A children's morning health check oral image communication system

Abstract

The present invention relates to the technical field of medical image processing, and particularly relates to a children's morning physical examination oral image communication system. The system includes: an acquisition module, which is used to determine the tooth area by referring to the grayscale image and the oral grayscale image; a standard model construction module, which is used for image feature analysis, extracting edge and corner features, and constructing a standard model; a corner analysis module, which is used to perform position matching on the comparison corners and the standard corners in the standard model to determine the comprehensive offset of all comparison corners, and screen out abnormal images; an edge analysis module, which is used to perform gradient direction analysis of the edges to determine the abnormal comparison edges of the abnormal images; and a communication module, which is used to label and transmit the abnormal comparison edges of all abnormal images. The present invention can improve the accuracy of abnormal tooth area image analysis and the accuracy of auxiliary assessment of children's individual tooth health.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a children's morning check oral image communication system. Background Art

[0002] Currently, in order to ensure the dental health of children, morning checks are usually arranged to conduct routine medical examinations on the dental conditions of children. By taking pictures of children's teeth and analyzing them, medical images are obtained, and thus abnormal problems such as tooth dislocation and loss, dental calculus, and tooth decay are determined based on the medical images.

[0003] In the related art, by matching the tooth images of children with a standard template to determine abnormal tooth images. In this way, due to the large differences in tooth structures among children of different ages and different individuals, it will interfere with the medical detection of abnormal tooth areas, resulting in low accuracy in the analysis of abnormal tooth images and low accuracy in the auxiliary assessment of the dental health of individual children. Summary of the Invention

[0004] In order to solve the technical problems that due to the large differences in tooth structures among children of different ages and different individuals, it will interfere with the medical detection of abnormal tooth areas, resulting in low accuracy in the analysis of abnormal tooth images and low accuracy in the auxiliary assessment of the dental health of individual children, the present invention provides a children's morning check oral image communication system, and the specific technical solutions adopted are as follows:

[0005] The present invention proposes a children's morning check oral image communication system, including:

[0006] An acquisition module, configured to acquire a reference grayscale image of a child's historical normal teeth and an oral grayscale image during morning check, and determine the tooth areas in the reference grayscale image and the oral grayscale image;

[0007] A standard model construction module, configured to perform image feature analysis on the tooth areas in the reference grayscale image and the oral grayscale image respectively, extract the reference edges and reference corner points in the reference grayscale image and the comparison edges and comparison corner points in the oral grayscale image at the same scale, and construct a standard model for each child's teeth based on the reference edge curvature, the position distribution characteristics of the reference corner points, and the relationship between the reference corner points and the overall tooth morphology. The standard model includes the standard corner points and standard edges of each tooth;

[0008] A corner point analysis module, configured to perform position matching between the comparison corner points and the standard corner points in the standard model to determine the comprehensive offset of all comparison corner points; and screen out abnormal images from the oral grayscale image according to the comprehensive offset;

[0009] An edge analysis module, configured to determine the direction consistency of the comparison edge in each abnormal image according to the gradient direction of the comparison edge and the standard edge obtained by matching in the abnormal image; and determine the abnormal comparison edge of the abnormal image according to the direction consistency.

[0010] A communication module, configured to label and transmit the abnormal comparison edges of all abnormal images.

[0011] Further, the determination of the tooth regions in the reference grayscale image and the oral grayscale image includes:

[0012] Performing semantic segmentation on the reference grayscale image and the oral grayscale image respectively to determine the tooth regions.

[0013] Further, performing image feature analysis on the tooth regions in the reference grayscale image and the oral grayscale image respectively, and extracting the reference edges and reference corner points in the reference grayscale image and the comparison edges and comparison corner points in the oral grayscale image at the same scale, including:

[0014] Normalizing the reference grayscale image and the oral grayscale image based on a geometric calibration method to align the tooth regions in the reference grayscale image and the oral grayscale image to a unified scale;

[0015] Performing Canny edge detection on the reference grayscale image to obtain reference edges, and the reference edges form a reference edge image, and then performing corner point detection to obtain reference corner points;

[0016] Performing Canny edge detection on the oral grayscale image to obtain comparison edges, and the comparison edges form a comparison edge image, and performing corner point detection on the comparison edge image to obtain comparison corner points.

