A tooth crack detection method based on 3D-DIC

Abstract

The present invention discloses a tooth crack detection method based on 3D-DIC, comprising the following steps: S1: spraying black ink on the tooth surface with an airbrush so that the ink is evenly distributed on the tooth surface in the form of dense spots, wherein the spot set is called a speckle pattern, acquiring an image of the tooth through a binocular camera, wherein the image is divided into a left view and a right view, and obtaining the left and right views without distortion after correction, S2: extracting the area where the teeth are located in the left view and the right view, and obtaining an image containing only the teeth, S3: performing stereo matching on the area where the teeth are located in the left view and the right view, S4: calculating and obtaining a disparity map, S5: post-processing the disparity map, S6: generating a local point cloud map and generating a complete point cloud map after fusion, S7: applying pressure to the tooth and repeating the above steps to obtain another point cloud, S8: comparing the point clouds using a 3D-DIC method to obtain a three-dimensional strain field, and S9: analyzing the strain field to obtain the main crack position.

Description

Technical Field

[0001] The present invention relates to the technical field of tooth crack detection, and in particular to a tooth crack detection method based on 3D-DIC. Background Art

[0002] Hidden cracks in teeth are tiny and difficult to detect, which can cause a series of lesions in the tooth body, pulp, periapical and periodontal tissues. In modern medicine, the identification process of hidden cracks in teeth is increasingly dependent on digitization, and a series of methods based on computer vision have been born. In the existing technology, the radiation from CT can cause certain harm to the human body, and non-visual methods such as TOF cannot intuitively reconstruct the details of the teeth. Visual methods are more intuitive for humans, and the equipment cost is lower than other methods. However, due to the limitations of the accuracy of existing cameras, the measurement accuracy cannot be met in some cases, so digital imaging technology is needed for analysis and detection, and three-dimensional imaging can more intuitively present the crack situation. Therefore, a method that can detect tooth cracks and perform three-dimensional imaging has become necessary. Summary of the invention

[0003] The purpose of the present invention is to provide a tooth crack detection method based on 3D-DIC, which solves the problems in the above-mentioned background technology through various improvements.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A tooth crack detection method based on 3D-DIC, comprising the following methods:

[0006] S1: Use an airbrush to spray black ink on the tooth surface so that the ink is evenly distributed on the tooth surface in dense spots. The spot set is called a speckle pattern. The image of the tooth is obtained by a binocular camera. The image is divided into a left view and a right view. After correction, the distortion-free left view and right view are obtained.

[0007] S2: extracting the area where the teeth are located in the left view and the right view to obtain an image containing only the teeth;

[0008] S3: Stereo matching of the tooth area in the left view and the right view. Stereo matching means finding the matching pixel in the right view for a pixel in the left view, and calculating the matching cost for each pair of pixels.

[0009] S4: Calculate the disparity value of each pair of pixels to obtain a disparity map;

[0010] S5: post-processing the disparity map to obtain a smooth disparity map;

[0011] S6: According to the triangulation principle, the disparity map is mapped into the three-dimensional space to form a point cloud. Then the next pair of images is processed to obtain the point cloud, which is then registered with the point cloud of the previous pair of images to obtain a fused point cloud.

[0012] S7: Apply pressure parallel to the tooth cross section to the side of the target tooth, and the pressure does not exceed 2N. Repeat steps S1-S6 for the tooth after the pressure is applied to obtain another point cloud;

[0013] S8: Compare the two point clouds obtained in S7, calculate the displacement field according to the 3D-DIC method, and further calculate the strain field.

[0014] S9: Analyze the strain field and obtain the location of the main crack.

[0015] Preferably, in S1, the camera rotates about the central axis of the tooth, acquires a pair of images every 5 degrees of rotation, and corrects the images.

