A point cloud-based method for establishing a three-dimensional model of an immovable cultural relic

Abstract

This invention discloses a method for establishing a 3D model of immovable cultural relics based on point clouds, belonging to the fields of surveying and mapping geographic information and cultural relic protection technology. The method includes the following steps: Step 1, Data Acquisition and Preliminary Modeling: A 3D point cloud model of the cultural relic is obtained using both laser scanning and close-range photogrammetry. Step 2, Point Cloud Model Registration: One of the two 3D point cloud models obtained in Step 1 is kept in its original position while the other is moved and rotated until the two models are essentially overlapped. Finally, iterative calculations are performed on the points of the moved 3D model to ensure that the average distance between corresponding points of the two models is less than a given threshold. Step 3, Cultural Relic Model Repair: For the 3D model obtained in Step 2, the model with higher point cloud density is fully retained, while only the non-overlapping parts of the model with lower point cloud density are retained, resulting in the final 3D model of the cultural relic. The 3D model obtained by this invention not only has high mathematical accuracy but also preserves the texture details of the cultural relic.

Description

Technical Field

[0001] This invention patent belongs to the fields of surveying and mapping geographic information and cultural relic protection technology, specifically involving a method for establishing a three-dimensional model of immovable cultural relics based on point clouds. Background Technology

[0002] Historical and cultural heritage is an irreplaceable and invaluable resource. "Strengthening the protection and utilization of cultural relics and the preservation and inheritance of cultural heritage" is an important part of building cultural confidence. Through the continuous efforts of several generations in China, the protection of cultural relics has been significantly strengthened, and the preservation status of cultural relics has continued to improve. There are 767,000 immovable cultural relics, 108 million pieces / sets of state-owned museum collections, and a vast number of privately collected cultural relics that have traversed historical time and space, spread across the vast land, and enrich people's lives.

[0003] In recent years, the use of oblique photogrammetry and 3D laser scanning technologies to create realistic 3D models has gradually become mainstream in the geographic information industry. Aerial and terrestrial photography can acquire 3D point clouds and build models; terrestrial laser scanning can acquire large-area 3D point cloud scenes and perform high-precision scanning of individual targets to build models; simultaneously, handheld laser scanners can acquire high-precision 3D point clouds of smaller targets. Therefore, using photogrammetry and laser scanning technologies to acquire 3D scene point clouds and images of cultural relics and historical sites, and to realistically reconstruct large-scale immovable cultural relics and build centimeter-level or even millimeter-level 3D models, is an effective technical method to meet the purpose of cultural relic protection.

[0004] However, both photogrammetric modeling and 3D laser scanning modeling are limited by the occlusion of ground features and the inherent characteristics of the technologies themselves. A single method may suffer from scanning gaps and texture loss, making it difficult to achieve satisfactory results. For example, 3D laser scanning boasts very high mathematical accuracy, but the laser points are discrete, making it difficult to acquire texture information effectively; while photogrammetric modeling can acquire texture information well, but its mathematical accuracy is relatively lower. Summary of the Invention

[0005] The purpose of this invention is to solve the problems mentioned in the background art and provide a method for establishing a three-dimensional model of immovable cultural relics based on point cloud, which can not only obtain a model with high mathematical accuracy, but also make the three-dimensional model have the texture details of cultural relics.

[0006] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for establishing a 3D model of an immovable cultural relic based on point clouds, characterized by the following steps:

[0008] Step 1: Data Collection and Preliminary Modeling

[0009] The first three-dimensional point cloud model of the cultural relic was obtained by laser scanning and software processing; the second three-dimensional point cloud model of the cultural relic was obtained by close-range photogrammetry and software processing.

[0010] Step 2: Point cloud model registration:

[0011] Using the first 3D point cloud model as the reference model and the second 3D point cloud model as the registration model, while keeping the position of the reference model unchanged, the registration model is moved so that its center point coincides with the center point of the reference model; then the registration model is rotated around the center point so that the two models basically coincide; finally, the 3D model points of the registration model are iteratively calculated so that the average distance between the 3D model points of the registration model and the corresponding points of the 3D model of the reference model is less than a given threshold.

