基于灰度共生矩阵的连续多幅散斑图像纹理.pdf

47, 051002 (2010) ?2010 中国激光杂志社 051002-1 doi: 10.3788/lop47.051002 基于灰度共生矩阵的连续多幅散斑图像纹理 郭 冰 周晨波 赵亚伟 王 刚 (烟台大学光电信息科学技术学院，山东 烟台 264005) 摘要 采用激光散斑的实验方法，拍摄了连续多幅不同加工工艺、不同表面粗糙度标准样块的散斑图像。利用基于 灰度共生矩阵的纹理分析方法对图像的 4 个特征参数——角二阶矩、惯性矩、相关性和熵进行了计算、统计与分析。 研究表明，连续多幅激光散斑图像的各种参数取值会有波动，但整体趋于平稳；验证了表面粗糙度和特征参数之间 存在着规律性的曲线关系；进一步消除了外界干扰和装置随时间的动态不确定性，使得实验结果更可靠，同时扩大 了表面粗糙度和 4 个特征参量之间的关系数据库，验证了激光散斑的方法对测量表面粗糙度具有较高的可信度。 关键词 图像处理；纹理；灰度共生矩阵；连续；激光散斑 中图分类号 TN247，O436.2 OCIS 100.2960 290.0290 文献标识码 A Texture of Continuous Image Based on Gray Co-Occurrence Matrix Guo Bing Zhou Chenbo Zhao Yawei Wang Gang (College of Electro-Optical Information, Yantai University, Yantai, Shandong 264005, China) Abstract A large number of continuous speckle images with different processing techniques and different surface roughness are taken using the laser speckle measurement method. Through the texture analysis method based on gray-level co-occurrence matrix, four characteristic parameters of the images, angle of second-order moment, moment of inertia, relevance and entropy are calculated and analyzed. The research indicates that, the parameters values of the continuous laser speckle images are stable with little fluctuation and the regular relationship curves of the parameters and surface roughness are testified and verified. Outside interference and dynamic uncertainty of the device over time are eliminated, making results more reliable. The relation database of the parameters and surface roughness is expanded at the same time, and the laser speckle measurement method having high precision measurement of surface roughness is testified. Key words image processing; texture; gray co-occurrence matrix; continuous; laser speckle 1 引 言 表面粗糙度是表征和评定产品表面形貌的重要参数，对产品的使用性能和寿命有较大的影响[1]，其测 量方法可分为接触式测量和非接触式测量[2]。传统的机械触针法就是接触式测量的主要方法之一，该方法 易损伤被测表面，遗漏表面高频信息，不能很好地满足现代工业高自动化程度对在线测量的要求[3]；非接 触式测量具有非接触、无损、高精度、高灵敏度和实时快速等优点[4