Mixed tooth position identification method of local oral cavity image, storage medium and device

Abstract

The invention relates to the technical field of oral cavity image recognition, and discloses a mixed tooth position recognition method of a local oral cavity image, a storage medium and a device. The method comprises the following steps: collecting local images of a clinical oral cavity, numbering tooth positions and manually marking category information to form an original data set; replacing tooth position numbers of deciduous teeth in the original data set with permanent tooth numbers corresponding to the same positions, and deleting category information to form a tooth position labeling data set; deleting tooth position numbers in the original data set, and numeralizing category information to form a classification annotation data set; training a tooth position recognition model through the tooth position labeling data set; training a classification identification model through the classification annotation data set; and obtaining a to-be-recognized oral cavity image, recognizing the tooth position numbers of the teeth in the oral cavity image through the tooth position recognition model, and recognizing the types of the teeth in the oral cavity image through the classification recognition model. The method can improve the stability and accuracy of identifying the tooth position from the local oral cavity image.

Description

Technical Field

[0001] The present invention relates to the technical field of oral image recognition, and particularly relates to a method, a storage medium and a device for recognizing mixed tooth positions in local oral images. Background Art

[0002] With the rapid development of deep learning technology in the field of medical image processing, it has been widely used in the diagnosis of oral problems. Existing tooth position recognition methods mostly require obtaining full tooth images of the oral cavity through oral scanning or reconstructing a three-dimensional model of the oral cavity, and using the stable panoramic information of the oral cavity to accurately identify individual tooth positions. However, in actual clinical applications, oral observation instruments have gradually become an important auxiliary means for dental diagnosis. However, oral observation instruments can only obtain high-definition images of local teeth, resulting in a lack of a complete dentition structure in a single-frame image, and factors such as posture, occlusion, and perspective changes during imaging, making it difficult for existing deep learning models to stably and accurately predict the tooth positions in local oral images obtained by oral observation instruments. Moreover, for children and adolescents, there may be permanent teeth and deciduous teeth in the oral cavity at the same time, that is, there is a situation of "mixed teeth". In existing tooth recognition schemes, usually the algorithm needs to classify each tooth separately. The types of human teeth include 28 types of permanent teeth and 20 types of deciduous teeth, a total of 48 types. Learning the 48 types of tooth categories greatly increases the training difficulty of the model, and may also affect the accuracy and stability of recognition due to feature confusion between categories, further increasing the difficulty of stably identifying tooth positions from images lacking a complete dentition structure. Therefore, it is necessary to design a new tooth position recognition scheme applicable to oral observation instrument images to solve the above technical problems. Summary of the Invention

[0003] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a method for recognizing mixed tooth positions in local oral images, which can improve the stability and accuracy of recognizing tooth positions in local oral images obtained by an oral observation instrument.

[0004] To solve the above problems, the technical solution adopted by the present invention is as follows: A method for recognizing mixed tooth positions in local oral images includes the following steps: Collect a large number of clinical local oral images, and manually label the tooth position numbers and category information of each tooth in the local oral images; Replace the tooth position numbers of deciduous teeth in the labeled local oral images with the corresponding permanent tooth numbers according to the position mapping relationship between deciduous teeth and permanent teeth, and delete the category information of each tooth to form a tooth position labeled data set; Delete the tooth position numbers in the manually labeled local oral images, and numericalize the category information of each tooth to form a classification labeled data set; Train a model with a tooth position annotation dataset to obtain a tooth position recognition model that focuses on the relative position information of teeth; Train a model with a classification annotation dataset to obtain a classification recognition model that focuses on the category information of teeth; Obtain an oral cavity image to be recognized, identify the tooth position numbers of the teeth in the oral cavity image through the tooth position recognition model, and identify the categories of the teeth in the oral cavity image through the classification recognition model.

