Vision-based Toothbrush Quality Evaluation

Abstract

The toothbrush is of key importance for oral health of human beings. Therefore it is necessary to make an accu-rate evaluation of toothbrush quality, which should include its cleaning ability and durability. In this paper, a vision-based toothbrush quality evaluation system is proposed, which consists of a tartar simulant measurement module and a bristle estimation module. In the tartar simulant measurement module, the cleaning ability of a toothbrush can be evaluated by comparing the variation of tartar simulant after tooth brushing. The tooth is firstly detected with deep learning method. Secondly, the tartar simulant is extracted by the combination of both thresholding and intensity-based denoising. Thirdly, the tartar extraction result is refined by memory-based denoising. Finally, the region of tartar is precisely measured and compared during the brushing process. In the bristle estimation module, the durability of a toothbrush may be estimated by its bristle deformation after the specified amount of tooth-brushing. The tooth bristle is quickly extracted by thresholding in a low-resolution tooth-brush bristle image. The boundary of the toothbrush bristle is accurately detected in a high-resolution sub-image with the approach of maximal connected components and a hole-filling method. The toothbrush bristle deformation is estimated based on the accurate segmentation of toothbrush bristle. Experiments applied on collected practical cases demonstrate that our system achieves desired performance for toothbrush quality evaluation in practical industrial scenarios.