[0012] Step 3: Repairing the artifact model:

[0013] For the 3D model obtained in step two, determine the overlapping and non-overlapping parts of the two models. Then, retain the model with higher point cloud density, discard the overlapping parts of the model with lower point cloud density, and retain the non-overlapping parts of the model with lower point cloud density. The combination of all retained parts of the model is the final 3D model of the cultural relic.

[0014] As a preferred method, the first three-dimensional point cloud model acquisition method described in step one is as follows: set up measuring stations around the cultural relic and deploy spherical connecting targets, scan the cultural relic with a laser to directly obtain the laser point cloud of each measuring station, and the overlap of the point cloud coverage of adjacent measuring stations is required to be more than 50%; then, using the supporting software, the point clouds of each scanning measuring station are stitched together using spherical targets to construct the first three-dimensional point cloud model of laser scanning.

[0015] As a preferred method, the second three-dimensional point cloud model acquisition method described in step one is as follows: take photos of the cultural relic with a digital camera, requiring the photos to fully cover the cultural relic and adjacent photos to have an overlap of more than 65%; then perform matching calculations in the supporting software to construct a three-dimensional patch model, extract the vertices of the three-dimensional patch model, and thus construct the second three-dimensional point cloud model of close-range photogrammetry.

[0016] As a preferred method, the model center point determination method in step two involves: measuring the target coordinates using a surveying instrument, then importing the global coordinates into the laser point cloud to obtain the point cloud coordinates of the 3D model, and then using the formula...

[0017]

[0018]

[0019]

[0020] Calculate the center coordinates of the two models;

[0021] Where X, Y, and Z are the three coordinate components of the model's center coordinate value, X i ,Y i Z i The three-dimensional coordinates of the i-th point in the model.

[0022] As a preferred option, the iterative calculation method in step two is as follows:

[0023] Let the 3D point set of the baseline model be U1, and the 3D point set of the registration model be U2. The specific calculation steps are as follows:

[0024] a. Calculate the nearest point in the U1 point set for each point in U2;

[0025] b. Find the rigid body transformation that minimizes the average distance between the corresponding points mentioned above, and find the translation and rotation parameters.

[0026] c. Apply the translation and rotation parameters obtained in the previous step to U2 to obtain a new set of transformation points U3;

[0027] d. If the average distance between the two point sets U3 and U1 is less than a given threshold, then stop the iterative calculation; otherwise, repeat steps a to c with U3 as the new U2 until the latest transformed point set U3 and U1 satisfy the condition that the average distance between the two point sets is less than the given threshold.

[0028] Preferably, the method for determining overlapping and non-overlapping parts in step three includes the following steps:

[0029] Step (1) Obtain the normal vector of the point cloud of the reference model. Use the reference model or registration model as the reference model and the other model as the non-reference model; take the centroid of the reference model as the starting point, and define the normal vector of the point cloud as pointing outwards as the positive direction, and calculate the normal vector of each point in the reference model;

[0030] Step (2) Starting with the points of the reference model, find the points of the non-reference model along the positive and negative directions of the normal to obtain the overlapping area. Before obtaining the overlapping area, mark the category field value of all points of the reference model and the non-reference model as "1".

[0031] The specific method for determining the overlapping region is as follows:

[0032] First, construct a rotating cylinder with the normal as the axis of rotation and a specified distance value a as the radius;

[0033] Next, find the points that fall inside the cylinder; if a point is found, save it as a candidate point dataset, and consider the reference model and non-reference model to be overlapping parts at this point;

[0034] Finally, calculate the distance from each point in the candidate point dataset to the normal, mark the point with the smallest distance, and change its category field value to "6"; at the same time, change the category field value of the reference model point corresponding to this candidate point dataset to "2".

