Titanium network intelligent positioning method, device, equipment and storage medium

Abstract

The present invention relates to the field of image processing technology, and in particular to a titanium mesh intelligent positioning method, device, equipment and storage medium. The method comprises the following steps: obtaining a multi-directional image of the oral cavity; extracting oral cavity optical flow feature points based on the multi-directional image of the oral cavity; performing super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image; performing feature point spatial matching on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points, thereby obtaining feature point spatial position data; performing point cloud contour standardization processing on the super-resolution oral cavity image to obtain a tooth socket point cloud contour map; performing three-dimensional structural reconstruction of the tooth socket point cloud contour map through the feature point spatial position data to construct a three-dimensional point cloud model of the oral tooth socket; and obtaining an oral titanium mesh image based on a preset database. The present invention realizes accurate titanium mesh positioning.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a titanium mesh intelligent positioning method, device, equipment and storage medium. Background Art

[0002] With the development of modern medical technology, dental restoration technology is becoming more and more refined and personalized. As a widely used dental restoration material, titanium mesh brackets play an important role in repairing tooth defects and enhancing tooth mechanical properties. However, the traditional titanium mesh bracket installation method relies on the doctor's manual measurement and experience operation, which has the problems of low accuracy and poor repeatability, and cannot meet the increasingly refined diagnosis and treatment needs. Therefore, an intelligent titanium mesh positioning method is needed. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a titanium mesh intelligent positioning method, device, equipment and storage medium to solve at least one of the above technical problems.

[0004] To achieve the above object, the present invention provides a titanium mesh intelligent positioning method, comprising the following steps:

[0005] Step S1: acquiring a multi-directional image of the oral cavity; extracting oral cavity optical flow feature points based on the multi-directional image of the oral cavity; and performing super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image;

[0006] Step S2: performing spatial matching of feature points of the oral multi-directional image based on the oral optical flow feature points to obtain spatial position data of the feature points; performing point cloud contour normalization processing on the super-resolution oral image to obtain a tooth socket point cloud contour map; reconstructing the three-dimensional structure of the tooth socket point cloud contour map through the feature point spatial position data to construct a three-dimensional point cloud model of the oral cavity;

[0007] Step S3: acquiring an oral titanium mesh image based on a preset database; visually positioning the oral titanium mesh image and marking the titanium mesh positioning points; performing multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning points to generate multiple titanium mesh positioning registration samples;

[0008] Step S4: performing point-by-point positioning probability prediction on multiple titanium mesh positioning and registration samples to generate positioning point probability prediction values; performing confidence mapping based on the positioning point probability prediction values ​​to construct a titanium mesh positioning confidence map;

[0009] Step S5: segment the surrounding area of ​​the titanium mesh positioning confidence map to obtain a multi-scale confidence image; obtain an image sub-pixel offset value based on the multi-scale confidence image; perform sub-pixel positioning fine-tuning based on the image sub-pixel offset value to obtain a sub-pixel fine-embedded positioning point;

[0010] Step S6: Generate personalized bone structure feature data based on the oral alveolar three-dimensional point cloud model; reposition the titanium mesh error on the personalized bone structure feature data based on sub-pixel fine embedding positioning points, so as to obtain the optimal positioning result of the titanium mesh.

[0011] In this specification, a titanium mesh intelligent positioning device is also provided, including:

[0012] The behavior sequence module is used to obtain the real-time payment information of the enterprise; generate multi-dimensional payment information based on the real-time payment information of the enterprise; perform time series fitting on the multi-dimensional payment information to generate a payment behavior sequence;

[0013] The risk behavior feature module is used to obtain the local dependency of the payment sequence based on the payment behavior sequence; the payment behavior sequence is subjected to implicit risk feature analysis using the local dependency of the payment sequence to obtain implicit payment risk behavior feature data;

[0014] A topological network module is used to obtain multiple payment behavior nodes based on the payment behavior sequence; use implicit payment risk behavior feature data to dynamically connect multiple payment behavior nodes to form a distributed behavior node network;

[0015] The risk dynamic evolution module is used to perform risk probability distribution analysis on the distributed behavior node network to obtain node risk probability distribution data; perform risk propagation dynamic evolution analysis based on the node risk probability distribution data to generate risk propagation evolution data;

[0016] The risk trend prediction module is used to quantitatively predict risk trends one by one based on risk propagation evolution data to obtain the risk trend prediction value of the payment node; mark high-risk payment nodes based on the risk trend prediction value of the payment node; reconstruct the global potential association path of the high-risk payment node to obtain the payment risk deduction chain;

[0017] The risk monitoring module is used to conduct risk monitoring decision mining on the distributed behavior node network based on the payment risk deduction chain, build an intelligent payment risk monitoring engine, and execute enterprise payment operations based on the intelligent payment risk monitoring engine.

[0018] The present invention also provides a computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, and wherein the processor implements the steps of any one of the above-mentioned titanium mesh intelligent positioning methods when executing the computer program.

[0019] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any of the above-mentioned titanium mesh intelligent positioning methods are implemented.

