煤岩微观图像的K-均值和数学形态学分割算法研究

摘要

利用SEM扫描观测平煤十矿戊10-20230工作面煤岩的微观形态,针对灰度区域不明显的煤样SEM图像,探究K-means和数学形态学中先开后闭运算相结合分割算法的可行性与有效性,并采用相关分割算法作对比试验.结果表明:本文算法匹配率75.48％,误分率16.96％,相较于其他4种算法匹配率最高,误分率最低,有明显优势,能够较准确地将孔隙从扫描图像中分割出来;本算法TPF值34.79％,TNF值97.18％,正确分割比例最大,而FNF值65.21％,FPF值2.82％,错误分割比例最小,TPF值与FNF值和为1,TNF值与FPF值和也为1,说明评价指标的正确性.%We use SEM scanning to observe the micro morphology of coal and rock in V10-20230 coal face of Pingdingshan No.10 coal mine.For the SEM images of coal samples without obvious gray area,we explore the feasibility and effectiveness of the first open and then close operation segmentation algorithm in K-means and mathematical morphology,and adopt relevant segmentation algorithm for contrast test.The results show that the matching rate of the algorithm in this paper is 75.48％,and the classification error rate is 16.96％.Compared with the other 4 algorithms,this algorithm enjoys obvious advantages for its higher matching rate and lower classification error rate,therefore,it can segment the pores from the scanned images more accurately.Through this algorithm,when the TPF is 34.79％ and the TNF is 97.18％,the correct segmentation ratio is the highest.When the FNF value is 65.21％,and the FPF value is 2.82％,the error segmentation ratio is the lowest.When the sum of the TPF value and the FNF value is 1,and the sum of the TNF value and the FPF value is also 1,the evaluation index is correct.