[0035] In step (2), if no point is found, it is considered to be a non-overlapping area; at the same time, the category field value of this reference model point is modified to "3".

[0036] After the above process, the category field value of all points in the reference model becomes "2" or "3", and the category field value of all points in the non-reference model becomes "1" or "6". Point clouds with category field values ​​of "2" and "6" are overlapping areas, and those with category field values ​​of "1" or "3" are non-overlapping areas.

[0037] Preferably, the specified distance value a in step (2) is twice the average point spacing between the reference model and the registration model.

[0038] The beneficial effects of this invention are:

[0039] 1. By combining and filtering the 3D point cloud models obtained by laser scanning and close-range photogrammetry, the 3D stationery model composed of the retained 3D point cloud not only has high mathematical accuracy, but also has texture details, and the modeling effect is far superior to the existing technology.

[0040] 2. Iterative calculation is used to make the matching degree of the two models precise and controllable. The method is simple, improves efficiency, and ensures high accuracy in matching the two models.

[0041] 3. The method for determining overlapping and non-overlapping areas of the model intuitively distinguishes the point cloud of non-overlapping areas, thus providing a basis for the selection and combination of two 3D models. This enables the repair of missing details in the 3D model of cultural relics, resulting in higher accuracy and texture details in the 3D model of cultural relics. Attached Figure Description

[0042] Figure 1 This is a flowchart of the present invention;

[0043] Figure 2A This is a schematic diagram of the stone dragon head, an immovable cultural relic.

[0044] Figure 2B A 3D point cloud model of the stone dragon head constructed using the laser scanning method;

[0045] Figure 2C A 3D point cloud model of the stone dragon head constructed using close-range photogrammetry;

[0046] Figure 2DA partial enlarged view of the 3D point cloud model of the stone chi head constructed by the laser scanning method;

[0047] Figure 3A A 3D point cloud diagram of overlapping two models after the registration model is rotated and aligned around the center point;

[0048] Figure 3B A 3D point cloud model diagram of the stone chi head after iterative calculation;

[0049] Figure 4A An elevation rendering diagram of the repaired 3D point cloud model;

[0050] Figure 4B A local enlarged view before and after model repairing. DETAILED DESCRIPTION

[0051] Embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0052] It should be noted that terms such as "upper", "lower", "left", "right", "front" and "rear" cited in the invention are only used for clarity of description, and are not intended to limit the scope of the invention. The change or adjustment of their relative relationship shall also be regarded as within the implementable scope of the invention without substantial change to the technical content.

[0053] As Figure 1 shown, the 3D model construction method for immovable cultural relics based on point cloud provided by the present invention comprises three steps: data acquisition and modeling, point cloud model registration, and cultural relic model repairing. The following "3D modeling of stone chi head" is taken as an embodiment for specific description:

[0054] The first step, data acquisition and modeling; the first step comprises the following sub-steps:

[0055] (1) Data acquisition. The P40 laser scanner is adopted to scan Figure 2A the stone chi head cultural relic shown, set up survey stations around the stone chi head and arrange spherical connection targets, directly obtain the laser point cloud of each survey station, and a total of 745355 laser points are obtained; a high-resolution digital camera is used to take photos of the stone chi head cultural relic.

[0056] (2) Data inspection. Check whether the coverage overlap of point clouds between adjacent survey stations of laser scanning is greater than 50%; check whether the coverage of the target by adjacent photos meets the requirement of more than 65%. If the overlap of survey station point clouds or adjacent photos is insufficient, it is required to go back to the site for supplementary data acquisition.