[0016] Preferably, the stereo matching in S3 includes cost calculation and cost aggregation, and the cost calculation criterion selects ZNCC (normalized cross-correlation criterion). The pairing point of each pixel of the left view is found in the right view. The higher the cost value, the higher the similarity. By using the limit constraint, the search range can be reduced and the consumption of computing resources can be reduced. The epipolar constraint, that is, in the left view and the right view taken by the calibrated binocular camera, each pair of matching pixels should be on a straight line.

[0017] Preferably, after the cost calculation in S3 is completed, a cost space is obtained, and cost aggregation is performed in the cost space. The cost aggregation is combined with global information to optimize the pixel cost value to obtain a more accurate cost value. After the cost matching is completed, each pixel selects its lowest cost value to calculate the disparity map.

[0018] Preferably, the disparity map in S4 still has certain noise due to mismatching after stereo matching, and needs to be post-processed such as left-right consistency check, median filtering, and filling holes. The left-right consistency check is to swap the left and right views and perform stereo matching again. Pixels whose offset distances before and after the check exceed the limit are regarded as mismatches and are deleted. The median filtering is used to remove noise, smooth the disparity map, and fill holes, which is to fill the missing holes.

[0019] Preferably, after the second group of point clouds are generated in S6, due to errors, the generated point clouds are offset by a certain distance, and the two point clouds need to be registered and fused into one point cloud, and the fusion adopts the nearest neighbor iterative algorithm.

[0020] Preferably, the above steps are repeated until all images are calculated, and preliminary three-dimensional point cloud data of the teeth is obtained. After the point cloud image is obtained, it needs to be post-processed. The post-processing includes: filtering to remove outliers, filling holes, etc., and the point cloud is reconstructed to obtain a three-dimensional model. The surface reconstruction adopts the Poisson reconstruction method.

[0021] Preferably, in S7, pressure is applied to both sides of the tooth, and steps S1 to S7 are repeated to obtain a three-dimensional model of the tooth after the pressure is applied.

[0022] Preferably, the three-dimensional tooth model before and after stress deformation is compared, and the points with higher correlation are matched through the three-dimensional digital image correlation method (3D-DIC), and the displacement information and strain information of the point pair are calculated. After completing the matching of all points in the point cloud set, the three-dimensional strain field of the entire tooth is obtained.

[0023] Preferably, the method utilizes the characteristic that the crack area is more sensitive to pressure to find the position with the largest strain in the three-dimensional strain field, that is, the position where the crack is located. According to the above principle, a threshold is first set to eliminate the interference of small crack branches on the main crack, and then the search step size is set. After iterative extraction, the high strain area in the three-dimensional strain field is the area where the crack is located.

[0024] The tooth crack detection method based on 3D-DIC provided by the present invention has the following beneficial effects:

[0025] The present invention extracts the tooth area in the image, performs stereo matching on the left and right views, obtains a disparity map, and post-processes the disparity map, thereby reducing the search range and reducing the consumption of computing resources. Compared with existing detection methods, the present invention requires simple equipment, has low cost, and can intuitively display the location of tooth cracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Schematic diagram of the process of tooth crack detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Embodiment 1

[0028] like Figure 1 As shown, the tooth crack detection method based on 3D-DIC provided by the present invention comprises the following steps:

[0029] S1: Use an airbrush to spray black ink on the tooth surface so that the ink is evenly distributed on the tooth surface in dense spots. The spot set is called a speckle pattern. The image of the tooth is obtained by a binocular camera. The image is divided into a left view and a right view. After correction, the distortion-free left view and right view are obtained.

[0030] S2: extracting the area where the teeth are located in the left view and the right view to obtain an image containing only the teeth;

[0031] S3: Stereo matching is performed on the area where the teeth are located in the left view and the right view. Stereo matching means finding the matching pixel in the right view for a pixel in the left view, and calculating the matching cost for each pair of pixels. Stereo matching includes cost calculation and cost aggregation. The cost calculation criterion uses ZNCC (normalized cross-correlation criterion). The pairing point of each pixel of the left view is found in the right view. The higher the cost value, the higher the similarity. By using the limit constraint, the search range can be reduced and the consumption of computing resources can be reduced. The epipolar constraint means that each pair of matching pixels in the left view and the right view taken by the calibrated binocular camera should be on a straight line. After the cost calculation is completed, the cost space is obtained, and cost aggregation is performed in the cost space. The cost aggregation combines global information to optimize the pixel cost value to obtain a more accurate cost value. After the cost matching is completed, each pixel selects its lowest cost value to calculate the disparity map.

