【作者】 杨华东；

【导师】 简淼夫；

【作者基本信息】 南京工业大学 ， 材料加工工程， 2005， 硕士

【摘要】 粉体的颗粒大小与形状对粉体的性质极为重要。因此需要对粉体的大小和形状进行测量。在众多的颗粒测量技术中，显微图像法以其直观和全面的形态信息，以及容易实现的优越性得到了很大的重视。 本文研究了显微图像法颗粒测量系统的四个组成部分，配置了试验平台，由显微镜、CCD摄像头和计算机构成的图像采集系统进行图像采集以及图像的识别和分析处理。 本文建立了在MATLAB平台上用于颗粒分析的显微图像处理系统，进行相应的颗粒图像处理研究。用CCD对载玻片上的颗粒显微摄像，采集的图像进行处理，转化为前背景分离的图像后，利用相应算法实现图像中粘连颗粒的分割，使之成为独立的颗粒系统，然后实现对颗粒的面积、周长等参数的检测，最后通过对所有检测颗粒的特征统计，得出粉体的粒度分布。 本文研究了新的颗粒图像分析方法。对采集的每一幅图中的颗粒信息进行筛选，选取其中信息比较集中，容易提取的区域作为检测处理的对象，并消除掉被边缘切割的颗粒部分，提高图像处理效率，完成多幅图的大量颗粒测量。研究表明，此方法可以快速高效的实现颗粒的测量，在处理速度和处理效果上比传统的对图像中所有颗粒信息提取的测量方法好。 在进行颗粒图像的识别与分析处理的过程中，图像分割是其中最为关键的环节之一。本文针对颗粒图像的特点，进行颗粒自动分割的研究，在分水岭变换的基础上，改进分水岭算法，综合利用距离变化，分水岭变换和形态重构，建立了新的分割算法，很好的解决了颗粒图像的分割。 本文在进行颗粒的统计研究时提出了一个新的方法，是对图像进行栅格扫描，逐步减少颗粒的像素，无需对颗粒进行反复的扫描，就可以得出颗粒的边缘轮廓，然后对颗粒的轮廓进行进一步的紧缩，就可以完成对颗粒用点代替的处理，统计点的个数，完成计数统计任务。为避免对边缘收缩产生误差，采用不同方向上的重复扫描。 本文对颗粒的参数检测时，通过判断边缘像素的边的个数，获得颗粒的周长参数；对颗粒面积的检测则是直接通过统计颗粒中像素的个数，进而通过与单位更多还原

【Abstract】 Particle size and shape are important for the characters of powder. It is necessary to measure the size and shape of particles of powder. In many methods of particle measurement, microimage method is one of the most popular methods for its intuitionistic, comprehensive morphological information and easiness to measure.The four parts of measurement system based on microimage method are studied at first in this paper, and then a series of apparatus for experimentation are assembled. Images can be gotten by microscope with CCD and computer.The system of particle measurement based on microimage method is built to study the image processing with the software of MATLAB. The microimages of particles on the glass slide are collected by CCD camera, and transformed into a new image whose foreground and background are separate. Distribution of powder can be obtained after that all the parameters of area and perimeter of particles have been measured.A new method of particle image analyzing is studied in the paper. The area of every image is chosen as object for process, where information is concentrated and easy to recognize. Those particles which are segmented by the border of image are discarded. The efficiency of image processing is improved through these steps. Contrast with traditional methods of image processing, this method is a speedier processing and has a better result.A new segmentation algorithm is developed in this paper, which is used to segment the overlapping parts in image. Image segmentation is one of the most important steps during particle image recognition and analysis processing. The algorithm based on watershed transform is improved. The distance transform, watershed transform and morphological reconstruction are integrated to segment images commendably.A new method to calculate the number of particles is presented in this paper. The number of pixels of particle is decreased by scanning the image, the border can be更多还原