[0017] Further, based on the reference edge curvature, the position distribution characteristics of the reference corner points, and the relationship of the reference corner points in the overall tooth shape, constructing a standard model for each child's tooth, including:

[0018] Determining the curvature of the reference edges, the position distribution of the reference corner points on the reference edges, and the distribution law of the reference corner points on the tooth contour in all reference grayscale images of the same child's tooth as the construction information;

[0019] Inputting the construction information into a preset tooth image model, and performing model training on the preset tooth image model to obtain a standard model, where the standard model is a standardized image model corresponding to the child's tooth region.

[0020] Further, the position matching of the comparison corner points and the standard corner points in the standard model to determine the comprehensive offset of all comparison corner points includes:

[0021] Map the positions of each pair of comparison corner points in the oral cavity grayscale image to the standard model at the same scale, perform model matching based on the positions of the comparison corner points and the standard corner points, and match the comparison corner point with the closest position to the standard corner point as a pair of matching points;

[0022] Take the mean of the Euclidean distances of all pairs of matching points as the first offset index;

[0023] Calculate the quantity difference between all comparison corner points and the comparison corner points among the matching points, and perform maximum-minimum normalization as the second offset index;

[0024] Perform maximum-minimum normalization on the product of the first offset index and the second offset index as the comprehensive offset of all comparison corner points.

[0025] Further, screen out abnormal images from the oral cavity grayscale image according to the comprehensive offset, including:

[0026] Take the oral cavity grayscale image to which the comparison corner points with the comprehensive offset greater than the preset offset threshold belong as the abnormal image.

[0027] Further, determine the direction consistency of the comparison edges in each abnormal image according to the gradient directions of the comparison edges and the standard edges obtained by matching, including:

[0028] When at least two comparison corner points included in a certain comparison edge match the corresponding standard corner points in the standard edge, regard this comparison edge and the standard edge as the matching edges;

[0029] Determine the comparison gradient direction of the comparison edge and the standard gradient direction of the standard edge according to the gradient direction of each pixel point in the comparison edge and the standard edge;

[0030] Determine the angular value of the included angle between the comparison gradient direction of the comparison edge and the standard gradient direction of the matching standard edge as the edge matching angle difference;

[0031] Perform maximum-minimum normalization on the opposite number of the edge matching angle difference as the direction consistency.

[0032] Further, determine the comparison gradient direction of the comparison edge and the standard gradient direction of the standard edge according to the gradient direction of each pixel point in the comparison edge and the standard edge, including:

[0033] Determine the gradient direction of each pixel point in the comparison edge, and take the sum of the gradient directions of all pixel points as the comparison gradient direction of the comparison edge; similarly, take the sum of the gradient directions of all pixel points in the standard edge as the standard gradient direction of the standard edge.

[0034] Further, according to the direction consistency, determining the abnormal comparison edges of the abnormal image includes:

[0035] Regarding the comparison edges with the direction consistency less than the preset consistency threshold as the abnormal comparison edges.

[0036] Further, labeling and transmitting the abnormal comparison edges of all abnormal images includes:

[0037] Regarding the preset neighborhood area of the abnormal comparison edge in the abnormal image as the area to be analyzed;

[0038] Performing lossless image compression transmission on the area to be analyzed.

[0039] The present invention has the following beneficial effects:

