摘要
针对现有煤岩纹理特征提取采用局部二值模式算法存在分类准确率欠佳、算法运行效率较低及旋转纹理识别鲁棒性较差等缺陷,提出了一种基于变差函数和局部方差图的煤岩纹理特征提取算法。该算法首先在局部二值模式理论框架中逐像素计算局部方差得到局部方差图,然后在局部方差图中利用变差函数计算不同方向的变差函数向量,最后组合变差函数向量作为纹理特征,将所提取特征与局部二值模式特征融合完成煤岩纹理分类与识别。实验结果表明,该算法能够有效地提取局部方差的空间分布信息,实现对局部二值模式丢失信息的再利用,分类结果优于多种经典的局部二值模式纹理特征提取算法,分类准确率达到86%。
In view of problems of low classification accuracy and algorithmic running efficiency and poor robust property of rotation texture recognition existed in local binary patterns for texture feature extraction of coal-rock,a texture feature extraction algorithm of coal-rock image based on variogram and local variance image was proposed.Firstly,local variance image was got by calculating local variance with pixel by pixel in theoretic framework of local binary patterns.Then,the variogram vectors with different direction were calculated by variogram in local variance image.Finally,combination variogram vectors were taken as the texture feature,classification and recognition of texture of coal-rock was realized combining the texture feature and local binary patterns feature.Experiment results show that the algorithm can effectively extract spatial distribution information of the local variance image,realize information reuse missed by local binary patterns,and its classification results are better than other algorithms of texture extraction based on local binary patterns,the classification precision reaches 86%.
作者
黄蕾
郭超亚
HUANG Lei , GUO Chaoya(School of Mechanical Electronic and Information Engineering , China University of Mi ning and Technology （ Bei jing） , Beijing 100083, China)
出处
《工矿自动化》
北大核心
2018年第4期62-68,共7页
Journal Of Mine Automation
基金
国家重点研发计划项目(2016YFC0801800)
国家自然科学基金项目(51674269)