[0032] S4: Calculate the disparity value of each pair of pixels to obtain a disparity map. After stereo matching, the disparity map still has certain noise due to mismatching, etc., and needs to be post-processed by left-right consistency check, median filtering, filling holes, etc. The left-right consistency check is to swap the left and right views and perform stereo matching again. Pixels whose offset distances before and after the check exceed the limit are regarded as mismatches and deleted. The median filtering is used to remove noise points, smooth the disparity map, and fill holes, which is to fill the missing holes.

[0033] S5: post-processing the disparity map to obtain a smooth disparity map;

[0034] S6: According to the triangulation principle, the disparity map is mapped into the three-dimensional space to form a point cloud. Next, the next pair of images is processed to obtain the point cloud, which is then registered with the point cloud of the previous pair of images to obtain a fused point cloud. After the second group of point clouds is generated, due to errors, the generated point clouds have a certain distance offset. The two point clouds need to be registered and fused into one point cloud. The fusion adopts the nearest neighbor iteration algorithm.

[0035] S7: Apply pressure parallel to the tooth cross section to the side of the target tooth. When the pressure does not exceed 0.5N, apply pressure to both sides of the tooth. Repeat the steps to obtain a three-dimensional model of the tooth after applying pressure. Compare the two tooth models. Calculate the time of their displacement field for 1 second. Find the location where the displacement is significant, which is the location of the crack. Repeat the above steps until all pictures are calculated. At this time, preliminary three-dimensional point cloud data of the tooth is obtained. After obtaining the point cloud map, it needs to be post-processed. The post-processing includes: filtering to remove outliers, filling holes, etc., and performing surface reconstruction on the point cloud to obtain a three-dimensional model. The surface reconstruction adopts the Poisson reconstruction method.

[0036] S8: Compare the two point clouds obtained in S7, calculate the displacement field according to the 3D-DIC method, and further calculate the strain field. Finally, analyze the strain field to obtain the position of the main crack, compare the three-dimensional model of the tooth before and after stress deformation, match the points with high correlation through the three-dimensional digital image correlation method (3D-DIC), and calculate the displacement information and strain information of the point pair. After completing the matching of all points in the point cloud set, the three-dimensional strain field of the entire tooth is obtained.

[0037] S9: Taking advantage of the fact that the crack area is more sensitive to pressure, the position with the largest strain in the three-dimensional strain field, that is, the position where the crack is located, is found. According to the above principle, the threshold is first set to eliminate the interference of small crack branches on the main crack, and then the search step size is set. After iteration, the high strain area in the three-dimensional strain field is extracted, which is the area where the crack is located.

[0038] Embodiment 2

[0039] like Figure 1 As shown, the tooth crack detection method based on 3D-DIC of the present invention comprises the following steps:

[0040] S1: Use an airbrush to spray black ink on the tooth surface so that the ink is evenly distributed on the tooth surface in dense spots. The spot set is called a speckle pattern. The image of the tooth is obtained by a binocular camera. The image is divided into a left view and a right view. After correction, the distortion-free left view and right view are obtained.

