Method and system for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning

Abstract

The application relates to the technical field of three-dimensional model reconstruction, in particular to a paleontological fossil three-dimensional reconstruction method and system based on industrial CT scanning. The method comprises the following steps: collecting initial three-dimensional density image data of a paleontological fossil region by using an image sensor, determining an initial parent image based on the initial three-dimensional density image data, and optimizing the image through an iterative optimization mechanism based on the determined initial parent image; generating a three-dimensional voxel array according to the optimized image, and sequentially performing pixel spacing detection and structure connectivity verification on the three-dimensional voxel array; and constructing a paleontological three-dimensional model based on the three-dimensional voxel array that passes the two verifications. The application adjusts boundary points through density contrast, generates a stable structure recognition image through image enhancement, realizes the refinement of structure recognition, and performs real-time checking and automatic correction on the constructed three-dimensional model, so that the whole-link automatic processing from image recognition to model reconstruction of paleontological fossils is realized.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional model reconstruction technology, specifically to a method and system for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning. Background Technology

[0002] In existing 3D reconstruction techniques for paleontological fossils, most methods rely solely on conventional stacking of CT slice images for modeling, lacking sufficient optimization for image structure recognition accuracy and struggling to handle issues such as blurred boundaries between images and discontinuous slice structures. For determining the structural connectivity after voxel reconstruction, most existing methods rely on manual experience for adjustment, lacking automated and standardized verification mechanisms. Furthermore, once the 3D model is constructed using traditional methods, subsequent model corrections typically rely on manual editing, lacking real-time feedback and adjustment capabilities, resulting in low efficiency and poor controllability in the modeling process.

[0003] Therefore, there is an urgent need for a complete system with functions such as structural recognition, intelligent modeling, continuous verification, and model feedback adjustment to solve problems such as structural fracture, blurred boundaries, and high misidentification rate. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning, comprising the following steps:

[0006] Scanning and identifying paleontological fossil structures, specifically:

[0007] Initial three-dimensional density image data of the paleontological fossil area was acquired using an image sensor. The initial three-dimensional density image data was divided into N scanning slices according to the z-axis coordinate. The degree of boundary gradient change of each scanning slice was calculated and used as the structural clarity index of the image area. The image with the highest structural clarity was selected as the initial parent image from all scanning slice images. Based on the determined initial parent image, the first generation recognition image and the second generation recognition image were constructed sequentially through an iterative optimization mechanism. The paleontological fossil structure was identified based on the second generation recognition image.

[0008] A three-dimensional model is constructed based on the identified paleontological fossil structures, specifically as follows:

[0009] A three-dimensional voxel array is generated based on the optimized image, and the pixel spacing and structural connectivity of the three-dimensional voxel array are sequentially detected. A three-dimensional paleontological model is constructed based on the three-dimensional voxel array that has passed the two verifications, and the three-dimensional paleontological model is corrected in real time based on the local structural verification results of the three-dimensional model.

[0010] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the iterative optimization mechanism includes:

[0011] For high grayscale gradient regions in the parent image, multiple boundary points are randomly selected, and the density values ​​of the selected boundary points are compared with the corresponding regions in adjacent images. The region boundaries are adjusted according to the comparison results. After all boundary points have been compared, the resulting image set is the first generation of recognition images.

[0012] For the generated first-generation recognition image, edge enhancement processing is performed on the image after each implementation of the boundary region adjustment strategy in the first-generation recognition image, and the second-generation recognition image is constructed based on the result of the edge enhancement processing.

[0013] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the density value comparison is as follows:

[0014] Based on the determined high gray-level gradient boundary region, a boundary point (x′, y′) is randomly selected from the high gray-level gradient boundary region. At the same time, a boundary point (x′, y′) at the same position is selected from the adjacent image, and the gray-level difference at the same position in the adjacent image is compared. The comparison result of the gray-level difference is the density difference between the selected boundary point and the same position in the adjacent image, which is used to determine whether to execute the boundary region adjustment strategy.

[0015] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the determination of whether to execute the boundary region adjustment strategy is as follows:

[0016] Set density difference threshold θ D Based on the set density difference threshold, it is determined whether to execute the boundary region adjustment strategy.