[0020] The titanium mesh intelligent and precise positioning method, device, equipment and storage medium provided by the present invention have the following beneficial effects:

[0021] The present invention provides detailed surface information of the inside of the oral cavity by acquiring multi-directional images of the oral cavity, and provides basic data for subsequent titanium mesh positioning. The extraction of oral optical flow feature points can capture the movement information in the oral cavity image, and help determine the position of the teeth and alveolar bones. The super-resolution fine structure restoration processing improves the resolution of the oral cavity image by enhancing the details and clarity of the image, and helps to accurately locate the titanium mesh. The spatial matching of the oral optical flow feature points provides an accurate correspondence between the teeth and the alveolar bones, and provides a spatial reference for the subsequent titanium mesh positioning. The point cloud contour standardization processing can extract the shape information of the alveolar and convert it into standardized point cloud data, which provides a basis for the subsequent three-dimensional structure reconstruction. The construction of the oral alveolar three-dimensional point cloud model provides an accurate representation of the three-dimensional structure inside the oral cavity, and provides accurate spatial information for titanium mesh positioning. The preset database provides a large amount of oral titanium mesh image data, and provides rich samples for the training and optimization of titanium mesh positioning. The visual positioning frame selection and the marking of the titanium mesh positioning points can accurately locate the position of the titanium mesh in the oral cavity image, and provide a reference for the subsequent positioning accuracy. The multi-positioning point registration iterative simulation can simulate a variety of titanium mesh positioning situations , generate multiple titanium mesh positioning and registration samples, providing more possibilities for the selection of the optimal positioning result. The positioning point probability prediction can predict the positioning accuracy of each positioning point by analyzing multiple titanium mesh positioning and registration samples, and provide an evaluation for the reliability of the positioning result. The confidence mapping constructs the titanium mesh positioning confidence map by mapping the positioning point probability prediction value to space, which can intuitively represent the positioning credibility of the titanium mesh and provide a basis for the subsequent positioning result analysis and selection. The peripheral area segmentation of the titanium mesh positioning confidence map can segment the image into areas with different confidences, providing a multi-scale The confidence information of the degree is helpful to locate the titanium mesh more accurately. The calculation of the sub-pixel offset value of the image can provide the sub-pixel accuracy of the positioning point. By analyzing the sub-pixel offset value, the positioning result can be fine-tuned to further improve the positioning precision and accuracy. The personalized bone structure feature data of the oral alveolar 3D point cloud model can provide the feature information required for titanium mesh positioning and provide a basis for the generation of the optimal positioning result. The titanium mesh error repositioning combines the sub-pixel fine-embedded positioning point and the personalized bone structure feature data to realize the correction and adjustment of the titanium mesh positioning error and obtain the optimal positioning result of the titanium mesh. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of the steps of a titanium mesh intelligent positioning method of the present invention;

[0023] Figure 2 Detailed implementation flow chart of step S1;

[0024] Figure 3 Detailed implementation flow chart of step S2;

[0025] Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0027] This application example provides a titanium network intelligent positioning method, system, device and storage medium. The execution subjects of the titanium network intelligent positioning method, system, device and storage medium include but are not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the device can be regarded as the general computing node of this application, and the data processing platform includes but is not limited to: at least one of an audio and image management device, an information management device, and a cloud data management device.

[0028] See also Figures 1 to 4 The present invention provides a titanium mesh intelligent positioning method, the method comprising the following steps:

[0029] Step S1: acquiring a multi-directional image of the oral cavity; extracting oral cavity optical flow feature points based on the multi-directional image of the oral cavity; and performing super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image;

[0030] Step S2: performing spatial matching of feature points of the oral multi-directional image based on the oral optical flow feature points to obtain spatial position data of the feature points; performing point cloud contour normalization processing on the super-resolution oral image to obtain a tooth socket point cloud contour map; reconstructing the three-dimensional structure of the tooth socket point cloud contour map through the feature point spatial position data to construct a three-dimensional point cloud model of the oral cavity;

[0031] Step S3: acquiring an oral titanium mesh image based on a preset database; visually positioning the oral titanium mesh image and marking the titanium mesh positioning points; performing multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning points to generate multiple titanium mesh positioning registration samples;

[0032] Step S4: performing point-by-point positioning probability prediction on multiple titanium mesh positioning and registration samples to generate positioning point probability prediction values; performing confidence mapping based on the positioning point probability prediction values ​​to construct a titanium mesh positioning confidence map;

[0033] Step S5: segment the surrounding area of ​​the titanium mesh positioning confidence map to obtain a multi-scale confidence image; obtain an image sub-pixel offset value based on the multi-scale confidence image; perform sub-pixel positioning fine-tuning based on the image sub-pixel offset value to obtain a sub-pixel fine-embedded positioning point;

[0034] Step S6: Generate personalized bone structure feature data based on the oral alveolar three-dimensional point cloud model; reposition the titanium mesh error on the personalized bone structure feature data based on sub-pixel fine embedding positioning points, so as to obtain the optimal positioning result of the titanium mesh.