[0005] Compared with the prior art, the beneficial effects of the present invention are as follows: The tooth position recognition process is split into two processes: tooth position number recognition and tooth type recognition, and recognition models are trained separately. When training the recognition model for tooth position numbers, the manually annotated tooth position numbers of primary teeth are replaced with the tooth position numbers of permanent teeth at the corresponding positions, so that the recognition of a total of 48 types of permanent and primary teeth is simplified to the tooth position recognition of 28 types of permanent teeth and the recognition of 2 types of teeth, namely permanent teeth and primary teeth. Thus, the number of recognition categories of the tooth positions is greatly reduced, the model training difficulty is reduced, the probability of feature confusion between categories is avoided, and the accuracy and stability of tooth position recognition in local oral cavity images lacking a complete dentition structure are improved. At the same time, this method uses manually annotated local oral cavity images to train the tooth position recognition model, which can enable the model to learn the relative position relationships of teeth in local oral cavity images under different postures, perspectives, and occlusions, thereby further improving the recognition accuracy of tooth positions in local oral cavity images obtained by an oral cavity observation instrument.

[0006] The above-mentioned mixed tooth position recognition method for local oral cavity images further includes the steps of: Use the prior knowledge of tooth positions to correct the recognized tooth position numbers with discontinuous spatial positions.

[0007] For the mixed tooth position recognition method of local oral cavity images, the step of training a model with a tooth position annotation dataset to obtain a tooth position recognition model that focuses on the relative position information of teeth includes: Randomly occlude the images in the tooth position annotation dataset to obtain a relative position dataset; Use the relative position dataset to train a model to obtain a tooth position recognition model that focuses on the relative position information of teeth.

[0008] For the above-mentioned mixed tooth position recognition method of local oral cavity images, the step of using the prior knowledge of tooth positions to correct the recognized tooth position numbers with discontinuous spatial positions includes: Establish a tooth position matrix corresponding to the actual spatial positions of each tooth position number; Fill the number of each tooth position number recognized from the oral cavity image into the corresponding position of the tooth position matrix; If there is a situation where a certain tooth position number in the tooth position matrix is repeatedly recognized, then compare the number of recognized tooth positions on both sides of the repeatedly recognized tooth position number, and translate the section with fewer tooth positions by the number of repeated recognitions in the direction of the smaller number of tooth positions, and fill the vacancy generated after translation; If there is a vacancy in the recognized tooth position section in the tooth position matrix, then compare the number of recognized tooth positions on both sides of the vacancy, and translate the section with fewer tooth positions towards the section with more tooth positions to fill the vacant tooth positions.

[0009] A storage medium stores a computer program, and when the computer program is called and executed by a processor, it implements the above-mentioned method for identifying mixed tooth positions in a local oral image.

[0010] A device for identifying mixed tooth positions in a local oral image, characterized in that it includes a processor and a memory, the processor and the memory are electrically connected, and the processor implements the above-mentioned method for identifying mixed tooth positions in a local oral image by calling and executing the computer program in the memory.

[0011] A device for identifying mixed tooth positions in a local oral image includes: an acquisition module for acquiring a local oral image of the oral cavity; a tooth position recognition model for recognizing the tooth position numbers of the teeth in the acquired local oral image, the tooth position recognition model being formed by training with a tooth position annotation data set obtained from local oral images of the oral cavity in which the tooth position numbers of each tooth are manually marked, and replacing the tooth position numbers of the primary teeth in the local oral image of the oral cavity with the corresponding permanent tooth numbers according to the position mapping relationship between the primary teeth and the permanent teeth; a classification recognition model for recognizing the categories of the teeth in the acquired local oral image, the classification recognition model being formed by training with a classification annotation data set obtained from manually marking the category information of each tooth and numericalizing the category information of each tooth.

[0012] The above-mentioned device for identifying mixed tooth positions in a local oral image further includes a correction module, and the correction module is used to correct the recognized tooth position numbers with discontinuous spatial positions by using the prior knowledge of tooth positions.

[0013] In the training process of the above-mentioned tooth position recognition model of the device for identifying mixed tooth positions in a local oral image, first randomly occlude the images in the tooth position annotation data set to obtain a relative position data set, and then use the relative position data set to train the tooth position recognition model.

