Segmentation of Dwarf Rocks Based on Bayesian Hierarchical Mixture Model

Abstract

The image recognition of dwarf rocks is the key to the study of maritime navigation, especially to the navigation of unmanned ships, and the segmentation of images plays an important role in the analysis of dwarf rocks images. Aiming at solving the problem that the image segmentation method of large dwarf rocks must be given the number of segmentations in advance, this paper introduces the complete data likelihood method, hierarchical priori and recognizable priori into the mixture model to establish a Bayesian Hierarchical Mixture Model with a priori weak dependence, and uses the reversible jump Markov chain Monte Carlo algorithm to estimate the model parameters, thus realizing jump sampling in variable- dimensional parameter spaces. In the analysis of Dwarf rocks images, the number of image components obtained in the sample is consistent with the image intensity characteristics, and the sampling results are in agreement with the original image. The results of data analysis show that the Bayesian hierarchical mixture model method can reflect the actual characteristics of the dwarf reef image well and realize the automatic segmentation of the dwarf reef image.