[0017] If the calculated density difference ΔI(x′,y′) satisfies the formula |ΔI(x′,y′)|≤θ when compared with the set density difference threshold. D This indicates that the selected boundary points have not undergone structural changes in adjacent images, and no boundary region adjustment is performed;

[0018] If the calculated density difference satisfies the formula |ΔI(x′,y′)|>θ when compared with the set density difference threshold. D This indicates that the selected boundary points undergo structural changes in adjacent images, and boundary region adjustment is performed, specifically:

[0019] If the density of the corresponding location of the selected boundary point in the adjacent image is greater, the boundary expands outward; otherwise, the boundary contracts inward.

[0020] Images are collected after each boundary region adjustment. Once all boundary nodes have been compared, the first-generation recognition image is generated.

[0021] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the edge enhancement processing is specifically performed as follows:

[0022] The image after each execution of the boundary region adjustment strategy in the first generation of recognition images is set. The image has a width of W and a height of H, and is divided into a left half and a right half along the vertical midline.

[0023] Based on the divided left and right images, perform a horizontal flip, specifically as follows:

[0024] Replace the left half of the image with a mirror image of the right half, and replace the right half of the image with a mirror image of the left half.

[0025] For the horizontally flipped image, the gradients corresponding to the boundary points in the image after the region adjustment strategy is applied in the first generation recognition image are enhanced by the enhancement coefficient. The enhanced result is the density matrix corresponding to the boundary points in the second generation recognition image set. All images after the region adjustment strategy is applied in the first generation recognition image are enhanced to form the second generation recognition image.

[0026] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the identification of paleontological fossil structures based on second-generation recognition images is specifically as follows:

[0027] Based on the optimized image Construct a structural stability judgment function f i Used to identify the structure of ancient fossils, specifically:

[0028] Set a threshold f for the structural integrity of paleontological fossils. T If the stability judgment function satisfies the formula f when compared with the set threshold for the structural integrity of paleontological fossils. i ≥f T If the result is positive, it indicates that the fossil structure identified in the current image is stable; otherwise, it indicates that the fossil structure identified in the image is unstable. The initial three-dimensional density image data of the fossil area is then re-acquired, and the image is optimized until the fossil structure identified in the optimized image is complete.

[0029] In a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the pixel spacing detection is specifically as follows:

[0030] For the optimized image, the images are stacked sequentially according to the order of the scan slices of the initial parent image to generate an initial three-dimensional voxel array;

[0031] For the generated initial 3D voxel array, the stacked image is inspected. Three locations are randomly selected, and the pixel spacing between any two locations is checked to see if they are equal. If the detected pixel spacing is not equal, three locations are selected again for pixel spacing inspection. If the pixel spacing is not equal in three consecutive inspections, it means that the spacing correction of the current 3D voxel array has failed. The 3D voxel array is reconstructed until the spacing correction is passed.

[0032] If the detected pixel spacing is equal, it means that the spacing correction of the current 3D voxel array has passed and no resampling operation is required.

[0033] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the structural connectivity verification is specifically as follows:

[0034] The resampled three-dimensional voxel array is divided into M regions, and one region is randomly selected from the divided regions as the initial region for structural connectivity verification. At the same time, the adjacent regions of the initial region are located.

[0035] Determine the boundary structure of the initial region and the boundary structure of adjacent regions, calculate the pixel distance between two adjacent structures, and verify the structural connectivity based on the results. Then, we have...

[0036] If the pixel distance between two adjacent structures is 0 or the total pixel distance within the adjacent regions, it indicates that the structures between the adjacent regions in the initial region are connected; otherwise, it indicates that the structures between the two regions are not connected.

[0037] The adjacent regions are used as the second-generation initial regions. The pixel distance between the adjacent structures of the second-generation initial regions and their adjacent regions is calculated until the selected adjacent regions are the initial regions. The number of times the structures between the two regions are not connected is counted, and the structural connectivity is verified based on the number of disconnections.

[0038] As a preferred embodiment of the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning described in this invention, the structural connectivity verification based on the number of disconnections is specifically as follows:

[0039] Set a threshold N for the number of times the structure is disconnected. T If the number of disconnections counted during the verification of the 3D voxel array is N UC Compared with the set threshold for the number of times the structure is not connected, it satisfies the formula N. UC ≥N TIf the current three-dimensional voxel array structure fails verification, the three-dimensional voxel array will be reconstructed until the spacing correction is passed.