[0014] The above-mentioned mixed dentition recognition device for local oral images, wherein the correction module includes a tooth position matrix corresponding to the actual spatial positions of each tooth position number, a tooth position filling module, and a tooth position moving module. The tooth position filling module is used to fill the number of each tooth position number recognized by the tooth position recognition model into the corresponding position of the tooth position matrix. If there is a situation of repeated recognition of a certain tooth position number in the tooth position matrix, the tooth position moving module is used to compare the recognized tooth position quantities on both sides of the repeatedly recognized tooth position number, and translate the segment with fewer tooth position quantities by the number of tooth positions of the repeated recognition in the direction of fewer tooth position quantities, and fill the vacancies generated after translation. If there is a vacancy in the recognized tooth position segment in the tooth position matrix, the tooth position moving module is used to compare the recognized tooth position quantities on both sides of the vacancy, and translate the segment with fewer tooth position quantities towards the segment with more tooth position quantities to fill the vacant tooth positions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the method for recognizing mixed dentition of local oral images according to the first embodiment of the present invention.

[0016] Figure 2 It is a flowchart of the method for recognizing mixed dentition of local oral images according to the second embodiment of the present invention.

[0017] Figure 3 It is a schematic block diagram of the device for recognizing mixed dentition of local oral images according to the first embodiment of the present invention.

[0018] Figure 4 It is a schematic block diagram of the device for recognizing mixed dentition of local oral images according to the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following details the embodiments of the present invention. Referring to Figure 1 , the embodiments of the present invention provide a method for recognizing mixed dentition of local oral images, including the following steps: First, a large number of clinical local oral images are collected, and dental experts are asked to manually label the tooth position numbers and tooth categories of each tooth in the local oral images with different angles, different regions, and different occlusion situations, forming an original data set. Then, the tooth position numbers of the primary teeth in the labeled local oral images are replaced with the corresponding permanent tooth numbers at the same position according to the position mapping relationship between the primary teeth and the permanent teeth. For example, the tooth position numbers 51 and 52 of the primary teeth are replaced with the tooth position numbers 11 and 12 of the permanent teeth at the same position, and the category information of each tooth is deleted, forming a tooth position labeled data set H_dataset, which contains the labels of 28 categories of permanent teeth. Meanwhile, delete the tooth position numbers in the marked oral local images, and numericalize the category information of each tooth to form a classification and annotation dataset HR_dataset. In this embodiment, since the main recognition object is the oral cavity of children or adolescents, the classification and annotation dataset HR_dataset mainly focuses on two types of teeth: permanent teeth and deciduous teeth. The annotation of permanent teeth can be numericalized as "1", and the annotation of deciduous teeth can be numericalized as "2".

[0020] Subsequently, train the model with the tooth position annotation dataset H_dataset to obtain a tooth position recognition model HM_model that focuses on the relative position information of teeth; and train the model with the classification and annotation dataset HR_dataset to obtain a classification recognition model HR_model that focuses on the category information of teeth. Finally, obtain the oral image to be recognized. Use the tooth position recognition model HM_model to recognize the position information of each tooth in the obtained oral image, that is, the tooth position number of each tooth, and use the classification recognition model HR_model to recognize the category information of each tooth in the obtained oral image, that is, the permanent and deciduous tooth information. After the recognition is completed, integrate the recognized tooth position numbers and tooth category information of each tooth and output the result to complete the recognition of the tooth position.