[0057] (3) Construction of cultural relic point cloud model. Using the software matched with the laser hardware equipment, connecting by spherical targets, splicing the point clouds of each scanning survey station, and constructing a 3D point cloud model by laser scanning, as shown in Figure 2BAs shown. Close-up photographs were used in software for matching calculations to construct a 3D patch model. The vertices of this 3D patch model were extracted to create a 3D point cloud model based on close-up photogrammetry. A total of 544,360 image matching points were obtained, as shown below. Figure 2C As shown. From Figure 2A and Figure 2B , Figure 2C In comparison, it can be found that the point cloud model obtained by the laser scanning method can reflect the three-dimensional outline and detailed convex and concave features of the stone dragon head, but the model has some missing parts, such as... Figure 2D As indicated by the middle arrow.

[0058] The second step is point cloud model registration;

[0059] The coordinates of the spherical target are measured using precision surveying instruments with an accuracy of 0.1 mm. The global coordinates are then imported into a 3D point cloud model. Before registering the point cloud model, the "roles" of the models acquired through different methods must be determined. The point cloud model obtained through laser scanning serves as the "reference model," while the point cloud model obtained through image modeling serves as the "registration model." The second step involves registering the point cloud models constructed using the two acquisition methods, and includes the following sub-steps:

[0060] (1) Coarse registration based on center coordinate alignment. The center coordinates of the two models are calculated using the following formula:

[0061]

[0062]

[0063]

[0064] Where X, Y, and Z are the three coordinate components of the model's center coordinate value, X i ,Y i Z i The model corresponds to the 3D coordinates of the i-th point. The center coordinates of the reference model remain unchanged, so that the center coordinates of the registration model are consistent with those of the reference model. This is called moving the registration model, and the amount of movement is the difference between the center coordinates of the two models.

[0065] (2) Manually rotate the models to roughly align the two models. Keep the reference model in place and rotate the registration model around its center point. Determine the rotation angle based on the overlap of the point clouds of the two models, ultimately ensuring the two models are essentially aligned. For example... Figure 3A As shown, the dark color represents the point cloud of the registration model, and the light color represents the point cloud of the reference model.

[0066] (3) Use ICP, or Iterative Closest Point Method, for fine registration. The reference model remains stationary, and registration calculations are performed on the registration model using the ICP method. Note that the ICP method is implemented by a computer program and does not require manual calculation. Let the 3D point set of the reference model be U1, and the 3D point set of the registration model be U2. The registration steps of the ICP method are as follows:

[0067] a. Calculate the nearest point in the U1 point set for each point in U2;

[0068] b. Find the rigid body transformation that minimizes the average distance between the corresponding points mentioned above, and find the translation and rotation parameters.

[0069] c. Apply the translation and rotation parameters obtained in the previous step to U2 to obtain a new set of transformation points U3;

[0070] d. If the average distance between the two point sets U3 and U1 is less than a given threshold, then stop the iterative calculation; otherwise, repeat steps a to c with U3 as the new U2 until the latest transformed point set U3 and U1 satisfy the condition that the average distance between the two point sets is less than the given threshold. The finely registered 3D point cloud model of the stone dragon head is shown below. Figure 3B As shown, the dark color represents the point cloud of the registration model, and the light color represents the point cloud of the reference model.

[0071] The third step is the restoration of the artifact model. After registering the point cloud models obtained from the two methods, the superior data blocks are selected to construct a new point cloud model. That is, the model with high point cloud density is used as the base data, and data obtained from other methods are used to repair the defects in this base data, mainly repairing areas lacking detail. The third step includes the following sub-steps (note that this step is also implemented by a computer program and does not require manual calculation):

[0072] (1) Obtain the normal vector of the point cloud of the reference model. Use the reference model or registration model as the reference model, and the other model as the non-reference model. In this real-time example, the reference model is used as the reference model and the registration model is used as the non-reference model. The centroid of the reference model is used as the starting point, and the normal vector of the point cloud is defined as pointing outward from the point cloud. Calculate the normal vector of each point in the reference model.

[0073] (2) Starting with the points of the reference model, find the points of the non-reference model along the positive and negative directions of the normal to obtain the overlapping region. Before obtaining the overlapping region, mark the category field value of all points of the reference model and the non-reference model as "1".