[0040] In the embodiment of the present invention, by obtaining the reference gray-scale image of the historical normal teeth and the oral gray-scale image during morning check-up, it is convenient to construct the standard model and perform historical matching analysis subsequently, providing a data basis for the subsequent image analysis process. Through the standard model construction module, image feature analysis is performed on the tooth area in the reference gray-scale image to obtain the reference edges and reference corner points, and a standard model of each child's teeth is constructed through the reference edge curvature, the position distribution characteristics of the reference corner points, and the relationship of the reference corner points in the overall tooth shape, determining the standard corner points and standard edges; the construction of the standard model combines the reference gray-scale images of multiple historical normal teeth of the same child, enabling the standard model to accurately represent the adaptive tooth characteristics of each child, avoiding the influence caused by ignoring the tooth differences of different individuals when analyzing according to a unified tooth template, and improving the accuracy of subsequent standard analysis; then, the comprehensive offset is determined through corner point analysis, and the abnormal images are screened. The abnormal comparison edges are determined from the abnormal images through edge analysis. That is to say, image analysis of the abnormal tooth edges is performed using two dimensions of corner point matching and edge matching, enabling accurate analysis of the normal sulcus area and the abnormal tooth area. The abnormal comparison edges of all abnormal images are labeled and transmitted to improve the transmission efficiency. Labeling the abnormal comparison edges in the abnormal images improves the accuracy of subsequent abnormal recognition. In summary, the embodiment of the present invention constructs a standard model adaptive to different child individuals, combines the analysis of corner points and edges under the standard model, improves the accuracy of abnormal analysis, can more accurately analyze the abnormal tooth image area, improves the accuracy of image analysis of the abnormal tooth area, and improves the accuracy of auxiliary assessment of the dental health of child individuals. Description of the Drawings

[0041] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0042] Figure 1 The structural diagram of a children's morning oral health check image communication system provided by an embodiment of the present invention. Specific embodiments

[0043] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in conjunction with the drawings and preferred embodiments, details the specific embodiments, structures, features, and effects of a children's morning oral health check image communication system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0045] The following specifically describes the specific solution of a children's morning oral health check image communication system provided by the present invention in conjunction with the drawings.

[0046] Please refer to Figure 1 , which shows the structural diagram of a children's morning oral health check image communication system provided by an embodiment of the present invention, including: an acquisition module 101, a standard model construction module 102, a corner point analysis module 103, an edge analysis module 104, and a communication module 105.

[0047] The acquisition module 101 is used to acquire the reference grayscale image of the child's normal teeth in history and the oral grayscale image during the morning health check, and determine the tooth regions in the reference grayscale image and the oral grayscale image;

[0048] The standard model construction module 102 is used to perform image feature analysis on the tooth regions in the reference grayscale image and the oral grayscale image respectively, extract the reference edges and reference corner points in the reference grayscale image and the comparison edges and comparison corner points in the oral grayscale image at the same scale, and construct a standard model for each child's teeth based on the reference edge curvature, the position distribution characteristics of the reference corner points, and the relationship of the reference corner points in the overall tooth morphology. The standard model includes the standard corner points and standard edges of each tooth;

[0049] The corner point analysis module 103 is used to perform position matching between the comparison corner points and the standard corner points in the standard model to determine the comprehensive offset of all comparison corner points; and screen out abnormal images from the oral grayscale images according to the comprehensive offset;

[0050] The edge analysis module 104 is used to determine the direction consistency of the comparison edges in each abnormal image according to the gradient directions of the comparison edges and the matched standard edges in the abnormal images; and determine the abnormal comparison edges of the abnormal images according to the direction consistency;

[0051] The communication module 105 is used to label and transmit the abnormal comparison edges of all abnormal images.

[0052] The embodiment of the present invention aims to screen the oral images of children during morning physical examinations based on prior analysis, so as to determine abnormal tooth images, which is convenient for improving the accuracy of auxiliary assessment of children's individual tooth health. Among them, tooth abnormalities mainly include tooth dislocation and loss, dental calculus, and tooth decay. They all affect the shape, color, and structure of teeth. Therefore, specific abnormal analysis is mainly carried out through tooth color and shape.

[0053] It should be noted that in the embodiment of the present invention, the acquisition module 101 mainly realizes the data acquisition function and provides a data basis for the analysis of children's oral images during morning physical examinations. In the embodiment of the present invention, a reference grayscale image of normal teeth in the past is acquired, and this reference grayscale image is used to construct an adaptive tooth template for different children. The oral grayscale image during morning physical examination is an oral image obtained by shooting during morning physical examination, and this oral image is grayscale processed to obtain the oral grayscale image.

[0054] Since the oral cavity contains tooth regions, gum regions, and other corresponding regions, such as the tongue region, etc., in order to accurately analyze teeth, it is necessary to extract the tooth regions. In the embodiment of the present invention, semantic segmentation is performed on the reference grayscale image and the oral grayscale image respectively to determine the tooth regions.