[0035] The present invention provides detailed surface information of the inside of the oral cavity by acquiring multi-directional images of the oral cavity, and provides basic data for subsequent titanium mesh positioning. The extraction of oral optical flow feature points can capture the movement information in the oral cavity image, and help determine the position of the teeth and alveolar bones. The super-resolution fine structure restoration processing improves the resolution of the oral cavity image by enhancing the details and clarity of the image, and helps to accurately locate the titanium mesh. The spatial matching of the oral optical flow feature points provides an accurate correspondence between the teeth and the alveolar bones, and provides a spatial reference for the subsequent titanium mesh positioning. The point cloud contour standardization processing can extract the shape information of the alveolar and convert it into standardized point cloud data, which provides a basis for the subsequent three-dimensional structure reconstruction. The construction of the oral alveolar three-dimensional point cloud model provides an accurate representation of the three-dimensional structure inside the oral cavity, and provides accurate spatial information for titanium mesh positioning. The preset database provides a large amount of oral titanium mesh image data, and provides rich samples for the training and optimization of titanium mesh positioning. The visual positioning frame selection and the marking of the titanium mesh positioning points can accurately locate the position of the titanium mesh in the oral cavity image, and provide a reference for the subsequent positioning accuracy. The multi-positioning point registration iterative simulation can simulate a variety of titanium mesh positioning situations , generate multiple titanium mesh positioning and registration samples, providing more possibilities for the selection of the optimal positioning result. The positioning point probability prediction can predict the positioning accuracy of each positioning point by analyzing multiple titanium mesh positioning and registration samples, and provide an evaluation for the reliability of the positioning result. The confidence mapping constructs the titanium mesh positioning confidence map by mapping the positioning point probability prediction value to space, which can intuitively represent the positioning credibility of the titanium mesh and provide a basis for the subsequent positioning result analysis and selection. The peripheral area segmentation of the titanium mesh positioning confidence map can segment the image into areas with different confidences, providing a multi-scale The confidence information of the degree is helpful to locate the titanium mesh more accurately. The calculation of the sub-pixel offset value of the image can provide the sub-pixel accuracy of the positioning point. By analyzing the sub-pixel offset value, the positioning result can be fine-tuned to further improve the positioning precision and accuracy. The personalized bone structure feature data of the oral alveolar 3D point cloud model can provide the feature information required for titanium mesh positioning and provide a basis for the generation of the optimal positioning result. The titanium mesh error repositioning combines the sub-pixel fine-embedded positioning point and the personalized bone structure feature data to realize the correction and adjustment of the titanium mesh positioning error and obtain the optimal positioning result of the titanium mesh.

[0036] In the embodiment of the present invention, refer to Figure 1 , is a schematic diagram of the steps of a titanium mesh intelligent positioning method of the present invention. In this example, the steps of the titanium mesh intelligent positioning method include:

[0037] Step S1: acquiring a multi-directional image of the oral cavity; extracting oral cavity optical flow feature points based on the multi-directional image of the oral cavity; and performing super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image;

[0038] In this embodiment, a 3D scanning device (such as CBCT, CT, etc.) is used to image the patient's oral area from different angles to obtain multi-perspective 3D image data. Through this multi-perspective imaging, the three-dimensional information of the internal structure of the oral cavity can be fully collected. The optical flow algorithm in computer vision technology is used to extract stable feature points from the multi-perspective image of the oral cavity. The optical flow feature points can describe the subtle changes and dynamic information of the internal structure of the oral cavity. The correspondence between the feature points is used to restore and enhance the subtle structure of the original low-resolution oral image through a super-resolution reconstruction algorithm. In this way, a series of high-resolution and detail-rich oral 3D images can be obtained, providing more detailed image data for subsequent precise positioning.

[0039] Step S2: performing spatial matching of feature points of the oral multi-directional image based on the oral optical flow feature points to obtain spatial position data of the feature points; performing point cloud contour normalization processing on the super-resolution oral image to obtain a tooth socket point cloud contour map; reconstructing the three-dimensional structure of the tooth socket point cloud contour map through the feature point spatial position data to construct a three-dimensional point cloud model of the oral cavity;

[0040] In this embodiment, feature matching algorithms, such as SIFT and SURF, are used to perform spatial correspondence matching on feature points on multi-directional images of the oral cavity collected from different angles. The coordinate position data of each feature point in three-dimensional space can be obtained through the correspondence between the feature points. The high-resolution oral 3D image data is converted into a point cloud format, and contour extraction, smoothing, resampling and other processing are performed on the point cloud data to obtain a standardized point cloud contour map of the alveolar area. The three-dimensional coordinate data of the feature points obtained above are combined with the standardized point cloud contour map of the alveolar area to reconstruct the three-dimensional model. Through technical means such as point cloud registration and mesh reconstruction, a fine three-dimensional point cloud model of the oral alveolar area is constructed.

[0041] Step S3: acquiring an oral titanium mesh image based on a preset database; visually positioning the oral titanium mesh image and marking the titanium mesh positioning points; performing multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning points to generate multiple titanium mesh positioning registration samples;

[0042] In this embodiment, based on the constructed oral titanium mesh image database, a large number of real oral titanium mesh image samples are obtained. These image samples should cover different patients, different angles, and different types of titanium mesh implantation scenarios. For each titanium mesh image in the database, manual target selection and titanium mesh positioning point annotation are performed. During the annotation process, attention should be paid to the positions of key feature points such as the titanium mesh edge and screw holes. A titanium mesh positioning point dataset containing a large number of annotated samples is established. The aforementioned constructed oral alveolar three-dimensional point cloud model and the annotated titanium mesh positioning point dataset are used. A registration algorithm (such as ICP) is used to iteratively optimize and match these two types of data to simulate different titanium mesh installation scenarios. Through multiple iterative simulations, a large number of registration samples representing different titanium mesh installation positions are generated.