[0040] Conversely, it indicates that the structure of the current three-dimensional voxel array has been verified and a corresponding three-dimensional paleontological model has been constructed based on the current three-dimensional voxel array.

[0041] As a preferred embodiment of the three-dimensional reconstruction system for paleontological fossils based on industrial CT scanning described in this invention, the system includes: a structure recognition module, a three-dimensional model building module, and a feedback adjustment module; the structure recognition module is used to scan and recognize the structure of paleontological fossils; the three-dimensional model building module constructs a three-dimensional model of the paleontological fossils based on the recognized structures; and the feedback adjustment module is used to verify the three-dimensional model of the paleontological fossils and adjust the three-dimensional model in real time based on the verification results.

[0042] The beneficial effects of this invention are:

[0043] This invention adjusts boundary points through density contrast and generates stable structure recognition images through image enhancement, thereby achieving refined structure recognition. It also selects the "optimal recognition layer" based on the boundary gradient and generates image sequences with higher recognition accuracy through image enhancement.

[0044] By introducing a "structural connectivity verification mechanism" and "statistical threshold logic for disconnected regions", the robustness of the model is enhanced, the error of 3D modeling is reduced, and the accuracy of reconstructing fractured and broken fossils is improved.

[0045] It performs real-time verification and automatic correction of the constructed 3D model, solving the problem of tedious manual correction in the later stage of modeling, demonstrating a high degree of intelligent processing, and realizing full-link automated processing of paleontological fossils from image recognition to model reconstruction. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1 This is a schematic diagram of the overall method steps of the present invention for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0049] Example 1

[0050] Reference Figure 1 This is the first embodiment of the present invention, providing a method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning, including the following steps:

[0051] S1: Scan and identify ancient fossil structures.

[0052] Specifically, the scanning and identification of paleontological fossil structures involves acquiring image data of the paleontological fossil area and, based on the acquired image data, scanning and identifying the paleontological fossil structures. The specific implementation is as follows:

[0053] Initial three-dimensional density image data I(x,y,z) of the paleontological fossil area were acquired using an image sensor;

[0054] Based on the initial three-dimensional density image data of the area where paleontological fossils were collected, the paleontological fossil structures were scanned and identified, specifically as follows:

[0055] The initial 3D density image data is divided into N scan slices according to the z-axis coordinate. The degree of boundary gradient change in each scan slice is calculated, and the calculated boundary gradient change is used as the structural sharpness index of the image region. From all the scan slice images, the image with the highest structural sharpness is selected as the initial parent image. Then, we have...

[0056] Dividing the area into N scan slices according to the z-axis coordinate, we have:

[0057] I i (x,y),i∈{1,2,...,N}

[0058] Among them, I i (x,y) represents the two-dimensional density matrix of the i-th scan slice, where i represents the sequence number of the scan slice and N represents the total number of scan slices.

[0059] Calculate the degree of boundary gradient change for each scan slice, and use the result as an indicator of structural sharpness of the image region. Then, we have...

[0060]

[0061] Where A represents the image region Ω i The total number of pixels, Ω i This represents the selected region in the i-th scan slice used to evaluate boundary sharpness, which is set by the implementer according to the actual application scenario. Let G represent the gradient of the i-th scan slice at (x, y). i The boundary sharpness index of the i-th scan slice is used to select the initial parent image, specifically:

[0062] The boundary sharpness index of all scan slices is calculated sequentially. After the boundary sharpness index of all scan slices has been calculated, the scan slice with the largest boundary sharpness index is selected as the initial parent image.

[0063] Based on the determined initial parent image, the first generation recognition image is determined, specifically as follows:

[0064] For high grayscale gradient regions in the parent image, multiple boundary points are randomly selected, and their density values ​​are compared with corresponding regions in adjacent images. The region boundaries are adjusted based on the comparison results. Once all boundary points have been compared, the resulting image set constitutes the first-generation recognition image. Therefore,

[0065] The initial parent image is set to I. P Adjacent images are I P+1 and I P-1 And set the grayscale gradient threshold θ G Based on the set grayscale gradient threshold, high grayscale gradient regions in the initial parent image are selected, then...

