The shapes and sizes of teeth are greatly different in different
crowds, and in addition, the problems of
crowding, missing, overlapping and the like also exist, so that the complexity of tooth positioning is increased. At present, tooth segmentation and positioning are mainly based on tooth X-
ray films, and shooting of the X-
ray films in remote areas with backward medical resources is limited. In order to solve the above problems, a deep
convolutional neural network is planned to be used in the research, and teeth are positioned and classified based on intraoral images. The method comprises the following steps: firstly, preprocessing a collected
data set to improve the capturing capability of important features; secondly, a mixed attention mechanism is introduced into the project, and important information in feature maps of different scales is accurately captured; and finally, the convergence and generalization ability of the model are improved through a residual module. The specific innovation points are as follows: 1, a mixed attention mechanism is introduced into a basic framework based on U-Net + + to dynamically adjust the weight of each feature map in the network; 2, the low-level features and the high-level features are connected by adopting skip connection, so that the problem of local
information loss possibly caused by the low-level features and the high-level features can be solved;