[0043] Step S4: performing point-by-point positioning probability prediction on multiple titanium mesh positioning and registration samples to generate positioning point probability prediction values; performing confidence mapping based on the positioning point probability prediction values ​​to construct a titanium mesh positioning confidence map;

[0044] In this embodiment, a large number of titanium mesh positioning and registration samples generated above are used as training data to train a machine learning model, which can be a target detection network based on deep learning. The input is the oral alveolar three-dimensional point cloud model, and the output is the titanium mesh positioning probability of each point. Through the reasoning of the model, the titanium mesh positioning probability of the new oral alveolar three-dimensional point cloud data is predicted point by point, and the positioning probability prediction value of each point output by the model is mapped to a confidence value between 0 and 1. These confidence values ​​are visualized as a pseudo-color titanium mesh positioning confidence map, which reflects the positioning confidence of each area. In this map, the warmer the color, the higher the positioning confidence, and the colder the color, the lower the positioning confidence.

[0045] Step S5: segment the surrounding area of ​​the titanium mesh positioning confidence map to obtain a multi-scale confidence image; obtain an image sub-pixel offset value based on the multi-scale confidence image; perform sub-pixel positioning fine-tuning based on the image sub-pixel offset value to obtain a sub-pixel fine-embedded positioning point;

[0046] In this embodiment, starting from the original titanium mesh positioning confidence map, an image segmentation algorithm (such as K-Means clustering) is used to segment the map. The goal of segmentation is to separate high-confidence areas from low-confidence areas, and obtain multiple confidence areas of different scales. These areas can reflect positioning information at different levels of accuracy, providing a basis for subsequent sub-pixel positioning. For each confidence area obtained by segmentation, a sub-pixel image processing algorithm is used for analysis, and the grayscale gradient, edge information and other features within the area are used to calculate the sub-pixel offset value of each area. These offset values ​​describe the subtle position information within the area, providing a basis for subsequent fine positioning. The sub-pixel offset values ​​calculated above are applied to the original titanium mesh positioning points for fine-tuning. In this way, the positioning accuracy and details can be further improved on the basis of rough positioning, and finally a fine positioning point after sub-pixel fine-tuning can be obtained.

[0047] Step S6: Generate personalized bone structure feature data based on the oral alveolar three-dimensional point cloud model; reposition the titanium mesh error on the personalized bone structure feature data based on sub-pixel fine embedding positioning points, so as to obtain the optimal positioning result of the titanium mesh.

[0048] In this embodiment, a bone structure segmentation algorithm is used to extract the individual bone structure information of the patient, and feature extraction is performed on the segmented bone structure data, such as shape features, size features, curvature features, etc. These personalized bone structure feature data can provide prior information for subsequent titanium mesh positioning optimization. Based on fine sub-pixel positioning points, a comparative analysis is performed with the personalized bone structure feature data, and an optimization algorithm (such as feature-based rigid body transformation) is used to best match the two, and the final positioning error of the titanium mesh relative to the bone structure is calculated. Based on this error information, the original titanium mesh positioning result is optimized and adjusted to obtain a more accurate final positioning result.

[0049] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0050] Step S11: Acquire multi-directional images of the oral cavity;

[0051] Step S12: tracking the optical flow feature points of the oral cavity multi-directional image to extract the oral cavity optical flow feature points;

[0052] Step S13: performing image region segmentation on the oral cavity multi-directional image based on the oral cavity optical flow feature points to obtain a plurality of oral cavity region images;

[0053] Step S14: performing regional resolution calculation on a plurality of oral region images to generate oral region image resolution;

[0054] Step S15: performing super-resolution fine structure restoration processing on the oral multi-directional image based on the oral region image resolution to construct a super-resolution oral image.

[0055] In this embodiment, an intraoral imaging device (such as a 3D scanner or a high-definition camera) is used to collect multi-view image data of the patient's oral cavity from different angles. These image data cover the overall appearance and details of the oral cavity, providing sufficient visual information for subsequent image processing and analysis. An optical flow algorithm (such as the Lucas-Kanade algorithm) is used to track the feature points of the collected multi-view image sequence. By analyzing the motion trajectory of the midpoints in the image sequence, stable oral optical flow feature points are extracted. These feature points can describe the local motion characteristics of the oral tissue, providing a basis for subsequent regional segmentation. Using the extracted optical flow feature points, an image segmentation algorithm (such as K-Means clustering) is used to divide the original image into regions. The goal of segmentation is to separate regions with similar The areas with similar motion features are divided into independent sub-areas, so that multiple different local oral region images can be obtained, which provides a basis for subsequent resolution analysis and super-resolution processing. For each segmented local oral region image, its actual spatial resolution information is calculated. The resolution information can be inferred through the pixel size of the image, the parameters of the imaging device, etc., to generate a resolution map of each oral region, which provides a reference for subsequent super-resolution processing. For each oral region image, a super-resolution algorithm (such as SRCNN) is used to restore the fine structure. The resolution information calculated above is used to perform targeted super-resolution reconstruction on the undersampled image area, and finally a set of high-resolution full-view oral images are obtained.