[0041] S2: extracting the area where the teeth are located in the left view and the right view to obtain an image containing only the teeth;

[0042] S3: Stereo matching is performed on the area where the teeth are located in the left view and the right view. Stereo matching means finding the matching pixel in the right view for a pixel in the left view, and calculating the matching cost for each pair of pixels. Stereo matching includes cost calculation and cost aggregation. The cost calculation criterion uses ZNCC (normalized cross-correlation criterion). The pairing point of each pixel of the left view is found in the right view. The higher the cost value, the higher the similarity. By using the limit constraint, the search range can be reduced and the consumption of computing resources can be reduced. The epipolar constraint means that each pair of matching pixels in the left view and the right view taken by the calibrated binocular camera should be on a straight line. After the cost calculation is completed, the cost space is obtained, and cost aggregation is performed in the cost space. The cost aggregation combines global information to optimize the pixel cost value to obtain a more accurate cost value. After the cost matching is completed, each pixel selects its lowest cost value to calculate the disparity map.

[0043] S4: Calculate the disparity value of each pair of pixels to obtain a disparity map. After stereo matching, the disparity map still has certain noise due to mismatching, etc., and needs to be post-processed by left-right consistency check, median filtering, filling holes, etc. The left-right consistency check is to swap the left and right views and perform stereo matching again. Pixels whose offset distances before and after the check exceed the limit are regarded as mismatches and deleted. The median filtering is used to remove noise points, smooth the disparity map, and fill holes, which is to fill the missing holes.

[0044] S5: post-processing the disparity map to obtain a smooth disparity map;

[0045] S6: According to the triangulation principle, the disparity map is mapped into the three-dimensional space to form a point cloud. Next, the next pair of images is processed to obtain the point cloud, which is then registered with the point cloud of the previous pair of images to obtain a fused point cloud. After the second group of point clouds is generated, due to errors, the generated point clouds have a certain distance offset. The two point clouds need to be registered and fused into one point cloud. The fusion adopts the nearest neighbor iteration algorithm.

[0046] S7: Apply pressure parallel to the tooth cross section to the side of the target tooth. When the pressure does not exceed 1N, apply pressure to both sides of the tooth. Repeat the steps to obtain a three-dimensional model of the tooth after applying pressure. Compare the two tooth models. Calculate the time of their displacement field for 2 seconds. Find the location where the displacement is significant, which is the location of the crack. Repeat the above steps until all pictures are calculated. At this time, preliminary three-dimensional point cloud data of the tooth is obtained. After obtaining the point cloud map, it needs to be post-processed. The post-processing includes: filtering to remove outliers, filling holes, etc., and performing surface reconstruction on the point cloud to obtain a three-dimensional model. The surface reconstruction adopts the Poisson reconstruction method.

[0047] S8: Compare the two point clouds obtained in S7, calculate the displacement field according to the 3D-DIC method, and further calculate the strain field. Finally, analyze the strain field to obtain the location of the main crack, compare the three-dimensional model of the tooth before and after stress deformation, match the points with high correlation through the three-dimensional digital image correlation method (3D-DIC), and calculate the displacement information and strain information of the point pair. After completing the matching of all points in the point cloud set, the three-dimensional strain field of the entire tooth is obtained.

[0048] S9: Taking advantage of the fact that the crack area is more sensitive to pressure, the position with the largest strain in the three-dimensional strain field, that is, the position where the crack is located, is found. According to the above principle, the threshold is first set to eliminate the interference of small crack branches on the main crack, and then the search step size is set. After iteration, the high strain area in the three-dimensional strain field is extracted, which is the area where the crack is located.

[0049] Embodiment 3

[0050] like Figure 1As shown, the tooth crack detection method based on 3D-DIC provided by the present invention comprises the following steps:

[0051] S1: Use an airbrush to spray black ink on the tooth surface so that the ink is evenly distributed on the tooth surface in dense spots. The spot set is called a speckle pattern. The image of the tooth is obtained by a binocular camera. The image is divided into a left view and a right view. After correction, the distortion-free left view and right view are obtained.