[0074] The specific method for determining the overlapping region is as follows:

[0075] First, construct a rotating cylinder with the normal as the axis of rotation and a radius of 5mm (usually 1 times the average point spacing of the model);

[0076] Next, find points that fall within the cylinder. If a point is found, save it as a candidate point dataset, and consider the two models to overlap at that point.

[0077] Finally, calculate the distance from each point in the candidate point dataset to the normal, mark the point with the smallest distance, and change the category field value of this point to "6"; at the same time, change the category field value of the reference model point corresponding to this candidate point dataset to "2".

[0078] (3) In step (2) above, if no points are found, it is considered a non-overlapping region. The category field value of the corresponding reference model point is then changed to "3". At this point, the points of the two models can be determined to overlap based on their category field values. There are three possible scenarios, as detailed in the table below:

[0079] one The category field value is 2 The category field value is 6 Overlapping area two The category field value is 3 - Non-overlapping areas, only referencing the model slightly three - The category field value is 1 Non-overlapping regions, only the non-reference model has some differences.

[0080] (4) Clarify the selection principles and complete model repair. Clarify the principle of retaining high-density model point clouds and make comprehensive selections of model point clouds.

[0081] If the point cloud density of the non-reference model is high, based on the results of the previous step (3), retain all the point clouds of the non-reference model and the point clouds of the reference model with a category value of 3, while discarding the point clouds of the reference model with a category value of 2. Note that after step (2), the category field values ​​of all points in the reference model have been modified to 2 or 3. In this way, a new cultural relic point cloud model is constructed, and the repair of the cultural relic point cloud model is completed.

[0082] The repaired point cloud model is as follows Figure 4A As shown, the total number of point clouds before merging was 1,289,715, and the number of points in the model after merging was 987,713. A zoom-in view is shown below. Figure 4B As shown, the upper half is before model repair, with white and gray dots representing local points of the non-reference model and the reference model, respectively. The lower half is after model repair, with all points representing local points after the repair and fusion of the non-reference model and the reference model.

[0083] If the point cloud density of the reference model is high, based on the results of the previous step (3), retain all the point clouds of the reference model and the point clouds with a class value of 1 in the non-reference model, while discarding the point clouds with a class value of 6 in the non-reference model. In this way, a new point cloud model of cultural relics is constructed, thus completing the repair of the point cloud model of cultural relics.

[0084] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for establishing a 3D model of an immovable cultural relic based on point clouds, characterized in that, Includes the following steps: Step 1: Data Collection and Preliminary Modeling The first three-dimensional point cloud model of the cultural relic was obtained by laser scanning and software processing; the second three-dimensional point cloud model of the cultural relic was obtained by close-range photogrammetry and software processing. Step 2: Point cloud model registration: Using the first 3D point cloud model as the reference model and the second 3D point cloud model as the registration model, while keeping the position of the reference model unchanged, the registration model is moved so that its center point coincides with the center point of the reference model; then the registration model is rotated around the center point so that the two models basically coincide; finally, the 3D model points of the registration model are iteratively calculated so that the average distance between the 3D model points of the registration model and the corresponding points of the 3D model of the reference model is less than a given threshold. Step 3: Repairing the artifact model: For the 3D model obtained in step two, determine the overlapping and non-overlapping parts of the two models. Then, retain the model with higher point cloud density, discard the overlapping parts of the model with lower point cloud density, and retain the non-overlapping parts of the model with lower point cloud density. The combination of all retained parts of the model is the final 3D model of the cultural relic.

2. The method for establishing a three-dimensional model of immovable cultural relics based on point clouds according to claim 1, characterized in that: The method for obtaining the first three-dimensional point cloud model described in step one is as follows: set up measuring stations around the cultural relic and deploy spherical connecting targets. Use a laser to scan the cultural relic to directly obtain the laser point cloud of each measuring station. The overlap of the point cloud coverage of adjacent measuring stations is required to be more than 50%. Then, using the supporting software, the point clouds of each scanning measuring station are stitched together using the spherical targets to construct the first three-dimensional point cloud model of laser scanning.

