基于岩石薄片偏光序列图的颗粒分割算法

PARTICLE SEGMENTATION ALGORITHM BASED ON POLARIZED LIGHT SEQUENCE IMAGES OF PETROGRAPHIC THIN SECTION

岩石薄片中矿物颗粒纹理丰富,颗粒间存在填隙物,在单偏光下颗粒边缘模糊难以辨别。为了解决岩石薄片颗粒分割的难题,提出一种新的颗粒分割算法。算法以岩石薄片固定视域的偏光序列图为基础,该偏光序列图包含一张单偏光图和多张连续变化偏光角度的正交偏光序列图。算法先利用正交偏光下矿物颗粒较明显的边缘轮廓进行分割。再根据矿物颗粒在单偏光图像中的颜色特征以及在正交偏光序列图像中的亮度变化趋势对过分割颗粒进行修复。经实验表明,该算法具备较好的颗粒分割效果,为岩石薄片的颗粒分割提供了新的思路。

The mineral particles in the petrographic thin section are rich in texture and there are interstitial fillings among particles,which make it difficult to discern the grain boundary in single-polarized light. In this paper,a new algorithm is presented to solve the problem of grain segmentation. It is based on the polarized light sequence images in a stationary view of petrographic thin section,which contains a single-polarized light pattern and a plurality of crosspolarized light pattern of continuous angle changes. The image segmentation operations made full use of obvious edge contour information of mineral particles in cross-polarized light. The over-segmented particles were repaired by the color features in the single-polarized image and the trend of brightness in the cross-polarized sequence image. The experimental results indicate that this algorithm achieves better effect in particle segmentation,which provides a new idea for the segmentation of mineral particles in the petrographic.