[0052] S2: extracting the area where the teeth are located in the left view and the right view to obtain an image containing only the teeth;

[0053] S3: Stereo matching is performed on the area where the teeth are located in the left view and the right view. Stereo matching means finding the matching pixel in the right view for a pixel in the left view, and calculating the matching cost for each pair of pixels. Stereo matching includes cost calculation and cost aggregation. The cost calculation criterion uses ZNCC (normalized cross-correlation criterion). The pairing point of each pixel of the left view is found in the right view. The higher the cost value, the higher the similarity. By using the limit constraint, the search range can be reduced and the consumption of computing resources can be reduced. The epipolar constraint means that each pair of matching pixels in the left view and the right view taken by the calibrated binocular camera should be on a straight line. After the cost calculation is completed, the cost space is obtained, and cost aggregation is performed in the cost space. The cost aggregation combines global information to optimize the pixel cost value to obtain a more accurate cost value. After the cost matching is completed, each pixel selects its lowest cost value to calculate the disparity map.

[0054] S4: Calculate the disparity value of each pair of pixels to obtain a disparity map. After stereo matching, the disparity map still has certain noise due to mismatching, etc., and needs to be post-processed by left-right consistency check, median filtering, filling holes, etc. The left-right consistency check is to swap the left and right views and perform stereo matching again. Pixels whose offset distances before and after the check exceed the limit are regarded as mismatches and deleted. The median filtering is used to remove noise points, smooth the disparity map, and fill holes, which is to fill the missing holes.

[0055] S5: post-processing the disparity map to obtain a smooth disparity map;

[0056] S6: According to the triangulation principle, the disparity map is mapped into the three-dimensional space to form a point cloud. Next, the next pair of images is processed to obtain the point cloud, which is then registered with the point cloud of the previous pair of images to obtain a fused point cloud. After the second group of point clouds is generated, due to errors, the generated point clouds have a certain distance offset. The two point clouds need to be registered and fused into one point cloud. The fusion adopts the nearest neighbor iteration algorithm.

[0057] S7: Apply pressure parallel to the tooth cross section to the side of the target tooth. When the pressure does not exceed 1.5N, apply pressure to both sides of the tooth. Repeat the steps to obtain a three-dimensional model of the tooth after applying pressure. Compare the two tooth models and calculate the time of their displacement field for 3 seconds. Find the location where the displacement is significant, which is the location of the crack. Repeat the above steps until all pictures are calculated. At this time, preliminary three-dimensional point cloud data of the tooth is obtained. After obtaining the point cloud map, it needs to be post-processed. The post-processing includes: filtering to remove outliers, filling holes, etc., and performing surface reconstruction on the point cloud to obtain a three-dimensional model. The surface reconstruction adopts the Poisson reconstruction method.

[0058] S8: Compare the two point clouds obtained in S7, calculate the displacement field according to the 3D-DIC method, and further calculate the strain field. Finally, analyze the strain field to obtain the location of the main crack, compare the three-dimensional model of the tooth before and after stress deformation, match the points with high correlation through the three-dimensional digital image correlation method (3D-DIC), and calculate the displacement information and strain information of the point pair. After completing the matching of all points in the point cloud set, the three-dimensional strain field of the entire tooth is obtained.

[0059] S9: Taking advantage of the fact that the crack area is more sensitive to pressure, the position with the largest strain in the three-dimensional strain field, that is, the position where the crack is located, is found. According to the above principle, the threshold is first set to eliminate the interference of small crack branches on the main crack, and then the search step size is set. After iteration, the high strain area in the three-dimensional strain field is extracted, which is the area where the crack is located.

[0060] Embodiment 4

[0061] like Figure 1 As shown, the tooth crack detection method based on 3D-DIC provided by the present invention comprises the following steps:

[0062] S1: Use an airbrush to spray black ink on the tooth surface so that the ink is evenly distributed on the tooth surface in dense spots. The spot set is called a speckle pattern. The image of the tooth is obtained by a binocular camera. The image is divided into a left view and a right view. After correction, the distortion-free left view and right view are obtained.

