Research And Application Of Intraoral Scanning Data Processing Technology Based On Image Recognition

In the past ten years,tooth segmentation and recognition algorithms based on deep learning have developed rapidly,and have become research hotspots in medical CT and other fields,but there are few studies and data sets for common scenarios.Therefore,this thesis focuses on the processing technology of intraoral scan data based on image recognition,including tooth image segmentation technology in preprocessing of intraoral scan data,postprocessing segmentation technology of intraoral scan 3D model data,tooth model recognition technology,and strongly related business Digital product digital oral impression instrument.The main work contents of this thesis include:(1)The image processing method is used to segment and recognize the 3D tooth model,and a seed point expansion tooth segmentation algorithm based on YOLOv5 is proposed.The process of manually selecting seed points in the tooth segmentation algorithm is optimized and improved,and the YOLOv5 network model is used to complete the target detection operation of the range box of each tooth,and the automatic marking of seed points is realized to improve the efficiency of the algorithm.The tooth segmentation algorithm based on deep learning can complete the tooth image segmentation in the preprocessing of the intraoral scan data and the postprocessing segmentation of the 3D model data of the intraoral scan,project the 3D model into an image from a suitable perspective for segmentation,and act on the patient’s tooth before and after the reconstruction of the 3D model Expect;(2)Further use CNN to automatically identify teeth and quickly obtain relevant information of patients.In this thesis,the three-dimensional model of the tooth is rotated to different positions to obtain the tooth view image at a suitable angle,and then the EfficientNet classification model is used to complete the tooth recognition task,which can help the dentist quickly understand the patient’s medical history,medication history,etc.,and help make correct decisions.Diagnosis and treatment plan,thereby improving work efficiency;(3)Separately make a data set for tooth segmentation and recognition and mark it well.Before the tooth 3D model is segmented,you need to use the MeshLab tool to import the tooth 3D model and project and export the 2D image.After establishing the connection between the two coordinate systems Then split.Two experiments on tooth segmentation and recognition based on deep learning were completed in common scenarios,and the specific operation steps of the two experiments were formulated respectively,and the effectiveness of the algorithm was verified by using relevant evaluation criteria;(4)Using the tooth segmentation and recognition algorithm based on deep learning in Chapter 3,combined with QT’s cross-platform software development platform,a digital oral impression instrument software system was developed using C/C++ and Python languages.The client includes core functions such as real-time scanning of 3D models,tooth segmentation and cutting,and tooth model recognition,and the server includes core functions such as device management and registered host management.