[0066] If the pixels in the initial parent image satisfy the formula Let (x, y) represent the gradient of the initial parent image at (x, y). Then, pixel (x, y) represents a pixel in a high grayscale gradient region. All pixels in the initial parent image that satisfy the formula are combined to form a high grayscale gradient boundary region B. P ;

[0067] Based on the identified high grayscale gradient boundary region, a boundary point (x′, y′) is randomly selected from this region. Simultaneously, a boundary point (x′, y′) at the same location is selected from adjacent images. The grayscale difference between this point and the point at the same location in the adjacent images is then compared. Therefore,

[0068]

[0069] Among them, I P (x′, y′) represents the density matrix corresponding to the boundary points selected from the high gray-level gradient boundary region, IP+1 (x′, y′) represents the neighboring image I P+1 The density matrix corresponding to the boundary point at the same position as the selected boundary point in the high gray-level gradient boundary region, I P-1 (x′, y′) represents the neighboring image I P-1 The density matrix corresponding to the boundary point at the same position as the selected boundary point in the high grayscale gradient boundary region, ΔI(x′,y′) represents the density difference between the selected boundary point and the same position in the adjacent image, and is used to determine whether to execute the boundary region adjustment strategy, specifically:

[0070] Set density difference threshold θ D Based on the set density difference threshold, it is determined whether to execute the boundary region adjustment strategy.

[0071] If the calculated density difference satisfies the formula |ΔI(x′,y′)|≤θ when compared with the set density difference threshold. D This indicates that the selected boundary points have not undergone structural changes in adjacent images, and no boundary region adjustment is performed;

[0072] If the calculated density difference satisfies the formula |ΔI(x′,y′)|>θ when compared with the set density difference threshold. D This indicates that the selected boundary points undergo structural changes in adjacent images, and boundary region adjustment is performed, specifically:

[0073] If the density of the corresponding location of the selected boundary point in the adjacent image is greater, the boundary expands outward; otherwise, the boundary contracts inward.

[0074] Images are acquired after each boundary region adjustment. Once all boundary nodes have been compared, the first-generation recognition image is generated.

[0075]

[0076] Where k represents the sequence number of the boundary region adjustment strategy executed, and K represents the total number of times the boundary region adjustment strategy was executed. I represents the image after the k-th execution of the boundary region adjustment strategy. (1) The first generation of recognition images generated.

[0077] For the generated first-generation recognition image, edge enhancement processing is performed on the image after each boundary region adjustment strategy is applied. A second-generation recognition image is then constructed based on the edge enhancement results. Simultaneously, by iterating between the first and second-generation recognition images, the recognition image is optimized, thereby achieving the identification of paleontological fossil structures. The specific implementation is as follows:

[0078] The image after each execution of the boundary region adjustment strategy in the first generation of recognition images is set. The image has a width of W and a height of H. Dividing the image along its vertical midline, we have:

[0079] In the first-generation image recognition, the pixels in the image after each implementation of the boundary region adjustment strategy satisfy the following formula.

[0080] Left half of the image The pixels in the left half of the image satisfy the formula x′∈[0,W / 2), and the pixels in the right half of the image... The pixels in the right half of the image satisfy the formula x′∈[W / 2,W], where W represents the width of the image after each execution of the boundary region adjustment strategy in the first generation of recognition images;

[0081] Based on the divided left and right images, perform a horizontal flip, specifically as follows:

[0082] If we replace the left half of the image with a mirror image of the right half, and vice versa, then we have:

[0083]

[0084] in, Let represent the density matrix corresponding to the boundary points in the image after the k-th execution of the boundary region adjustment strategy in the first-generation recognition image, and W represent the width of the image after each execution of the boundary region adjustment strategy in the first-generation recognition image. This represents the density matrix corresponding to the boundary points in the left half of the image after each implementation of the boundary region adjustment strategy in the first-generation recognition image. This represents the density matrix corresponding to the boundary points in the right half of the image after each execution of the boundary region adjustment strategy in the first-generation recognition image.

[0085] The image after horizontal flipping is then subjected to edge enhancement processing, specifically as follows:

[0086]

[0087] in, This represents the density matrix of the boundary points in the image after the k-th execution of the boundary region adjustment strategy in the first-generation recognition image. This represents the gradient corresponding to the boundary points in the image after the k-th execution of the boundary region adjustment strategy in the first-generation recognition image. This represents the density matrix corresponding to the boundary points in the second-generation recognition image set, and α represents the enhancement coefficient, which is set by the implementer according to the actual application scenario.