[0055] Among them, semantic segmentation is a well-known technology in the art. Through semantic segmentation, the tooth regions can be accurately divided. Of course, in some other embodiments of the present invention, other methods can also be used to determine the tooth regions, such as gray value segmentation. The tooth regions generally appear white or light yellow, which has an obvious difference in gray scale from the red color of other regions. Therefore, the tooth regions can be determined by gray scale division.

[0056] In the embodiment of the present invention, since the tooth shapes of different children vary greatly, it is impossible to analyze all children using a unified tooth template. It is necessary to set an adaptive template for different children respectively, that is, it is necessary to construct a standard model of each child's teeth.

[0057] It should be noted that the morning check-up frequency for children varies in different situations, ranging from once a month to once a year. For effective overall analysis and to avoid misidentification, scale normalization is required to ensure that the collected images are on the same scale.

[0058] In the embodiments of the present invention, the reference grayscale image and the oral grayscale image are normalized based on a geometric calibration method; the geometric calibration method is a method well-known to those skilled in the relevant art. By using the geometric calibration method to normalize the images, the tooth regions in different images are aligned to a unified scale, and the tooth information in the tooth regions of different images is analyzed and trained.

[0059] In the embodiments of the present invention, Canny edge detection is performed on the reference grayscale image to obtain reference edges, and the reference edges form a reference edge image. Then, corner detection is performed to obtain reference corner points; Canny edge detection is performed on the oral grayscale image to obtain comparison edges, and the comparison edges form a comparison edge image, and corner detection is performed on the comparison edge image to obtain comparison corner points.

[0060] Edge detection and corner detection can detect and analyze each image, thereby extracting edge features and corner features. Since they belong to the same child individual, that is, the tooth region of the oral grayscale image obtained from the morning check-up is similar to the tooth region of the reference grayscale image, and the inconsistent region is the abnormal region.

[0061] Moreover, since the number of reference grayscale images for the same child is multiple, there will also be certain differences in the corresponding tooth regions for each of them. Therefore, in the embodiments of the present invention, the tooth regions of all reference grayscale images for the same child are model-integrated to obtain the standard model of each child's teeth.

[0062] Furthermore, in some embodiments of the present invention, based on the curvature of the reference edge, the position distribution characteristics of the reference corner points, and the relationship of the reference corner points in the overall tooth morphology, a standard model of each child's teeth is constructed, including: determining the curvature of the reference edge of the teeth of the same child in all reference grayscale images, the position distribution of the reference corner points on the reference edge, and the distribution law of the reference corner points on the tooth contour as the construction information; inputting the construction information into a preset tooth image model and performing model training on the preset tooth image model to obtain a standard model, where the standard model is a standardized image model corresponding to the child's tooth region.

[0063] Among them, the construction information is the position information of the tooth edge, corner points, etc. In the embodiments of the present invention, the construction information mainly includes the curvature of the reference edge in the reference grayscale image, the position distribution of the reference corner points on the reference edge, and the distribution law of the reference corner points on the tooth contour. It should be noted that this part of the construction information can be used as big data information and set as training data, which is input into the big data model to be trained for model training to obtain a standard model. Among them, the big data model can be specifically, for example, a machine learning model, which performs machine learning through the tooth region features under normal conditions, that is, the reference edge and the reference corner points, so as to realize the training process.

[0064] In the embodiments of the present invention, after training to obtain a standard model, it is necessary to perform standard difference analysis on each oral grayscale image.

[0065] Further, in some embodiments of the present invention, the position matching is performed between the comparison corner points and the standard corner points in the standard model to determine the comprehensive offset of all comparison corner points, including: mapping the position of each comparison corner point in the oral grayscale image to the standard model at the same scale, and performing model matching according to the position of the comparison corner point and the position of the standard corner point, and matching the comparison corner point with the closest position to the standard corner point as a pair of matching points; taking the mean value of the Euclidean distances of all pairs of matching points as the first offset index; calculating the quantity difference between all comparison corner points and the comparison corner points among the matching points, and performing maximum-minimum normalization processing as the second offset index; multiplying the first offset index by the second offset index and performing maximum-minimum normalization processing as the comprehensive offset of all comparison corner points.