[0063] S2: extracting the area where the teeth are located in the left view and the right view to obtain an image containing only the teeth;

[0064] S3: Stereo matching is performed on the area where the teeth are located in the left view and the right view. Stereo matching means finding the matching pixel in the right view for a pixel in the left view, and calculating the matching cost for each pair of pixels. Stereo matching includes cost calculation and cost aggregation. The cost calculation criterion uses ZNCC (normalized cross-correlation criterion). The pairing point of each pixel of the left view is found in the right view. The higher the cost value, the higher the similarity. By using the limit constraint, the search range can be reduced and the consumption of computing resources can be reduced. The epipolar constraint means that each pair of matching pixels in the left view and the right view taken by the calibrated binocular camera should be on a straight line. After the cost calculation is completed, the cost space is obtained, and cost aggregation is performed in the cost space. The cost aggregation combines global information to optimize the pixel cost value to obtain a more accurate cost value. After the cost matching is completed, each pixel selects its lowest cost value to calculate the disparity map.

[0065] S4: Calculate the disparity value of each pair of pixels to obtain a disparity map. After stereo matching, the disparity map still has certain noise due to mismatching, etc., and needs to be post-processed by left-right consistency check, median filtering, filling holes, etc. The left-right consistency check is to swap the left and right views and perform stereo matching again. Pixels whose offset distances before and after the check exceed the limit are regarded as mismatches and deleted. The median filtering is used to remove noise points, smooth the disparity map, and fill holes, which is to fill the missing holes.

[0066] S5: post-processing the disparity map to obtain a smooth disparity map;

[0067] S6: According to the triangulation principle, the disparity map is mapped into the three-dimensional space to form a point cloud. Next, the next pair of images is processed to obtain the point cloud, which is then registered with the point cloud of the previous pair of images to obtain a fused point cloud. After the second group of point clouds is generated, due to errors, the generated point clouds have a certain distance offset. The two point clouds need to be registered and fused into one point cloud. The fusion adopts the nearest neighbor iteration algorithm.

[0068] S7: Apply pressure parallel to the tooth cross section to the side of the target tooth. When the pressure does not exceed 2N, apply pressure to both sides of the tooth. Repeat the steps to obtain a three-dimensional model of the tooth after applying pressure. Compare the two tooth models and calculate the time of their displacement field, which is 4 seconds. Find the location where the displacement is significant, which is the location of the crack. Repeat the above steps until all pictures are calculated. At this time, preliminary three-dimensional point cloud data of the tooth is obtained. After obtaining the point cloud map, it needs to be post-processed. The post-processing includes: filtering to remove outliers, filling holes, etc., and performing surface reconstruction on the point cloud to obtain a three-dimensional model. The surface reconstruction adopts the Poisson reconstruction method;

[0069] S8: Compare the two point clouds obtained in S7, calculate the displacement field according to the 3D-DIC method, and further calculate the strain field. Finally, analyze the strain field to obtain the location of the main crack, compare the three-dimensional model of the tooth before and after stress deformation, match the points with high correlation through the three-dimensional digital image correlation method (3D-DIC), and calculate the displacement information and strain information of the point pair. After completing the matching of all points in the point cloud set, the three-dimensional strain field of the entire tooth is obtained.

[0070] S9: Taking advantage of the fact that the crack area is more sensitive to pressure, the position with the largest strain in the three-dimensional strain field, that is, the position where the crack is located, is found. According to the above principle, the threshold is first set to eliminate the interference of small crack branches on the main crack, and then the search step size is set. After iteration, the high strain area in the three-dimensional strain field is extracted, which is the area where the crack is located.

[0071] The above embodiments of the present invention focus on extracting the area of ​​the teeth in the image, performing stereo matching on the left and right views, obtaining a disparity map, and post-processing the disparity map, which can reduce the search range and reduce computing resource consumption. Compared with existing detection methods, the present invention requires simple equipment, has low cost, and can intuitively display the location of tooth cracks.