[0088] It should be noted that the second-generation recognition image is constructed by performing edge enhancement processing on all the first-generation recognition images, thus, we have,

[0089]

[0090] Where k represents the sequence number of the boundary region adjustment strategy executed, and K represents the total number of times the boundary region adjustment strategy was executed. I represents the result of edge enhancement processing on the image after the k-th execution of the boundary region adjustment strategy. (2) The generated second-generation recognition image.

[0091] For the generated second-generation recognition image, it is iterated with the first-generation recognition image to optimize the image and thus achieve the identification of paleontological fossil structures. Specifically:

[0092] The iteration condition for the first generation of recognition images is: the average boundary continuity (edge ​​closure index) of the first generation of recognition images is greater than that of the initial parent image;

[0093] The iteration condition for the second-generation recognition image is: the density difference of any image in the second-generation recognition image is better than that of the first-generation recognition image;

[0094] The iteration stopping condition is: if all first-generation recognition images satisfy the iteration conditions for second-generation recognition images, then the second-generation recognition images generated iteratively from the current first-generation recognition images are taken as the optimized images. Based on the optimized images, paleontological fossil structures are identified, specifically:

[0095] Based on the optimized image The boundary sharpness index, edge closure index, and image registration residual are calculated for each image. Based on the calculation results, a structural stability judgment function is constructed, specifically as follows:

[0096] f i =w1·G i +w2·E i +w3·R i

[0097] Where w1, w2, and w3 represent weighting coefficients, which are set by the implementers according to the actual application scenario. G i E represents the edge sharpness index of the i-th image. i R represents the edge closure index of the i-th image. i f represents the registration residual of the i-th image, where i represents the image index sequence number. i The structural stability judgment function is constructed and used to identify the structure of ancient fossils, specifically as follows:

[0098] Set a threshold f for the structural integrity of paleontological fossils. T If the stability judgment function satisfies the formula f when compared with the set threshold for the structural integrity of paleontological fossils. i ≥f T If the result is positive, it indicates that the fossil structure identified in the current image is stable; otherwise, it indicates that the fossil structure identified in the image is unstable. The initial three-dimensional density image data of the fossil area is then re-acquired, and the image is optimized until the fossil structure identified in the optimized image is complete.

[0099] It should be noted that the edge closure index E in the stability judgment function... i The calculation is used to evaluate whether the edge contours in the image form closed regions, thereby verifying the integrity of the fossil contours. Specifically:

[0100] This involves edge detection processing of an image, often using the Canny operator to extract contour edges, generating a binarized edge image, and then performing connected component labeling on the extracted edge image. All contour regions are numbered and recorded. For each contour region, it is determined whether it is a closed shape. Then, we have...

[0101] If the outline is closed at both ends and not connected to the edge of the image, it is a closed figure;

[0102] The ratio of the number of closed contour regions in an image to the total number of contour regions is calculated, and the result is the edge closure index.

[0103] For the registration residual R in the stability judgment function i The calculation is used to measure the alignment between the current image and its neighboring images, as follows:

[0104] Select the current image and its neighboring images, and calculate the pixel mean square error between the neighboring images and the current image respectively. The average of the pixel mean square errors between the two neighboring images and the current image is the registration residual of the current image.

[0105] S2: Constructing a 3D model based on scanned paleontological fossil structures.

[0106] Specifically, the three-dimensional model constructed based on scanned paleontological fossil structures is built from optimized images, and the constructed three-dimensional model is corrected in real time. The specific implementation is as follows:

[0107] For the optimized image, stack the images sequentially according to the order of the scan slices of the initial parent image to generate an initial three-dimensional voxel array, then we have,

[0108]

[0109] Where V0(x,y,z) represents the generated initial three-dimensional voxel array, and z represents the stacking dimension. This represents the gray value of the z-th layer image at position (x, y).

[0110] Furthermore, for the generated initial 3D voxel array, 3D resampling is used to ensure consistent distances and uniform voxels between images. The structural connectivity of the resampled images is then verified, as follows:

[0111] 3D resampling involves detecting the pixel spacing of a 3D voxel array and performing 3D resampling based on the detection results. Specifically:

[0112] For the generated initial 3D voxel array, the stacked image is inspected. Three locations are randomly selected, and the pixel spacing between any two locations is checked to see if they are equal. If the detected pixel spacing is not equal, three locations are selected again for pixel spacing inspection. If the pixel spacing is not equal in three consecutive inspections, it means that the spacing correction of the current 3D voxel array has failed. The 3D voxel array is reconstructed until the spacing correction is passed.

