentation optimization algorithm is proposed. Can accurately and quickly realize the parasite egg microscopic image denoising segmentation. 3. In this paper, several common characteristic types of parasite egg microscopic image are introduced in detail, and the methods of extracting all kinds of characteristics are studied. To find out a group of best feature combinations as the input amount of parasite egg microscopic image to characterize the characteristics of various typical parasite egg microscopic images. 4. This paper studies several intelligent recognition algorithms, such as BP (Back Propagation) neural network, SVM (Support Vector Machine), which designs classifiers based on BP neural network and SVM classifier, respectively, according to the features of extracted parasite egg microscopic images. The recognition accuracy of the two classifiers is analyzed, and the recognition rates of the two classifiers are compared to select the high precision and high efficiency classifiers. Through the research and experiment in this paper, it is proved that the automatic recognition technology of parasite egg microscopic image based on machine vision is feasible and effective, and has certain practical value in medicine.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】