[0072] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A tooth crack detection method based on 3D-DIC, characterized in that: The following steps are involved: S1: Use an airbrush to spray black ink on the tooth surface so that the ink is evenly distributed on the tooth surface in dense spots. The spot set is called a speckle pattern. The image of the tooth is obtained by a binocular camera. The image is divided into a left view and a right view. After correction, the distortion-free left view and right view are obtained. S2: extracting the area where the teeth are located in the left view and the right view to obtain an image containing only the teeth; S3: Stereo matching of the tooth area in the left view and the right view. Stereo matching means finding the matching pixel in the right view for a pixel in the left view, and calculating the matching cost for each pair of pixels. The stereo matching in S3 includes cost calculation and cost aggregation. The cost calculation criterion uses the ZNCC normalized cross-correlation criterion to find the pairing point of each pixel of the left view in the right view, and use the limit constraint to reduce the search range and the consumption of computing resources. The limit constraint is that each pair of matching pixels in the left view and the right view taken by the calibrated binocular camera is on a straight line; S4: Calculate the disparity value of each pair of pixels to obtain a disparity map; after the cost calculation in S3 is completed, a cost space is obtained, and cost aggregation is performed in the cost space. The cost aggregation is combined with global information to optimize the pixel cost value to obtain a more accurate cost value. After the cost matching is completed, each pixel selects its lowest cost value to calculate the disparity map; S5: post-processing the disparity map to obtain a smooth disparity map; S6: According to the triangulation principle, the disparity map is mapped into the three-dimensional space to form a point cloud. Then the next pair of images is processed to obtain the point cloud, which is then registered with the point cloud of the previous pair of images to obtain a fused point cloud. S7: applying pressure parallel to the tooth cross section to the side of the target tooth, the pressure not exceeding 2N, repeating steps S1-S6 to the tooth after the pressure is applied, and obtaining another point cloud; S8: Compare the two point clouds obtained in S7, calculate the displacement field according to the 3D-DIC method, further calculate the strain field, and finally analyze the strain field to obtain the position of the main crack.

2. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: In the S1, the camera rotates around the central axis of the tooth, acquires a pair of images every 5 degrees of rotation, and corrects the images.

3. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: The disparity map in S4 needs to be processed by left-right consistency check, median filtering, and hole filling after stereo matching. The left-right consistency check is to swap the left and right views and then perform stereo matching again. Pixels whose offset distances before and after the check exceed the limit are considered to be mismatched and deleted. The median filtering is used to remove noise, smooth the disparity map, and fill the holes, which is to fill the missing holes.

4. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: After the second group of point clouds are generated in S6, the two point clouds need to be registered and fused into one point cloud using a nearest neighbor iterative algorithm.

5. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: Repeat steps S1-S7 until all images are calculated. At this time, preliminary three-dimensional point cloud data of the teeth is obtained. After obtaining the point cloud image, it needs to be post-processed. The post-processing includes: filtering to remove outliers, filling holes, and reconstructing the point cloud surface to obtain a three-dimensional model. The surface reconstruction adopts the Poisson reconstruction method.

6. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: In S7, pressure is applied to both sides of the tooth, and steps S1 to S7 are repeated to obtain a three-dimensional model of the tooth after the pressure is applied.

7. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: The three-dimensional tooth models before and after stress deformation are compared, and the points with high correlation are matched through the three-dimensional digital image correlation method 3D-DIC, and the displacement and strain information of the point pairs are calculated. After completing the matching of all points in the point cloud set, the three-dimensional strain field of the entire tooth is obtained.

8. The tooth crack detection method based on 3D-DIC according to claim 1, characterized in that: Taking advantage of the fact that the crack area is more sensitive to pressure, the position with the largest strain, that is, the position where the crack is located, is found in the three-dimensional strain field. According to the above principle, the threshold is first set to eliminate the interference of small crack branches on the main crack, and then the search step size is set. After iteration, the high strain area in the three-dimensional strain field is extracted, which is the area where the crack is located.

Images