[0113] If the detected pixel spacing is equal, it means that the spacing correction of the current 3D voxel array has passed and no resampling operation is required.

[0114] Furthermore, structural connectivity verification is performed on the resampled 3D voxel array by detecting 3D connected regions. Specifically:

[0115] The resampled three-dimensional voxel array is divided into M regions, and one region is randomly selected from the divided regions as the initial region for structural connectivity verification. At the same time, the adjacent regions of the initial region are located.

[0116] Determine the boundary structure of the initial region and the boundary structure of adjacent regions, calculate the pixel distance between two adjacent structures, and verify the structural connectivity based on the results. Then, we have...

[0117] If the pixel distance between two adjacent structures is 0 or the total pixel distance within the adjacent regions, it indicates that the structures between the adjacent regions in the initial region are connected; otherwise, it indicates that the structures between the two regions are not connected.

[0118] Using adjacent regions as the second-generation initial regions, the pixel distance between adjacent structures of the second-generation initial regions and their adjacent regions is calculated until a selected adjacent region becomes the initial region. The number of times the structures between two regions are not connected is counted, and the result is based on the number of disconnections N. UC If we perform structural connectivity verification, then we have:

[0119] Set a threshold N for the number of times the structure is disconnected. T If the number of disconnections counted during the verification of the 3D voxel array satisfies the formula N when compared with the set threshold for the number of structural disconnections. UC ≥N T If the current three-dimensional voxel array structure fails verification, the three-dimensional voxel array will be reconstructed until the spacing correction is passed.

[0120] Conversely, it indicates that the structure of the current three-dimensional voxel array has been verified and a corresponding three-dimensional paleontological model has been constructed based on the current three-dimensional voxel array.

[0121] Furthermore, real-time correction of the constructed 3D model involves performing local area verification on the constructed paleontological 3D model and making real-time corrections based on the verification results of the local structures. The specific implementation is as follows:

[0122] For the constructed 3D paleontological model, it is divided into m structural units. The boundary sharpness index and density consistency index of each structural unit are calculated sequentially. Based on the calculation results, the 3D paleontological model is corrected in real time. Specifically:

[0123] At the same time, threshold values ​​for boundary clarity and density consistency are set for each structural unit;

[0124] By comparing the calculation results with the set indicator thresholds, it is determined whether to make real-time corrections. Specifically:

[0125] If either the calculated boundary sharpness index or the density consistency index exceeds the set threshold, the current paleontological 3D model needs real-time correction, including boundary sharpness correction and density consistency correction. Conversely, if either exceeds the threshold, the current paleontological 3D model does not need real-time correction.

[0126] Furthermore, the boundary clarity correction involves locating the abnormal boundary gradient anomaly points based on the structural units with clear boundaries, deleting the anomaly points, and replacing the deleted anomaly points with adjacent normal points to ensure the integrity of the paleontological 3D model. This process continues until all gradient anomaly points have been deleted, indicating that the boundary clarity correction of the current paleontological 3D model is complete.

[0127] Density consistency correction is performed by calculating the median gray value of all voxels within the anomalous structural units based on the structural units with uneven density. Then, the voxels whose gray values ​​deviate from the calculated median gray value are corrected until the gray values ​​of all voxels are consistent with the calculated median gray value. This process is repeated until the density consistency correction of the current paleontological 3D model is complete.

[0128] Example 2

[0129] A second embodiment of the present invention provides a three-dimensional reconstruction system for paleontological fossils based on industrial CT scanning, comprising:

[0130] Specifically, the structure recognition module is used to scan and identify the structures of ancient fossils;

[0131] The 3D model building module constructs 3D models of paleontological fossils based on the identified fossil structures.

[0132] The feedback adjustment module is used to verify the 3D model of the paleontological fossil and adjust the establishment of the 3D model based on the verification results.

[0133] Furthermore, the structure recognition module acquires CT scan images, extracts the boundary gradient information of each slice, automatically selects the slice image with the best structural clarity, and generates a stable recognition image sequence based on boundary density differences and image enhancement strategies. The validity of the recognition image is determined by the structural integrity judgment function, and finally outputs a stable structural region for 3D model construction, providing accurate structural basis data for the 3D model building module. Image clarity and high structural stability directly affect model accuracy and construction efficiency.