[0056] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0057] Step S21: performing feature point spatial matching on the oral cavity multi-directional image based on the oral cavity optical flow feature points, thereby obtaining feature point spatial position data;

[0058] Step S22: performing alveolar morphological structure recognition on the super-resolution oral cavity image to generate alveolar morphological structure data;

[0059] Step S23: performing point cloud contour standardization processing on the alveolar morphological structure data to obtain an alveolar point cloud contour map;

[0060] Step S24: performing a planar connectivity analysis on the alveolar point cloud contour map to obtain alveolar cross-sectional spatial topological data;

[0061] Step S25: locating the three-dimensional position of the feature points of the tooth socket point cloud contour image according to the spatial position data of the feature points, thereby obtaining contour feature point positioning data;

[0062] Step S26: reconstructing the three-dimensional structure of the alveolar cross-section spatial topology data through the contour feature point positioning data to construct an oral alveolar three-dimensional point cloud model.

[0063] In this embodiment, according to the oral optical flow feature points, the corresponding feature points on the images of different view angles are linked by a feature point matching algorithm (such as SIFT and SURF). According to the two-dimensional coordinate position of the feature points in each view angle image, combined with the internal and external parameter information of the camera, the actual coordinates of each feature point in the three-dimensional space are calculated to obtain a set of three-dimensional position data of the feature points describing the overall spatial structure of the oral cavity, which provides a basis for subsequent three-dimensional reconstruction. By using image segmentation and pattern recognition technology, the alveolar area of ​​the high-resolution oral image constructed above is accurately extracted and modeled, and the alveolar structure features such as the position, shape, and size of each tooth are identified to generate a set of detailed alveolar morphological data. , these data describe the overall geometric structure of the oral alveolar, providing an important basis for subsequent 3D reconstruction. The identified alveolar morphological features are converted into discrete 3D point cloud data to form a digital contour expression of the alveolar. The point cloud data is subjected to a series of standardized processing such as filtering, smoothing, and refinement to make it more regular and continuous. Finally, a set of high-quality alveolar point cloud contour maps are obtained, which provide a good foundation for subsequent spatial topological analysis. The geometric relationship and spatial connectivity of each part in the point cloud contour map are analyzed, the continuous sections of the alveolar are identified, the plane equation and spatial position information of each section are calculated, and a set of spatial topological data describing the complex 3D structure of the alveolar is generated, laying the foundation for 3D reconstruction.

[0064] In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0065] Step S31: acquiring an oral titanium mesh image based on a preset database;

[0066] Step S32: visually positioning the oral titanium mesh image and marking the titanium mesh positioning point;

[0067] Step S33: extracting the structural features of the positioning points of the oral titanium mesh image based on the titanium mesh positioning points, thereby obtaining the structural feature data of the positioning points;

[0068] Step S34: performing deep convolution learning processing on the structural feature data of the positioning points to generate titanium mesh positioning point extraction rules;

[0069] Step S35: performing multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning point extraction rule to generate multiple titanium mesh positioning registration samples.

[0070] In this embodiment, a preset database containing oral titanium mesh images is established, covering titanium mesh implants of different types and sizes. A large number of high-quality titanium mesh images are extracted from the database to provide rich training samples for subsequent visual positioning and feature extraction, ensuring that these images can fully cover the various states and angles of titanium mesh implants in the oral cavity. The titanium mesh images in the database are manually or automatically positioned and marked using image segmentation and annotation tools. The boundary contour of the titanium mesh is accurately outlined in each image, and a set of key positioning points are marked. These positioning points describe the geometric position and spatial distribution characteristics of the titanium mesh in the image, laying the foundation for subsequent feature extraction. For each marked positioning point, the surrounding visual features such as texture, shape, and brightness are analyzed. , extract a set of data indicators that can accurately describe the structural characteristics of the positioning points, such as texture variance, gradient distribution, key point density, etc., and summarize these feature data into a complete feature vector for subsequent deep learning processing, oral positioning point structural feature data, train a deep convolutional neural network model, learn a set of visual feature extraction rules that can accurately identify the titanium mesh positioning points, import the constructed oral alveolar three-dimensional point cloud model, and apply the titanium mesh positioning point extraction rules for automatic identification, according to the distribution of multiple identified positioning points in three-dimensional space, use the registration algorithm (such as ICP) to iteratively optimize the model, and through multiple simulation experiments, generate a set of three-dimensional registration samples describing different titanium mesh positions and angles, providing a reliable basis for subsequent surgical planning.

[0071] In this embodiment, step S4 includes the following steps:

[0072] Step S41: Sliding window sampling is performed on multiple titanium mesh positioning and registration samples to extract multiple sample positioning point data;

[0073] Step S42: performing point-by-point positioning probability prediction on the multiple sample positioning point data to generate positioning point probability prediction values;

[0074] Step S43: performing positioning point marking segmentation on the oral alveolar three-dimensional point cloud model based on the multi-sample positioning point data to obtain multiple titanium mesh positioning maps;

[0075] Step S44: Confidence mapping is performed on multiple titanium mesh positioning maps based on the positioning point probability prediction values, thereby constructing a titanium mesh positioning confidence map.