[0021] This method splits the tooth position recognition process into two independent subtasks: tooth position recognition and tooth category recognition. And replace the tooth position numbers of deciduous teeth in the tooth position annotation dataset H_dataset for training the tooth position recognition model HM_model with the tooth position numbers of permanent teeth corresponding to the same position. Thus, the 48 tooth categories that the traditional model needs to learn are simplified to the learning of the tooth position numbers of 28 permanent teeth, and the learning of 2 tooth categories: permanent teeth and deciduous teeth, a total of 30 categories, greatly simplifying the number of categories that the model needs to learn, and thus reducing the probability of misrecognition due to feature confusion between categories. Split the tooth position category into a position category and a permanent / deciduous category, and let two models respectively learn the relevant features to exclude the interference of other features. Thus, the tooth position recognition model HM_model can focus more on the learning of the relative position relationship of each tooth position number, and the category recognition model HR_dataset can focus more on the learning of the morphological features of each tooth to improve the recognition accuracy and stability of the two models. By using the manually marked oral local images to construct the dataset for model training, the model can also learn the relative position relationship and pose features of teeth in the oral local images under different poses, perspectives, and occlusions, so that the trained model can better adapt to the oral local images and can more accurately and stably recognize the positions and categories of each tooth in the image in the absence of complete dentition structure information, further improving the recognition stability and accuracy.

[0022] It can be understood that in some possible embodiments, the missing tooth annotation values in the classification annotation dataset HR_dataset can also be numerically labeled as "0".

[0023] Referring to Figure 2 , in some possible embodiments, since the number of occlusion situations in the clinically collected local oral cavity is limited, in order to enable the tooth position recognition model HM_model to pay more attention to the relative position relationship between each tooth position number, so that the model can infer the tooth position number based on the surrounding information of the teeth, such as inferring the tooth position number of the occluded tooth based on the position relationship between the adjacent tooth position number and the reference features inside the oral cavity such as the tongue, it is necessary to randomly occlude the images in the tooth position annotation dataset H_dataset to form a relative position dataset HM_dataset, and then use the relative position dataset HM_dataset after random occlusion to train the tooth position recognition model HM_model. In this embodiment, each time half of the teeth are randomly occluded. For example, if there are three teeth in an image, one of the teeth is randomly set to zero in pixels, but the tooth position number marked for the tooth needs to be maintained.

[0024] Referring to Figure 2 , in some possible embodiments, in order to further improve the accuracy of the recognized tooth position numbers, after the tooth position numbers of each tooth in the obtained oral cavity image are recognized by the tooth position recognition model HM_model, the recognized tooth position numbers with discontinuous spatial positions can also be corrected through the prior indication of the tooth positions. Specifically, a tooth position matrix as shown in the following table is established according to the actual spatial position relationship of the tooth position numbers of each permanent tooth: Table 1 Original Tooth Position Matrix

[0025] Among them, the grids above and below each tooth position number are used to fill in the number of the corresponding tooth position number recognized in the oral cavity image by the tooth position recognition model HM_model. In practice, the tooth position matrix can be set in a storage medium in the form of a two-dimensional array, and the number of each tooth position number recognized by the tooth position recognition model HM_model is stored in the element with the corresponding subscript in the two-dimensional array according to the mapping relationship between the tooth position number and the row and column numbers of the tooth position matrix. Assume that the tooth position matrix is set in the storage medium as an array HM[a][b], where a represents the row of the tooth position matrix, taking values of 0 or 1, and b represents the column of the tooth position matrix, taking positive values from 0 to 13. If the tooth position recognition model HM_model recognizes that there is 1 tooth with the tooth position number 11 in the oral cavity image, the number "1" is written into HM[0][6].

[0026] In practice, the tooth position recognition model HM_model may recognize the tooth position numbers of two adjacent teeth in an oral image as the same tooth position number, resulting in duplicate recognition of the unique tooth position number, as shown in Table 2; or may misrecognize the tooth position number of a tooth, resulting in a discontinuous tooth position matrix in terms of space after filling, as shown in Table 3.