[0134] The 3D model building module stacks the recognized images sequentially to construct a 3D voxel array and performs pixel spacing detection on the voxel array to ensure spatial consistency. At the same time, it executes structural connectivity verification logic to eliminate disconnected or unstable regions. Finally, it generates a preliminary 3D fossil model based on the verified voxel array. The model built by the 3D model building module will be verified and judged by the feedback module to determine the structural stability and closure. The feedback module provides a verification and callback mechanism to drive this module to complete the model update.

[0135] The feedback adjustment module performs local integrity judgment based on the internal structural regions of the model, identifies structural anomalies such as geometric distortion, missing boundaries, and abnormal density distribution in the model, and then triggers a local region correction mechanism to reconstruct fractured segments or refill void areas in real time. The correction results are fed back to the 3D model building module to update the model version, forming a closed-loop optimization. The feedback adjustment module realizes dynamic closed-loop control from structural identification to modeling, solves unpredictable structural problems during model construction, and forms automatic optimization capabilities.

[0136] Furthermore, if the aforementioned function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0137] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0138] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0139] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning, characterized by: Includes the following steps, Scanning and identifying paleontological fossil structures, specifically: Initial three-dimensional density image data of the paleontological fossil area was acquired using an image sensor. The initial three-dimensional density image data was divided into N scanning slices according to the z-axis coordinate. The degree of boundary gradient change of each scanning slice was calculated and used as the structural clarity index of the image area. The image with the highest structural clarity was selected as the initial parent image from all scanning slice images. Based on the determined initial parent image, the first generation recognition image and the second generation recognition image were constructed sequentially through an iterative optimization mechanism. The paleontological fossil structure was identified based on the second generation recognition image. A three-dimensional model is constructed based on the identified paleontological fossil structures, specifically as follows: A three-dimensional voxel array is generated based on the optimized image, and the pixel spacing and structural connectivity of the three-dimensional voxel array are sequentially detected. A three-dimensional paleontological model is constructed based on the three-dimensional voxel array that has passed the two verifications, and the three-dimensional paleontological model is corrected in real time based on the local structural verification results of the three-dimensional model. Iterative optimization mechanisms include, For high grayscale gradient regions in the parent image, multiple boundary points are randomly selected, and the density values ​​of the selected boundary points are compared with the corresponding regions in adjacent images. The region boundaries are adjusted according to the comparison results. After all boundary points have been compared, the resulting image set is the first generation of recognition images. For the generated first-generation recognition image, edge enhancement processing is performed on the image after each implementation of the boundary region adjustment strategy in the first-generation recognition image, and the second-generation recognition image is constructed based on the result of the edge enhancement processing. The structural connectivity verification is as follows: The resampled three-dimensional voxel array is divided into M regions, and one region is randomly selected from the divided regions as the initial region for structural connectivity verification. At the same time, the adjacent regions of the initial region are located. Determine the boundary structure of the initial region and the boundary structure of adjacent regions, calculate the pixel distance between two adjacent structures, and verify the structural connectivity based on the results. Then, we have... If the pixel distance between two adjacent structures is 0 or the total pixel distance within the adjacent regions, it indicates that the structures between the adjacent regions in the initial region are connected; otherwise, it indicates that the structures between the two regions are not connected. The adjacent regions are used as the second-generation initial regions. The pixel distance between the adjacent structures of the second-generation initial regions and their adjacent regions is calculated until the selected adjacent regions are used as the initial regions. The number of times the structures between the two regions are not connected is counted, and the structural connectivity is verified based on the number of times they are not connected.

2. The method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning according to claim 1, characterized in that, The density value comparison is as follows: Based on the determined high gray-level gradient boundary region, a boundary point (x′, y′) is randomly selected from the high gray-level gradient boundary region. At the same time, a boundary point (x′, y′) at the same position is selected from the adjacent image, and the gray-level difference at the same position in the adjacent image is compared. The comparison result of the gray-level difference is the density difference between the selected boundary point and the same position in the adjacent image, which is used to determine whether to execute the boundary region adjustment strategy.

