基于PSO-GRG的背散射模式扫描电镜的数字处理及应用

A new method of extracting mineral compositions and pore space from backscatter electron imaging of sandstone based on PSO-GRG algorithm

根据背散射扫描电镜中灰度信息的多元高斯分布特征构造目标函数,应用粒子群-广义简约梯度(PSOGRG)联合优化算法求解岩石不同成分的灰度阈值,结合形态学滤波和图像融合算法得到典型元素图,并定量计算矿物体积分数及孔隙度。结果表明:所得岩石成分信息与氦气法、X衍射法所得结果一致性好,可靠性高;在取柱塞样品困难和样品规则度差的地区具有较好的推广性,可以为定量计算岩石成分信息提供新的手段,充分挖掘背散射模式扫描电镜图像中所蕴含的丰富地质信息。

Based on the multi-Gaussian distribution of grey histograms in backscattering electron images,a hybrid method combing the particle swarm optimization and the generalized reduced gradient algorithms（PSO-GRG） was developed to solve the nonlinear objective function,aiming for obtaining the thresholding values of different constituents of the rock. Typical element skeletons were then obtained by morphological filtering-image fusion methods. Thus,volumes of these solid constituents and the porosity could be extracted conveniently. By comparing the calculation results with porosity of helium method as well as mineral content of X-ray diffraction,it shows that the calculation results are reliable. This method can be widely applied to identify the porosity and mineral components with hard-take or random-shape samples which fully takes the advantage of the rich geological information from backscattering electron imaging.