基于改进模糊C均值聚类和区域合并的矿物颗粒分割方法

岩矿石薄片图像中矿物颗粒的分割提取是分析鉴定工作的前提，为了进一步提高矿物颗粒提取的准确性，提出一种新的矿物颗粒提取分割方法。该方法以一组岩矿石薄片正交偏光序列图作为输入，对每张正交偏光序列图采用改进的模糊 C 均值聚类(FCM)算法进行初始的边缘提取，并叠加多张边缘图像，再对得到的叠加边缘图像进行精细化处理并进行目标颗粒筛选，得到初始的分割提取图像。为了抑制颗粒的过分割现象，利用基于区域邻接图(RAG)的合并算法进行区域合并，首先依据颗粒的颜色纹理等特征进行初始合并操作，再根据矿物颗粒在不同正交偏光角度下的变化规律进行二次合并操作，从而得到最终的分割提取图像。通过对多组图像进行实验表明，该方法可以获得良好的效果。

MINERAL PARTICLE SEGMENTATION ALGORITHM BASED ON IMPROVED FCM AND REGION MERGING

Segmentation and extraction of mineral particles in rock ore slice images is the premise of analysis and identification. In order to further improve the accuracy of mineral particle extraction, a new method of mineral particle extraction and segmentation is proposed. A set of orthogonal polarized light sequence images of rock ore slice are token as input in this method.Firstly,an improved FCM algorithm is used in each orthogonal polarized light sequence image for initial edge extraction.Secondly, multiple edge images are superimposed.Thirdly, the resulting superimposed edge images are refined and target particles are filtered to obtain The initial segmentation image.Finally, In order to suppress the over-segmentation of the particles, the RAG-based merge algorithm is used to perform region merge. First,an initial merging operation is performed according to the color and texture characteristics of the particles, and then a secondary merging operation is performed according to the change rule of the mineral particles under different orthogonal polarization angles, so as to obtain the final segmentation image. Experiments on multiple sets of images show that this method can obtain good results.