3. The method for establishing a three-dimensional model of immovable cultural relics based on point clouds according to claim 2, characterized in that: The method for obtaining the second three-dimensional point cloud model described in step one is as follows: Use a digital camera to take pictures of the cultural relic, requiring the photos to fully cover the cultural relic and adjacent photos to have an overlap of more than 65%; then perform matching calculations in the supporting software to construct a three-dimensional patch model, extract the vertices of the three-dimensional patch model, and thus construct the second three-dimensional point cloud model of close-range photogrammetry.

4. The method for establishing a three-dimensional model of immovable cultural relics based on point clouds according to claim 3, characterized in that: The method for determining the model center point described in step two involves: measuring the target coordinates using a surveying instrument, then importing the global coordinates into the laser point cloud to obtain the point cloud coordinates of the 3D model, and finally using the formula... Calculate the center coordinates of the two models; Where X, Y, and Z are the three coordinate components of the model's center coordinate value. This represents the three-dimensional coordinate value corresponding to the coordinate value of the i-th point in the model.

5. The method for establishing a three-dimensional model of immovable cultural relics based on point clouds according to claim 4, characterized in that: The iterative calculation method described in step two is as follows: Let the 3D point set of the baseline model be U1, and the 3D point set of the registration model be U2. The specific calculation steps are as follows: a. Calculate the nearest point in the U1 point set for each point in U2; b. Find the rigid body transformation that minimizes the average distance between the corresponding points mentioned above, and find the translation and rotation parameters. c. Apply the translation and rotation parameters obtained in the previous step to U2 to obtain a new set of transformation points U3; d. If the average distance between the two point sets U3 and U1 is less than a given threshold, then stop the iterative calculation; otherwise, repeat steps a to c with U3 as the new U2 until the latest transformed point set U3 and U1 satisfy the condition that the average distance between the two point sets is less than the given threshold.

6. The method for establishing a three-dimensional model of immovable cultural relics based on point clouds according to claim 5, characterized in that: The method for determining overlapping and non-overlapping parts as described in step three includes the following steps: Step (1) Obtain the normal of the point cloud of the reference model. Use the base model or the registration model as the reference model and the other model as the non-reference model. Take the centroid of the reference model as the starting point and define the normal of the point cloud as pointing outwards as the positive direction. Calculate the normal vector of each point in the reference model. Step (2) Starting with the points of the reference model, search for the points of the non-reference model along the positive and negative directions of the normal to obtain the overlapping area. Before obtaining the overlapping area, mark the category field value of all points of the reference model and the non-reference model as "1". The specific method for determining the overlapping region is as follows: First, construct a rotating cylinder with the normal as the axis of rotation and a specified distance value a as the radius; Next, find the points that fall inside the cylinder; if a point is found, save it as a candidate point dataset, and consider the reference model and non-reference model to be overlapping parts at this point; Finally, calculate the distance from each point in the candidate point dataset to the normal, mark the point with the smallest distance, and change its category field value to "6"; at the same time, change the category field value of the reference model point corresponding to this candidate point dataset to "2". Step (3) In step (2), if no point is found, it is considered to be a non-overlapping area; at the same time, the category field value of this reference model point is modified to "3"; After the above process, the category field value of all points in the reference model becomes "2" or "3", and the category field value of all points in the non-reference model becomes "1" or "6". Point clouds with category field values ​​of "2" and "6" are overlapping areas, and those with category field values ​​of "1" or "3" are non-overlapping areas.

7. The method for establishing a three-dimensional model of immovable cultural relics based on point clouds according to claim 6, characterized in that: The specified distance value a in step (2) is twice the average point spacing between the baseline model and the registration model.

Images