A RESEARCH OF TREE IMAGE MARKOV RANDOM FIELD SEGMENTATION METHOD BASED ON GENETIC ALGORITHM

Under the natural background, the images of the trees have the characteristics of complex background, different colours and shapes of various tree species, and sensitivity to light. Through the analysis of the tree image crown colour, tree trunk colour, and texture features, this paper proposes a tree image Markov random field (MRF) segmentation algorithm based on genetic algorithm (GA). It fully considers the space constraint information and better preserves the texture information and edge information of the target tree. The method divides the image of the tree to be segmented into three types of regions: crown, trunk, and background. It supervises the selection of sample pixel sets of the three types of regions, labels the MRF based on the initial selection information, and then uses the GA to perform global energy optimisation , and finally gets the best solution. Compared with the segmentation results of commonly used optimisation methods, such as iterative conditional model algorithm (ICM), artificial bees colony algorithm (Bees), Gibbs sampling algorithm (Gibbs) and maximum mean difference algorithm (MMD), the experimental results show that the GA's optimised segmentation results have significantly improved the accuracy.