[0076] In this embodiment, a sliding window method is used to sample data for multiple three-dimensional titanium mesh positioning and registration samples. For each registration sample, the window is gradually slid according to a certain window size and step length, and the positioning point coordinate data within the window is extracted. In this way, a large number of three-dimensional coordinate data sets describing the distribution of titanium mesh positioning points can be obtained, which provides a basis for subsequent probability prediction. A deep learning method is used to train a prediction model that can predict the probability of a given three-dimensional positioning point coordinate being a titanium mesh positioning point. A multi-sample positioning point data set is input, and the probability prediction value of each positioning point is obtained after model training and optimization. This probability prediction value can reflect the reliability of the positioning point, which is provided for subsequent In order to provide a basis for confidence assessment, the multi-sample positioning point data obtained are superimposed on the original three-dimensional alveolar point cloud model, and each positioning point is marked and segmented to generate a series of two-dimensional positioning maps describing the position of the titanium mesh. High-probability areas are assigned higher confidence values, while low-probability areas are assigned lower confidence values. Through this weighted fusion, a complete titanium mesh positioning confidence map can be constructed, which intuitively shows the reliability of the positioning results. This confidence map provides an important basis for subsequent surgical planning, which can help doctors evaluate surgical risks and formulate more accurate plans. These positioning maps can intuitively reflect the distribution of titanium mesh in the alveolar model and provide a visual reference for surgical planning.

[0077] In this embodiment, the specific steps of step S5 are:

[0078] Step S51: performing non-maximum suppression on the titanium mesh positioning confidence map to extract local coarse positioning points;

[0079] Step S52: segmenting the surrounding area of ​​the titanium mesh positioning confidence map based on the local coarse positioning points to obtain a local confidence image;

[0080] Step S53: performing residual connection filling processing on the local confidence image to obtain a multi-scale confidence image;

[0081] Step S54: predicting the sub-pixel offset value of the multi-scale confidence image to obtain the image sub-pixel offset value;

[0082] Step S55: Based on the image sub-pixel offset value, the local coarse positioning point is input into the oral alveolar three-dimensional point cloud model, and sub-pixel positioning fine-tuning is performed to obtain sub-pixel fine-embedded positioning points.

[0083] In this embodiment, a non-maximum suppression algorithm is used to process it. Non-maximum suppression can remove points with low positioning confidence in the local area and only retain the local maximum point. In this way, a group of candidate coarse positioning points can be extracted to provide an initial reference for the subsequent fine positioning. For the extracted coarse positioning points, a local area is delineated on the titanium mesh positioning confidence map. The size of each local area can be adjusted as needed. Usually, a suitable window size is selected. In this way, a series of local confidence images can be obtained to provide input data for subsequent sub-pixel prediction. The residual connection structure of the convolutional neural network is used to fill the local confidence image generated in step S52. The residual connection can effectively extract The multi-scale features in the image generate confidence images containing rich detail information. These multi-scale confidence images provide richer input data for subsequent sub-pixel offset prediction. The generated multi-scale confidence images are used as input to train a deep learning model for predicting sub-pixel offset values. The model can learn the sub-pixel level detail information contained in the image and predict the offset of each pixel. Through this prediction, an image describing the distribution of sub-pixel offsets can be obtained, which provides a basis for subsequent fine positioning. The extracted coarse positioning points and the predicted sub-pixel offset values ​​are fine-tuned on the original three-dimensional alveolar point cloud model. By performing sub-pixel offset correction on the coarse positioning points, a more refined positioning result can be obtained.

[0084] In this embodiment, the specific steps of step S6 are:

[0085] Step S61: performing personalized bone structure analysis on the oral alveolar three-dimensional point cloud model to generate personalized bone structure feature data;

[0086] Step S62: performing optical flow posture correction on the personalized bone structure feature data based on the sub-pixel fine embedding positioning points to obtain personalized bone structure correction data;

[0087] Step S63: Use the personalized bone structure calibration data to reposition the sub-pixel fine-embedded positioning points to correct the titanium mesh errors, thereby obtaining the optimal positioning result of the titanium mesh.

[0088] In this embodiment, the obtained three-dimensional point cloud model of the oral cavity is used to extract the point cloud data of the alveolar bone area, and the alveolar bone point cloud data is refined and segmented and clustered to identify features such as bone density and structural texture. Based on these characteristic parameters, a feature vector representing the individual alveolar bone structure is constructed. This posture correction process can eliminate the positioning deviation caused by factors such as changes in the patient's alveolar posture. A group of finely embedded positioning points are marked on the three-dimensional point cloud model using sub-pixel feature points. These positioning points are aligned and proofread with the personalized bone structure feature data, and the posture of the positioning points is optimized by the optical flow algorithm to finally obtain the corrected personalized bone structure proofreading data. Based on the personalized bone structure proofreading data, the sub-pixel positioning points are error corrected, and the optimal installation position and angle of the titanium mesh in the three-dimensional space are calculated according to the corrected positioning point coordinates to generate a result containing precise positioning information.