3. The method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning according to claim 2, characterized in that, The specific steps for determining whether to execute the boundary region adjustment strategy are as follows: Set density difference threshold θ D Based on the set density difference threshold, it is determined whether to execute the boundary region adjustment strategy. If the calculated density difference ΔI(x′,y′) and the set density difference threshold satisfy the formula |ΔI(x′,y′)|≤θ D This indicates that the selected boundary points have not undergone structural changes in adjacent images, and no boundary region adjustment is performed; If the calculated density difference satisfies the formula |ΔI(x′,y′)|>θ when compared with the set density difference threshold. D This indicates that the selected boundary points undergo structural changes in adjacent images, and boundary region adjustment is performed, specifically: If the density of the corresponding location of the selected boundary point in the adjacent image is greater, the boundary expands outward; otherwise, the boundary contracts inward. Images are collected after each boundary region adjustment. Once all boundary nodes have been compared, the first-generation recognition image is generated.

4. The method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning according to claim 3, characterized in that, The edge enhancement processing is performed as follows: The image after each execution of the boundary region adjustment strategy in the first generation of recognition images is set. The image has a width of W and a height of H, and is divided into a left half and a right half along the vertical midline. Based on the divided left and right images, perform a horizontal flip, specifically as follows: Replace the left half of the image with a mirror image of the right half, and replace the right half of the image with a mirror image of the left half. For the horizontally flipped image, the gradients corresponding to the boundary points in the image after the region adjustment strategy is applied in the first generation recognition image are enhanced by the enhancement coefficient. The enhanced result is the density matrix corresponding to the boundary points in the second generation recognition image set. All images after the region adjustment strategy is applied in the first generation recognition image are enhanced to form the second generation recognition image.

5. The method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning according to claim 4, characterized in that, The specific steps for identifying paleontological fossil structures based on second-generation recognition images are as follows: Based on the optimized image Construct a structural stability judgment function f i Used to identify the structure of ancient fossils, specifically: Set a threshold f for the structural integrity of paleontological fossils. T If the stability judgment function satisfies the formula f when compared with the set threshold for the structural integrity of paleontological fossils. i ≥f T If the result is positive, it indicates that the fossil structure identified in the current image is stable; otherwise, it indicates that the fossil structure identified in the image is unstable. The initial three-dimensional density image data of the fossil area is then re-acquired, and the image is optimized until the fossil structure identified in the optimized image is complete.

6. The method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning according to claim 5, characterized in that, The pixel spacing detection is as follows: For the optimized image, the images are stacked sequentially according to the order of the scan slices of the initial parent image to generate an initial three-dimensional voxel array; For the generated initial 3D voxel array, the stacked image is inspected. Three locations are randomly selected, and the pixel spacing between any two locations is checked to see if they are equal. If the detected pixel spacing is not equal, three locations are selected again for pixel spacing inspection. If the pixel spacing is not equal in three consecutive inspections, it means that the spacing correction of the current 3D voxel array has failed. The 3D voxel array is reconstructed until the spacing correction is passed. If the detected pixel spacing is equal, it means that the spacing correction of the current 3D voxel array has passed and no resampling operation is required.

7. The method for three-dimensional reconstruction of paleontological fossils based on industrial CT scanning according to claim 6, characterized in that, The structural connectivity verification based on the number of disconnections is as follows: Set a threshold N for the number of times the structure is disconnected. T If the number of disconnections counted during the verification of the 3D voxel array is N UC Compared with the set threshold for the number of times the structure is not connected, it satisfies the formula N. UC ≥N T If the current three-dimensional voxel array structure fails verification, the three-dimensional voxel array will be reconstructed until the spacing correction is passed. Conversely, it indicates that the structure of the current three-dimensional voxel array has been verified and a corresponding three-dimensional paleontological model has been constructed based on the current three-dimensional voxel array.

8. A three-dimensional reconstruction system for paleontological fossils based on industrial CT scanning, applied to the three-dimensional reconstruction method for paleontological fossils based on industrial CT scanning as described in any one of claims 1 to 7, characterized in that, It includes a structure recognition module, a 3D model building module, and a feedback adjustment module; The structure recognition module is used to scan and identify the structure of ancient fossils; The three-dimensional model building module constructs a three-dimensional model of the paleontological fossil based on the identified paleontological fossil structure. The feedback adjustment module is used to verify the three-dimensional model of the paleontological fossil and to correct and adjust the three-dimensional model in real time based on the verification results.

Images