金刚石砂轮磨粒光学显微层析图像分析与识别方法研究

发布时间：2018-06-08 12:23

本文选题：金刚石磨粒 + 改进的Otsu　；参考：《华侨大学》2017年硕士论文

【摘要】：金刚石磨粒作为参与磨削加工的主要材料,是砂轮的重要组成部分,其在金刚石砂轮空间的分布特性直接影响磨削加工结果。通过对金刚石的分布状态、出露高度、形状特征等信息进行研究,并预测磨粒均匀磨损量与破碎深度,能够对金刚石砂轮磨削性能进行评估。因此,对金刚石砂轮磨粒进行分析和形状识别,有利于指导磨削加工工艺,优化砂轮磨削性能。本文基于光学显微层析测量技术,采用聚焦合成在位测量仪获取不同深度的磨粒层析图像,通过对金刚石砂轮磨粒的光学显微层析图像的分析,提取磨粒特征参数,实现磨粒形状的识别和分类。主要的研究工作包括:(1)磨粒图像预处理。分析不同深度的磨粒层析图像,针对不同因素引起的磨粒图像降质,研究不同的图像预处理方法。采用基于MSRCR与直方图均衡化结合的图像增强算法增强磨粒区域,获取磨粒较好的区域信息;(2)磨粒图像分割。基于直方图阈值,提出一种粒子群优化改进的Otsu自适应阈值分割算法,实现对不同层面破碎磨粒的分割;基于Hough直线检测与最小二乘法区域拟合,对分割后的磨粒区域进行区域合并;(3)磨粒特征提取与识别。研究单颗磨粒几何特征提取方法,采用小波描述子表示不同磨粒轮廓特征,并以此作为后续支持向量机磨粒形状识别的特征输入;分析多颗磨粒间的方位信息,获取单颗磨粒不同深度截层的特征参数;采用支持向量机分类器实现不同磨粒形状的识别与分类。实验结果表明,本文所研究的针对光学显微层析图像分析和识别的算法,较好的实现对金刚石砂轮磨粒的特征提取和识别分类。
[Abstract]:As the main material to participate in grinding, diamond abrasive particles are an important part of grinding wheel. The distribution characteristics of diamond abrasive particles in diamond wheel space directly affect the grinding results. By studying the information of diamond distribution, appearance height and shape characteristics, and predicting the uniform wear quantity and crushing depth of abrasive particles, the grinding performance of diamond wheel can be evaluated. Therefore, the analysis and shape recognition of diamond wheel abrasive particles are helpful to guide grinding process and optimize grinding performance. In this paper, based on the optical microscopic tomography technique, the wear particle tomography images of different depths were obtained by focusing synthetic in-situ measuring instrument. The characteristic parameters of the abrasive particles were extracted by analyzing the optical microscopic tomography images of diamond grinding wheels. The recognition and classification of wear particle shape are realized. The main research work includes: 1) image preprocessing of wear particles. The wear particle tomography images with different depth were analyzed and different image preprocessing methods were studied according to the degradation of wear particle images caused by different factors. The image enhancement algorithm based on MSRCR and histogram equalization is used to enhance the wear particle area, and the better region information is obtained. Based on histogram threshold, an improved Otsu adaptive threshold segmentation algorithm based on particle swarm optimization (PSO) is proposed to realize the segmentation of broken particles in different layers, which is based on Hough line detection and least square region fitting. The feature extraction and recognition of the segmented abrasive region are carried out. The geometric feature extraction method of single wear particle is studied. The wavelet descriptor is used to represent the contour features of different wear particles, which is used as the feature input for the shape recognition of wear particles in the subsequent support vector machine, and the azimuth information between several grains is analyzed. The characteristic parameters of different depth cut layers of single grain are obtained, and the support vector machine classifier is used to realize the recognition and classification of different wear particle shapes. The experimental results show that the algorithm of image analysis and recognition based on optical microscopic tomography is better for the feature extraction and classification of diamond grinding wheel abrasive particles.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG580.619.3;TP391.41

【相似文献】