[0066] It can be understood that abnormal phenomena such as tooth dislocation and shedding, dental calculus, and tooth decay will all cause changes in the shape of the teeth, resulting in changes in the position, quantity, and curvature of the corner points in the tooth region. For example, the dental calculus region will cause the teeth to be covered by dental calculus, and the covered region will cover the normal tooth edge, resulting in a large number of edges and corner points in the image. The position, quantity, curvature, and length of the corner points corresponding to these edges and corner points in the standard tooth model have all changed. Therefore, by comparing the differences in the edge curvature, length, and the position and quantity of the corner points in the tooth region with the standard model, if the change characteristics of the edge curvature or the length distribution deviate from the standard model, it can be determined whether this region is a potential abnormal region.

[0067] Among them, in the overall matching process, this solution performs abnormal analysis through two matching analysis processes: corner point matching and edge matching. First, the comprehensive offset is determined through corner point matching. The comprehensive offset in the embodiments of the present invention is the offset index corresponding to corner point matching, and the comprehensive offset analysis is performed through the distance between the already matched corner points and the proportion of the number of matched corner points.

[0068] In the embodiments of the present invention, the first offset index is the index information corresponding to the corner distance. The larger the Euclidean distance between the matching points, the farther the position of the compared corner point that is matched from the position of the standard corner point, that is, the more likely it is that the corner point offset is caused by some abnormal reasons. At this time, the value of the first offset index is larger.

[0069] In the embodiments of the present invention, the difference in the number of compared corner points among all compared corner points and matching points represents the number of compared corner points that are not matched. The larger this value, the larger the number of compared corner points that are not matched, that is, the more likely it is that abnormal phenomena occur as a whole, and the larger the value of the second offset index.

[0070] Therefore, by combining the first offset index and the second offset index, a comprehensive offset amount is calculated. Among them, since the larger the values of the first offset index and the second offset index, the more abnormal they can represent, the product value of the first offset index and the second offset index is directly subjected to maximum-minimum normalization processing and used as the comprehensive offset amount of all compared corner points.

[0071] Of course, in some other embodiments of the present invention, the sum value of the first offset index and the second offset index can also be directly calculated and normalized as the comprehensive offset amount.

[0072] It can be understood that the larger the value of the comprehensive offset amount, the worse the matching effect between the compared corner points in the oral gray-scale image and the standard model, that is, the more likely it is that abnormal phenomena occur correspondingly, such as abnormal phenomena such as tooth dislocation and shedding, dental calculus, and tooth decay. Therefore, further screening can be performed in combination with the comprehensive offset amount.

[0073] Further, in some embodiments of the present invention, abnormal images are screened from the oral gray-scale images according to the comprehensive offset amount, including: using the oral gray-scale image to which the compared corner points with a comprehensive offset amount greater than the preset offset threshold belong as the abnormal image.

[0074] Among them, the preset offset threshold is the threshold value of the comprehensive offset amount. In the embodiments of the present invention, the preset offset threshold can be set to 0.5, that is to say, when the comprehensive offset amount is greater than 0.5, the corresponding oral gray-scale image is screened as the abnormal image.

[0075] The abnormal images in the embodiments of the present invention indicate that there are corresponding abnormal effects in the images. However, since only corner point detection is used, the accuracy of the obtained abnormal analysis effect is not ideal. Therefore, the analysis of the edges needs to be added to effectively cope with complex tooth changes.

[0076] Among them, for edge analysis, the gradient direction is mainly analyzed. Since the gums themselves change with tooth use, especially in children, the changes in the normal growth of the gum edges and tooth edges in the oral cavity are relatively fast. Directly analyzing based on the position deviation of the edges has low accuracy and insufficient reliability. Also, under normal growth conditions in the tooth area of the oral cavity, the overall edge shape changes little. And due to phenomena such as tooth decay and dental calculus, they have a greater impact on the junction edge between the teeth and the gums, resulting in obvious deformation of the edges. That is, each edge has a similar gradient direction under normal circumstances and a large change in the gradient direction under abnormal circumstances. Therefore, in the embodiments of the present invention, abnormal analysis of the tooth edges is performed through the gradient direction.