[0089] In this specification, a titanium mesh intelligent positioning device is also provided, including:

[0090] An image processing module is used to obtain a multi-directional image of the oral cavity; extract oral cavity optical flow feature points based on the multi-directional image of the oral cavity; and perform super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image;

[0091] The three-dimensional point cloud module is used to perform spatial matching of feature points of multi-directional oral images based on oral optical flow feature points, so as to obtain spatial position data of feature points; perform point cloud contour standardization processing on super-resolution oral images to obtain alveolar point cloud contour map; perform three-dimensional structure reconstruction of alveolar point cloud contour map through feature point spatial position data, and construct an oral alveolar three-dimensional point cloud model;

[0092] The positioning and registration simulation module is used to obtain the oral titanium mesh image based on the preset database; perform visual positioning frame selection on the oral titanium mesh image and mark the titanium mesh positioning points; perform multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning points to generate multiple titanium mesh positioning and registration samples;

[0093] The probability prediction module is used to perform point-by-point positioning probability prediction on multiple titanium mesh positioning and registration samples to generate positioning point probability prediction values; perform confidence mapping based on the positioning point probability prediction values ​​to construct a titanium mesh positioning confidence map;

[0094] The sub-pixel fine-tuning module is used to segment the surrounding area of ​​the titanium mesh positioning confidence map to obtain a multi-scale confidence image; obtain the image sub-pixel offset value based on the multi-scale confidence image; perform sub-pixel positioning fine-tuning based on the image sub-pixel offset value to obtain a sub-pixel fine-embedded positioning point;

[0095] The optimal positioning module is used to generate personalized bone structure feature data based on the oral alveolar three-dimensional point cloud model; the personalized bone structure feature data is relocated based on the sub-pixel fine embedding positioning points to obtain the optimal positioning result of the titanium mesh.

[0096] The present invention also provides a computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, and wherein the processor implements the steps of any one of the above-mentioned titanium mesh intelligent positioning methods when executing the computer program.

[0097] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any of the above-mentioned titanium mesh intelligent positioning methods are implemented.

[0098] Those skilled in the art clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0099] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it is stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or partly or all or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media for storing program codes such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks.

[0100] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0101] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A titanium mesh intelligent positioning method, characterized in that: The following steps are involved: Step S1: acquiring a multi-directional image of the oral cavity; extracting oral cavity optical flow feature points based on the multi-directional image of the oral cavity; and performing super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image; Step S2: performing spatial matching of feature points of the oral multi-directional image based on the oral optical flow feature points to obtain spatial position data of the feature points; performing point cloud contour normalization processing on the super-resolution oral image to obtain a tooth socket point cloud contour map; reconstructing the three-dimensional structure of the tooth socket point cloud contour map through the feature point spatial position data to construct a three-dimensional point cloud model of the oral cavity; Step S3: acquiring an oral titanium mesh image based on a preset database; Visually locate and frame the oral titanium mesh image and mark the titanium mesh positioning point; Based on the titanium mesh positioning points, the three-dimensional point cloud model of the oral alveolar is iteratively simulated for multi-positioning point registration to generate multiple titanium mesh positioning and registration samples; Step S4: perform point-by-point positioning probability prediction on multiple titanium mesh positioning and registration samples to generate positioning point probability prediction values; perform confidence mapping based on the positioning point probability prediction values ​​to construct a titanium mesh positioning confidence map; the specific steps of step S4 are: Step S41: Sliding window sampling is performed on multiple titanium mesh positioning and registration samples to extract multiple sample positioning point data; Step S42: performing point-by-point positioning probability prediction on the multiple sample positioning point data to generate positioning point probability prediction values; Step S43: performing positioning point marking segmentation on the oral alveolar three-dimensional point cloud model based on the multi-sample positioning point data to obtain multiple titanium mesh positioning maps; Step S44: Confidence mapping is performed on multiple titanium mesh positioning maps based on the positioning point probability prediction value, thereby constructing a titanium mesh positioning confidence map; Step S5: segment the surrounding area of ​​the titanium mesh positioning confidence map to obtain a multi-scale confidence image; obtain the image sub-pixel offset value based on the multi-scale confidence image; perform sub-pixel positioning fine-tuning based on the image sub-pixel offset value to obtain a sub-pixel fine-embedded positioning point; the specific steps of step S5 are: Step S51: performing non-maximum suppression on the titanium mesh positioning confidence map to extract local coarse positioning points; Step S52: segmenting the surrounding area of ​​the titanium mesh positioning confidence map based on the local coarse positioning points to obtain a local confidence image; Step S53: performing residual connection filling processing on the local confidence image to obtain a multi-scale confidence image; Step S54: predicting the sub-pixel offset value of the multi-scale confidence image to obtain the image sub-pixel offset value; Step S55: inputting the local coarse positioning point into the oral alveolar three-dimensional point cloud model based on the image sub-pixel offset value, performing sub-pixel positioning fine adjustment, and obtaining the sub-pixel fine embedding positioning point; Step S6: generating personalized bone structure feature data based on the oral alveolar three-dimensional point cloud model; The titanium mesh error is repositioned based on the personalized bone structure feature data based on sub-pixel fine-embedded positioning points, thereby obtaining the optimal positioning result of the titanium mesh.