[0027] Table 2 Tooth position matrix with duplicate recognition

[0028] Table 3 Tooth position matrix with gaps

[0029] When the above situations occur, the recognized results need to be rewritten. The rewriting methods include "squeezing" the number of duplicate recognition positions of the duplicate recognized tooth position numbers to one side; or translating one of the two sections of tooth position numbers with gaps to the other end to fill the gap. In this embodiment, when there is a situation of duplicate recognition, compare the number of teeth on both sides of the duplicate recognized tooth position number, and "squeeze" the tooth position number to the side with fewer teeth. Taking the situation shown in Table 2 as an example, the tooth position numbers of two teeth are both recognized as 42, that is, one tooth with the tooth position number 42 is recognized repeatedly. Since only one tooth is recognized on its left side, move the tooth position numbers on the left side to the left by the number of repeated recognition positions, that is, move the "1" of the tooth position number 43 to the grid of the tooth position number 44, and fill the gap generated after the movement, that is, fill the grid of the tooth position number 43 that becomes "0" again after the movement with "1". After completion, change the number of the duplicate recognized tooth position number "42" to 1. The corrected tooth position matrix is shown in Table 4. When there is a misrecognition resulting in a gap, compare the number of recognized teeth on both sides of the gap, and translate the section with fewer teeth to the section with more teeth to fill the tooth position with a gap. Taking the situation shown in Table 3 as an example, due to misrecognition of the tooth position number, there is a gap at the tooth position number 41. Among them, 2 teeth with the tooth position numbers 42 and 43 are recognized on the left side of the gap, and 3 teeth with the tooth position numbers 31, 32, and 33 are recognized on the right side of the gap. Since the probability of all 3 being misrecognized is lower than the probability of 2 being misrecognized at the same time, move the left side with fewer teeth one position to the right, that is, move the number 1 filled in the grids corresponding to the tooth position numbers 43 and 42 in the tooth position matrix one position to the right, to the grids corresponding to the tooth position numbers 42 and 42, as shown in Table 5.

[0030] Table 4 Corrected tooth position matrix for Table 2

[0031] Table 5 Corrected tooth position matrix for Table 3

[0032] It can be understood that after the tooth position recognition model HM_model and the category recognition model HR_model respectively recognize the tooth position number and the permanent / deciduous category of the teeth in the oral image taken by the oral cavity observation instrument, the recognition results need to be integrated and marked on the oral image to output the recognition results. If the tooth at the same position is recognized as a permanent tooth, the tooth position number recognized by the tooth position recognition model HM_model can be directly output; if the tooth at the same position is recognized as a deciduous tooth, the tooth position number recognized by the tooth position recognition model HM_model needs to be restored to the tooth position number of the deciduous tooth at the same position and then marked at the corresponding position on the oral image. It can be understood that both the tooth position recognition model HM_model and the category recognition model HR_model can adopt neural network models such as Mask R-CNN, Faster R-CNN, or U-Net.

[0033] Based on the same inventive concept, an embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned mixed tooth position recognition method for local oral images can be implemented.

[0034] In some possible implementation manners, each aspect of the mixed tooth position recognition method for local oral images provided by the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a device, the program code is used to cause the control device to execute the steps in the mixed tooth position recognition method for local oral images according to various exemplary embodiments of the present application described above in this specification.

[0035] Refer to Figure 3 , based on the same inventive concept, an embodiment of the present invention further provides an identification device for implementing the above-mentioned mixed tooth position recognition method for local oral images, including a processor and a memory. The memory is electrically connected to the processor, and when the processor executes the computer program stored on the memory, the above-mentioned mixed tooth position recognition method for local oral images is implemented.

[0036] In a possible design, the processor may include one or more processing units. The processor and the memory may be implemented on the same chip or separately on independent chips. The processor may be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method for identifying mixed tooth positions in local oral images disclosed in combination with the embodiments of the present application may be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0037] As a non-volatile computer-readable storage medium, the memory can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. The memory may include at least one type of storage medium, for example, it may include flash memory, a hard disk, a multimedia card, a card-type memory, a random access memory (RAM), a static random access memory (SRAM), a programmable read-only memory (PROM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic memory, a magnetic disk, an optical disk, and so on. The memory is any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in the embodiments of the present application may also be a circuit or any other device capable of implementing a storage function, for storing program instructions and / or data.

[0038] By designing and programming the processor, the code corresponding to the method for identifying mixed tooth positions in local oral images introduced in the foregoing embodiments can be solidified into the chip, so that the chip can execute the steps of the method for identifying mixed tooth positions in local oral images of the embodiments shown in the present invention when running. How to design and program the processor is a well-known technology to those skilled in the art and will not be elaborated here.

