Improved RHGSO-FC Algorithm for GMM Clustering Segmentation of RGB-D Images
With the introduction of low-cost depth image sensors,reliable image segmentation in RGB-D images is a goal of computer vision,and how to segment images of cluttered scenes is a challenging problem.A new RGB-D image seg-mentation method based on the random Henry gas solubility optimization-fuzzy clustering(RHGSO-FC)is proposed.Firstly,the Henry gas solubility optimization(HGSO)algorithm is improved to propose the improved Henry gas solubility optimization(LRHGSO)algorithm,and the initialization labels are generated using the the improved Henry gas solubility optimization-kernel fuzzy clustering(LRHGSO-KFC).The initialized labels are then passed into Gaussian mixture model(GMM)clustering to obtain multiple clustering results.Finally these clustering results are segmented and merged by the aggregating superpixels method to get the final segmentation results.The experimental dataset uses NYU depth V2 indoor images and is compared with some of the existing segmentation methods:threshold segmentation algorithm,hard C means,fuzzy C-means,Gaussian mixture model clustering,kernel fuzzy clustering,fuzzy subspace clustering,chaotic Kbest gravitational search algorithm and random Henry gas solubility optimization algorithm.The results show that the RGB-D segmentation algorithm proposed in this paper outperforms the compared algorithms.
万方数据
维普