2. The titanium mesh intelligent positioning method according to claim 1, characterized in that: The specific steps of step S1 are: Step S11: Acquire multi-directional images of the oral cavity; Step S12: tracking the optical flow feature points of the oral cavity multi-directional image to extract the oral cavity optical flow feature points; Step S13: performing image region segmentation on the oral cavity multi-directional image based on the oral cavity optical flow feature points to obtain a plurality of oral cavity region images; Step S14: performing regional resolution calculation on a plurality of oral region images to generate oral region image resolution; Step S15: performing super-resolution fine structure restoration processing on the oral multi-directional image based on the oral region image resolution to construct a super-resolution oral image.

3. The titanium mesh intelligent positioning method according to claim 1, characterized in that: The specific steps of step S2 are: Step S21: performing feature point spatial matching on the oral cavity multi-directional image based on the oral cavity optical flow feature points, thereby obtaining feature point spatial position data; Step S22: performing alveolar morphological structure recognition on the super-resolution oral cavity image to generate alveolar morphological structure data; Step S23: performing point cloud contour standardization processing on the alveolar morphological structure data to obtain an alveolar point cloud contour map; Step S24: performing a planar connectivity analysis on the alveolar point cloud contour map to obtain alveolar cross-sectional spatial topological data; Step S25: locating the three-dimensional position of the feature points of the tooth socket point cloud contour image according to the spatial position data of the feature points, thereby obtaining contour feature point positioning data; Step S26: reconstructing the three-dimensional structure of the alveolar cross-section spatial topology data through the contour feature point positioning data to construct an oral alveolar three-dimensional point cloud model.

4. The titanium mesh intelligent positioning method according to claim 1, characterized in that: The specific steps of step S3 are: Step S31: acquiring an oral titanium mesh image based on a preset database; Step S32: visually positioning the oral titanium mesh image and marking the titanium mesh positioning point; Step S33: extracting the structural features of the positioning points of the oral titanium mesh image based on the titanium mesh positioning points, thereby obtaining the structural feature data of the positioning points; Step S34: performing deep convolution learning processing on the structural feature data of the positioning points to generate titanium mesh positioning point extraction rules; Step S35: performing multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning point extraction rule to generate multiple titanium mesh positioning registration samples.

5. The titanium mesh intelligent positioning method according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: performing personalized bone structure analysis on the oral alveolar three-dimensional point cloud model to generate personalized bone structure feature data; Step S62: performing optical flow posture correction on the personalized bone structure feature data based on the sub-pixel fine embedding positioning points to obtain personalized bone structure correction data; Step S63: Use the personalized bone structure calibration data to reposition the sub-pixel fine-embedded positioning points to correct the titanium mesh errors, thereby obtaining the optimal positioning result of the titanium mesh.

6. A titanium mesh precise positioning device based on image processing, characterized in that: Used to execute the titanium mesh intelligent positioning method according to claim 1, comprising: An image processing module is used to obtain a multi-directional image of the oral cavity; extract oral cavity optical flow feature points based on the multi-directional image of the oral cavity; and perform super-resolution fine structure restoration processing on the multi-directional image of the oral cavity based on the oral cavity optical flow feature points to construct a super-resolution oral cavity image; The three-dimensional point cloud module is used to perform spatial matching of feature points of multi-directional oral images based on oral optical flow feature points, so as to obtain spatial position data of feature points; perform point cloud contour standardization processing on super-resolution oral images to obtain alveolar point cloud contour map; perform three-dimensional structure reconstruction of alveolar point cloud contour map through feature point spatial position data, and construct an oral alveolar three-dimensional point cloud model; The positioning and registration simulation module is used to obtain the oral titanium mesh image based on the preset database; perform visual positioning frame selection on the oral titanium mesh image and mark the titanium mesh positioning points; perform multi-positioning point registration iterative simulation on the oral alveolar three-dimensional point cloud model based on the titanium mesh positioning points to generate multiple titanium mesh positioning and registration samples; The probability prediction module is used to perform point-by-point positioning probability prediction on multiple titanium mesh positioning and registration samples to generate positioning point probability prediction values; perform confidence mapping based on the positioning point probability prediction values ​​to construct a titanium mesh positioning confidence map; The sub-pixel fine-tuning module is used to segment the surrounding area of ​​the titanium mesh positioning confidence map to obtain a multi-scale confidence image; obtain the image sub-pixel offset value based on the multi-scale confidence image; perform sub-pixel positioning fine-tuning based on the image sub-pixel offset value to obtain a sub-pixel fine-embedded positioning point; The optimal positioning module is used to generate personalized bone structure feature data based on the oral alveolar three-dimensional point cloud model; the personalized bone structure feature data is relocated based on the sub-pixel fine embedding positioning points to obtain the optimal positioning result of the titanium mesh.

7. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the titanium mesh intelligent positioning method described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the titanium mesh intelligent positioning method described in any one of claims 1 to 5 are implemented.

Images