[0077] Further, in some embodiments of the present invention, according to the gradient directions of the comparison edges and the standard edges obtained by matching in the abnormal images, the direction consistency of the comparison edges in each abnormal image is determined, including: when at least two comparison corner points included in a certain comparison edge match the corresponding standard corner points in the standard edge, regarding this comparison edge and the standard edge as the matching edges; determining the comparison gradient direction of the comparison edge and the standard gradient direction of the standard edge according to the gradient directions of each pixel point in the comparison edge and the standard edge; determining the angular value of the included angle between the comparison gradient direction of the comparison edge and the standard gradient direction of the matching standard edge as the edge matching angle difference; performing maximum-minimum normalization processing on the opposite number of the edge matching angle difference as the direction consistency.

[0078] Among them, edge matching can be analyzed by combining the characteristics of corner point matching. That is, when at least two comparison corner points included in a certain comparison edge match the corresponding standard corner points in the standard edge, regarding this comparison edge and the standard edge as the matching edges. Since tooth decay, dental calculus, etc. can cause edge changes, directly performing matching has poor effects. Through the method of corner point matching, accurate matching analysis can be carried out, and its matching accuracy is higher.

[0079] Then, gradient direction analysis is performed. It can be understood that under normal growth conditions, the change in its gradient direction is small, while under abnormal conditions, the change in its gradient direction is large. The standard gradient direction represents the overall gradient direction of the standard edge, and the comparison gradient direction represents the overall gradient direction of the comparison edge. Therefore, in the embodiments of the present invention, it is necessary to perform difference analysis on the matching standard gradient direction and comparison gradient direction.

[0080] Further, in some embodiments of the present invention, according to the gradient directions of each pixel point in the comparison edge and the standard edge, determining the comparison gradient direction of the comparison edge and the standard gradient direction of the standard edge includes: determining the gradient direction of each pixel point in the comparison edge, and taking the sum of the gradient directions of all pixel points as the comparison gradient direction of the comparison edge; similarly, taking the sum of the gradient directions of all pixel points in the standard edge as the standard gradient direction of the standard edge.

[0081] Since there may be a large difference in the number of pixel points between the comparison edge and the standard edge, directly matching pixel points one by one has a poor effect. Therefore, the present invention integrates the sum vector of the gradient directions of all pixel points in the standard edge (by default, the vector length of each pixel point is recorded as 1), takes the direction of the corresponding sum vector as the standard gradient direction, integrates the sum vector of the gradient directions of all pixel points in the comparison edge, and takes the direction of the corresponding sum vector as the comparison gradient direction.

[0082] In the embodiments of the present invention, the analysis of the difference between the two directions is specifically carried out through the angular value of the angle between the two directions. Among them, the included angle is specifically an angle less than or equal to 180 degrees. That is to say, the larger the included angle between the two directions, the greater the difference between the two directions. Therefore, in the embodiments of the present invention, the edge matching angle difference is directly determined, and the opposite number of the edge matching angle difference is normalized by the maximum and minimum values as the direction consistency.

[0083] Among them, the direction consistency is a characteristic index representing direction consistency. That is, the larger the value of the direction consistency, the more consistent the overall gradient directions between the matching standard edge and the comparison edge.

[0084] Since the smaller the value of the direction consistency, the greater the difference in the overall gradient directions between the comparison edge and the matching standard edge, that is, the comparison edge has a direction change. This direction change may be caused by the holes affected by dental caries or the edge change caused by dental calculus. Therefore, the smaller the value of the direction consistency, the greater the abnormal feature.

[0085] In the embodiments of the present invention, the comparison edge with the direction consistency less than the preset consistency threshold is used as the abnormal comparison edge.

[0086] Among them, the preset consistency threshold is the threshold value of the direction consistency. The preset consistency threshold in the embodiments of the present invention can be specifically, for example, 0.5, that is, the comparison edge with the direction consistency less than 0.5 is used as the abnormal comparison edge.

[0087] Then this abnormal comparison edge indicates that in the abnormal image with poor corner matching effect, the worse the matching effect of the comparison edge, that is, the abnormal image during the morning check needs to be further processed, such as subsequent manual analysis or classified storage, etc.

[0088] In the embodiments of the present invention, the abnormal comparison edges of all abnormal images are marked and transmitted, including: taking the preset neighborhood area of the abnormal comparison edge in the abnormal image as the area to be analyzed; performing lossless image compression transmission on the area to be analyzed.