[0039] Refer to Figure 4, in some possible embodiments, the recognition device includes an acquisition module, a tooth position recognition model, a classification recognition model, and a correction module. The acquisition module is used to obtain a local oral image of the oral cavity through an oral cavity observation instrument. The tooth position recognition model is used to recognize the tooth position numbers of the teeth in the obtained local oral image. The tooth position recognition model is trained by a tooth position annotation data set formed by local oral images of each tooth manually annotated with tooth position numbers, and replacing the tooth position numbers of the deciduous teeth in the local oral image with the corresponding permanent tooth numbers according to the position mapping relationship between the deciduous teeth and the permanent teeth. The classification recognition model is used to recognize the categories of the teeth in the obtained local oral image. The classification recognition model is trained by a classification annotation data set formed by manually annotating the category information of each tooth and numericalizing the category information of each tooth. The correction module is used to correct the discontinuous tooth position numbers in the recognized spatial positions by using the prior knowledge of the tooth positions.

[0040] In some possible embodiments, in order to improve the accuracy and stability of the tooth position recognition model, during the training process of the tooth position recognition model, the images in the tooth position annotation data set are randomly occluded to obtain a relative position data set, and then the relative position data set is used to train the tooth position recognition model.

[0041] In some possible embodiments, the correction module includes a tooth position matrix corresponding to the actual spatial position of each tooth position number, a tooth position filling module, and a tooth position moving module. The tooth position filling module is used to fill the number of each tooth position number recognized by the tooth position recognition model into the corresponding position of the tooth position matrix shown in Table 1. If there is a situation of repeated recognition of a certain tooth position number in the tooth position matrix, the tooth position moving module is used to compare the recognized tooth position numbers on both sides of the repeatedly recognized tooth position number, and translate the segment with fewer recognized tooth positions by the number of tooth positions of the repeated recognition in the direction of fewer tooth positions, and fill the vacancies generated after translation, that is, correct the tooth position matrix shown in Table 2 above to the tooth position matrix shown in Table 4. If there is a vacancy in the recognized tooth position segment in the tooth position matrix, the tooth position moving module is used to compare the recognized tooth position numbers on both sides of the vacancy, and translate the segment with fewer tooth positions towards the segment with more tooth positions to fill the vacant tooth positions, that is, correct the tooth position matrix shown in Table 3 above to the tooth position matrix shown in Table 5.

[0042] Note that this application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one or more flows and / or blocks Figure 1 one or more flows and / or blocks Figure 1 or a plurality of blocks.

[0043] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in one or more flows and / or blocks Figure 1 one or more flows and / or blocks Figure 1 or a plurality of blocks.

[0044] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more flows and / or blocks Figure 1 one or more flows and / or blocks Figure 1 or a plurality of blocks.

[0045] The above embodiments are only the preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantive changes and substitutions made by those skilled in the art based on the present invention fall within the scope of protection required by the present invention.

Claims

1. A method for identifying the mixed dentition of local oral images, characterized in that, It includes the following steps: Collect a large number of clinical local oral images, and manually label the tooth position numbers and category information of each tooth in the local oral images; Replace the tooth position numbers of deciduous teeth in the labeled local oral images with the corresponding permanent tooth numbers according to the position mapping relationship between deciduous teeth and permanent teeth, and delete the category information of each tooth to form a tooth position annotation dataset; Delete the tooth position numbers in the manually labeled local oral images, and numericalize the category information of each tooth to form a classification annotation dataset; Train a model with the tooth position annotation dataset to obtain a tooth position recognition model that focuses on the relative position information of teeth; Train a model with the classification annotation dataset to obtain a classification recognition model that focuses on the category information of teeth; Obtain the oral image to be recognized, identify the tooth position numbers of the teeth in the oral image through the tooth position recognition model, and identify the categories of the teeth in the oral image through the classification recognition model.

2. The method for identifying the mixed dentition of a local oral image according to claim 1, wherein, It also includes the steps: Use the prior knowledge of tooth positions to correct the tooth position numbers with discontinuous recognized spatial positions.

