Research On Segmentation Method Of Rock Slice Image Based On Clustering

In the field of geology,quantitative analysis of mineral grains in rock slice images is a fundamental task in research,and rock slice image segmentation is the first step in quantitative analysis.The result of segmentation affects the analysis of mineral composition and grains size of rock slices,which in turn affects subsequent tasks such as geological exploration and oil development.The traditional segmentation of mineral grains mainly relies on the experience of geological professionals,which not only wastes human resources and consumes a lot of time,but also results in uneven segmentation results due to different personnel experience.The research on the automatic segmentation method of mineral grains in rock slice images can effectively reduce manpower and improve the segmentation accuracy,which is of great significance for the quantitative analysis of rock slices.This paper focuses on the clustering method for rock slice image segmentation.The main research contents and achievements are summarized as follows:(1)Aiming at the incomplete segmentation of plane-polarized image or cross-polarized image of rock slice,a joint segmentation method of rock slice image under different polarization systems is proposed.Firstly,the chromaticity information of the CB color space is intuitive and has little correlation with brightness,and the plane-polarized image and the cross-polarized image of the rock are decomposed to improve the chromaticity difference of the images;Then using bilateral filter is to smooth the image to reduce the noise in the chrominance information of the image.According to the difference between the chromaticity information of the mineral particles,the difference image is obtained to reduce the influence of the hetero-base and cement around the mineral particles;Then,the differential image is pre-segmented based on the K-means clustering algorithm,and most of the mineral particles are segmented;Finally,the watershed algorithm based on distance transformation is used for secondary segmentation to solve the adhesion problem of mineral particles to a certain extent.The experimental results show that this algorithm alleviates the incomplete segmentation of mineral particles to a certain extent,and outperforms the comparison algorithm in terms of particle accuracy,particle recall and average intersection ratio,with each index reaching more than 80%.(2)In order to further solve the problem of the adhesion of mineral grains,an improved superpixel segmentation method is proposed by fusing edge features on the basis of chromaticity features.Firstly,using the polarized images reconstruct feature to generate a new image,so as to reduce the contrast between mineral particles and increase the contrast with cements,hetero-bases,etc.Secondly,using the multiscale mathematical morphology filter extracts the edge features of the image to reduce the influence of noise;Considering the color features,spatial features and edge features of the image,the distance metric function of the simple linear iterative clustering algorithm is updated to enhance the superpixel boundary adhesion;Finally,the fuzzy clustering algorithm is used to merge the superpixel blocks,which further alleviates the adhesion problem between mineral grains.The experimental results show that the algorithm is superior to the comparison algorithm in terms of particle accuracy,particle recall and average cross-union ratio,with each index reaching more than 85%.