Edge Fuzzy Clustering by Eliminating Undesirable Features in Garment Texture Image Segmentation

Edge computing allows data to be processed at the nearest termination, reducing the requirement to transfer data back and forth between the cloud during the procedure of online knowledge graph construction for fashion design resources. In this paper, we proposed a DBI driven clustering technique for garment texture image segmentation installed on terminations which has two merits. One is that the technique embeds DBI index into the Shannon entropy term in the objective function to control feature weighting learning such that the learnt feature weights become more accurate. Another merit is that we propose a feature eliminating principle to filter undesirable features to thoroughly reduce the side effects caused by undesirable features. The proposed clustering technique is verified on different kinds of datasets including garment texture images and the experimental results indicate its promising performance.