3. The method for identifying the mixed dentition of a local oral image according to claim 1, characterized in that, The step of training a model with the tooth position annotation dataset to obtain a tooth position recognition model that focuses on the relative position information of teeth includes: Randomly occlude the images in the tooth position annotation dataset to obtain a relative position dataset; Train a model with the relative position dataset to obtain a tooth position recognition model that focuses on the relative position information of teeth.

4. The method for identifying the mixed dentition of a local oral image according to claim 2, wherein, The step of using the prior knowledge of tooth positions to correct the tooth position numbers with discontinuous recognized spatial positions includes: Establish a tooth position matrix corresponding to the actual spatial positions of each tooth position number; Fill the corresponding positions of the tooth position matrix with the quantities of each tooth position number recognized from the oral image; If there is a situation of repeated recognition of a certain tooth position number in the tooth position matrix, compare the recognized tooth quantities on both sides of the repeatedly recognized tooth position number, and translate the segment with fewer tooth quantities by the number of repeatedly recognized tooth positions in the direction of fewer tooth quantities, and fill the vacancies generated after translation; If there are vacancies in the recognized tooth segments in the tooth position matrix, compare the recognized tooth quantities on both sides of the vacancies, and translate the segment with fewer tooth quantities towards the segment with more tooth quantities to fill the vacant tooth positions.

5. A storage medium stores a computer program, characterized in that, When the computer program is called and executed by the processor, it implements the method for mixed tooth position recognition of local oral images according to any one of claims 1 to 4.

6. A mixed dentition recognition device for local oral images, characterized in that, It includes a processor and a memory, the processor and the memory are electrically connected, and the processor realizes the method for mixed tooth position recognition of local oral images according to any one of claims 1 to 4 by calling and executing the computer program in the memory.

7. A mixed dentition recognition device for local oral images, characterized in that, It includes: An acquisition module for acquiring local oral images of the oral cavity; A tooth position recognition model for recognizing the tooth position numbers of the teeth in the acquired local oral images, and the tooth position recognition model is trained by a tooth position annotation dataset formed by local oral images with the tooth position numbers of each tooth manually labeled and replacing the tooth position numbers of deciduous teeth in the local oral images with the corresponding permanent tooth numbers according to the position mapping relationship between deciduous teeth and permanent teeth; A classification recognition model is used to recognize the categories of teeth in the acquired local oral images. The classification recognition model is trained by a classification annotation dataset formed by manually annotating the category information of each tooth and then numericalizing the category information of each tooth.

8. The mixed dentition recognition device for local oral images according to claim 7, characterized in that, It further includes a correction module, which is used to correct the tooth position numbers with discontinuous spatial positions recognized by using the prior knowledge of tooth positions.

9. The mixed dentition recognition device for local oral images according to claim 7, characterized in that, During the training process of the tooth position recognition model, the images in the tooth position annotation dataset are randomly occluded to obtain a relative position dataset, and then the relative position dataset is used to train the tooth position recognition model.

10. The mixed dentition recognition device for local oral images according to claim 8, characterized in that, The correction module includes a tooth position matrix corresponding to the actual spatial position of each tooth position number, a tooth position filling module, and a tooth position moving module. The tooth position filling module is used to fill the quantity of each tooth position number recognized by the tooth position recognition model into the corresponding position of the tooth position matrix. If there is a situation of repeated recognition of a certain tooth position number in the tooth position matrix, the tooth position moving module is used to compare the recognized tooth position quantities on both sides of the repeatedly recognized tooth position number, translate the section with a smaller tooth position quantity by the number of tooth positions of the repeated recognition in the direction of the smaller tooth position quantity, and fill the vacancy generated after translation; if there is a vacancy in the recognized tooth position section in the tooth position matrix, the tooth position moving module is used to compare the recognized tooth position quantities on both sides of the vacancy and translate the section with a smaller tooth position quantity towards the section with a larger quantity to fill the vacant tooth position.

Images