Image Segmentation of Throat Swab Sampling Based on Mask R-CNN

Abstract

This study aims to provide a feasibility basis for image processing for automatic throat swab robot sampling. For target segmentation and identification in the oral environment, we use a robust instance segmentation algorithm Mask R-CNN. In order to solve the insufficiency of the data used, we use data augmentation methods based on the original data to increase the available effective data. Experiments have proved that good results have been achieved for the segmentation of the throat swab and the posterior pharyngeal wall.