[0089] Among them, for the preset neighborhood area, morphological dilation can be specifically performed on the abnormal comparison edge, and the dilation scale can be 10, that is, the area within 10 pixel points on both sides of the abnormal comparison edge is taken as the area to be analyzed, and lossless image compression transmission is performed on the area to be analyzed.

[0090] In the embodiments of the present invention, the image areas other than the area to be analyzed may not be transmitted, or lossy compression transmission may be performed to improve the transmission efficiency. And if there are subsequent manual analysis steps, it is convenient to focus on analyzing the abnormal area to be analyzed, improve the subsequent processing efficiency, and overall improve the accuracy of the auxiliary assessment of the dental health of children individuals.

[0091] In the embodiments of the present invention, by obtaining the reference gray-scale images of historical normal teeth and the oral gray-scale images during morning check-ups, it is convenient to perform standard model construction and historical matching analysis subsequently, providing a data basis for the subsequent image analysis process. The image feature analysis of the tooth area in the reference gray-scale image is performed through the standard model construction module, so as to obtain the reference edges and reference corner points, and a standard model of each child's teeth is constructed through the reference edge curvature, the position distribution characteristics of the reference corner points, and the relationship between the reference corner points in the overall tooth shape, and the standard corner points and standard edges are determined; the construction of the standard model combines the reference gray-scale images of multiple historical normal teeth of the same child, so that the standard model can accurately represent the adaptive tooth characteristics of each child, avoiding the influence caused by ignoring the tooth differences of different individuals when analyzing according to a unified tooth template, and improving the accuracy of subsequent standard analysis; then, the comprehensive offset is determined through corner point analysis, and abnormal images are screened, and the abnormal comparison edges are determined from the abnormal images through edge analysis, that is to say, image analysis of the abnormal tooth edges is performed using two dimensions of corner point matching and edge matching, so that the normal sulcus area and the abnormal tooth area can be accurately analyzed, the abnormal comparison edges of all abnormal images are marked and transmitted to improve the transmission efficiency, and the abnormal comparison edges in the abnormal images are marked, improving the accuracy of subsequent abnormality recognition. In summary, the embodiments of the present invention construct standard models that are adaptive to different child individuals, combine the analysis of corner points and edges under the standard model, improve the accuracy of abnormal analysis, so as to more accurately analyze the abnormal tooth image area, improve the accuracy of the image analysis of the abnormal tooth area, and improve the accuracy of the auxiliary assessment of the dental health of children individuals.

[0092] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A children's morning check-up oral image communication system, characterized in that: The system includes: An acquisition module, used to acquire a reference grayscale image of the child's historical normal teeth and an oral grayscale image during a morning checkup, and determine a tooth region in the reference grayscale image and the oral grayscale image; A standard model construction module is used to perform image feature analysis on the tooth areas in the reference grayscale image and the oral grayscale image, extract reference edges and reference corners in the reference grayscale image and contrast edges and contrast corners in the oral grayscale image at the same scale, and construct a standard model of each child's tooth based on the reference edge curvature, the position distribution characteristics of the reference corners, and the relationship between the reference corners and the overall morphology of the teeth, wherein the standard model includes the standard corners and standard edges of each tooth; A corner point analysis module is used to match the positions of the comparison corner points with the standard corner points in the standard model, determine the comprehensive offset of all comparison corner points; and screen the abnormal image from the oral grayscale image according to the comprehensive offset; An edge analysis module, used to determine the direction consistency of the contrast edge in each abnormal image according to the gradient direction of the contrast edge in the abnormal image and the matched standard edge; and determine the abnormal contrast edge of the abnormal image according to the direction consistency; The communication module is used to mark and transmit the abnormal contrast edges of all abnormal images.

2. A children's morning check-up oral image communication system as claimed in claim 1, characterized in that: The step of determining the tooth region in the reference grayscale image and the oral grayscale image comprises: Semantic segmentation is performed on the reference grayscale image and the oral grayscale image respectively to determine the tooth area.

3. The oral image communication system for children's morning checkup according to claim 1, characterized in that: Performing image feature analysis on the tooth regions in the reference grayscale image and the oral grayscale image respectively, extracting reference edges and reference corners in the reference grayscale image and contrast edges and contrast corners in the oral grayscale image at the same scale, including: The reference grayscale image and the oral grayscale image are normalized based on a geometric calibration method so that the tooth regions in the reference grayscale image and the oral grayscale image are aligned to a uniform scale; Performing Canny edge detection on the reference grayscale image to obtain reference edges, the reference edges constitute a reference edge image, and then performing corner point detection to obtain reference corner points; Performing Canny edge detection on the oral grayscale image to obtain contrast edges, the contrast edges constitute a contrast edge image, and performing corner point detection on the contrast edge image to obtain contrast corner points.

4. The oral image communication system for children's morning checkup according to claim 1, characterized in that: Based on the reference edge curvature, the position distribution characteristics of the reference corner points, and the relationship between the reference corner points and the overall shape of the teeth, a standard model of each child's teeth is constructed, including: Determine the curvature of the reference edge of the same child's tooth in all reference grayscale images, the position distribution of the reference corner points on the reference edge, and the distribution law of the reference corner points on the tooth contour as construction information; The construction information is input into a preset tooth image model, and model training is performed on the preset tooth image model to obtain a standard model, wherein the standard model is a standardized image model corresponding to a child's tooth area.

5. The oral image communication system for children's morning checkup according to claim 1, characterized in that: The step of matching the positions of the comparison corner points with the standard corner points in the standard model to determine the comprehensive offsets of all the comparison corner points includes: The position of each comparison corner point in the oral grayscale image is mapped to the standard model at the same scale, and the model is matched according to the position of the comparison corner point and the position of the standard corner point, and the comparison corner point and the standard corner point with the closest position are matched as a pair of matching points; The mean of the Euclidean distances of all pairs of matching points is used as the first offset indicator; Calculate the difference in the number of all contrast corner points and the contrast corner points in the matching points, and normalize the maximum and minimum values ​​as the second offset index; The product of the first offset index and the second offset index is normalized to the maximum and minimum values ​​to be used as the comprehensive offset of all the compared corner points.

6. The oral image communication system for children's morning checkup according to claim 1, characterized in that: Screening an abnormal image from the oral grayscale image according to the comprehensive offset includes: The oral grayscale image to which the comparison corner point having the comprehensive offset greater than the preset offset threshold belongs is regarded as an abnormal image.

7. A children's morning checkup oral image communication system as claimed in claim 5, characterized in that: According to the gradient direction of the contrast edge in the abnormal image and the matched standard edge, the direction consistency of the contrast edge in each abnormal image is determined, including: When at least two comparison corner points included in a comparison edge match corresponding standard corner points in the standard edge, the comparison edge and the standard edge are regarded as matched edges; Determine the contrast gradient direction of the contrast edge and the standard gradient direction of the standard edge according to the gradient direction of each pixel in the contrast edge and the standard edge; Determine the angle value of the contrast gradient direction of the contrast edge and the standard gradient direction of the matched standard edge as the edge matching angle difference; The opposite number of the edge matching angle difference is normalized to the maximum and minimum values ​​to be used as the direction consistency.

8. The oral image communication system for children's morning checkup as claimed in claim 7, characterized in that: According to the gradient direction of each pixel in the contrast edge and the standard edge, the contrast gradient direction of the contrast edge and the standard gradient direction of the standard edge are determined, including: Determine the gradient direction of each pixel in the contrast edge, and take the sum of the gradient directions of all pixels as the contrast gradient direction of the contrast edge; similarly, take the sum of the gradient directions of all pixels in the standard edge as the standard gradient direction of the standard edge.

9. The oral image communication system for children's morning checkup according to claim 1, characterized in that: Determining an abnormal contrast edge of the abnormal image according to the directional consistency includes: The contrast edge whose direction consistency is less than a preset consistency threshold is regarded as an abnormal contrast edge.

10. The oral image communication system for children's morning checkup according to claim 1, characterized in that: The abnormal contrast edges of all abnormal images are marked and transmitted, including: Taking a preset neighborhood area of ​​the abnormal contrast edge in the abnormal image as an area to be analyzed; The region to be analyzed is subjected to lossless image compression and transmission.

Images