Cultural relic fragment set visual restoration method, system and device based on certainty

Abstract

Compared with the prior art, the cultural relic fragment visual restoration method based on certainty provided by the invention has the advantages that based on the to-be-restored cultural relic fragment data set, the user can identify errors in the restoration process on three levels of category, object and fragment through the restoration model and in combination with size, category certainty and matching certainty indexes; the difficulty of splicing and repairing the mixed fragments from multiple types of objects is reduced; the invention further provides a visual repair system, the certainty indexes of cultural relic fragment repair are visualized through the bar chart and the relational graph, the splicing process and potential splicing schemes between the fragments and the object are displayed, the certainty indexes of different splicing schemes are compared, visual interaction analysis is combined, a user controls and intervenes in the repair process, and the repair efficiency is improved. Therefore, the repairing work is more accurate and systematized, the method can be expanded to the splicing work of objects of different types or fragments of different scales, and the accuracy and repairing efficiency of the splicing work of the cultural relic fragments are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cultural relic splicing, and in particular to a method, system and device for visually repairing a set of cultural relic fragments based on determinism. Background Art

[0002] Cultural relics carry the splendid civilization and are an indispensable part of historical and cultural research. However, a large number of cultural relics have inevitably been damaged in the long history, and most of them have been found as a pile of mixed fragments. Although professional restorers are able to restore most cultural relics, it is very time-consuming to manually find the correct splicing solution due to the huge fragment matching space and a large amount of trial and error costs.

[0003] Breakthroughs in computational geometry and machine learning techniques have opened up new prospects for fragment restoration. For the splicing of cultural relic fragments, existing methods include automatic splicing algorithms, immersive manual splicing methods, and human-machine combined splicing methods. These methods complete cultural relic restoration by handling two main tasks: 1) classifying fragments into categories with similar features (such as shape or color); 2) finding matches in fragments of the same category and splicing them. Although these methods can replace a lot of labor and provide a restoration preview to guide the actual restoration work, the reliability of these restoration results is low.

[0004] Cultural relic fragments usually have irregular shapes due to damage and erosion. When repairing cultural relic fragments, the existing technologies mostly splice them based on the characteristics of the fragments, but damage and erosion may erase the key features used for category prediction and fragment splicing.

[0005] Therefore, objects that have been automatically repaired often contain many incorrectly spliced ​​fragments, which urgently need to be discovered and corrected by the restorer; the immersive manual splicing method mainly focuses on the manual alignment of two fragments from a single object, and does not consider the classification and restoration problems faced by mixed fragments from multiple objects. The current human-machine combined method does not repair the splicing results of mixed multi-object fragments.

[0006] In addition, there are two challenges in correcting restoration errors of cultural relic fragments. First, it is difficult to check the restoration errors caused by the model. Although cultural relic restorers can manually judge the correctness of categories and splicing, this process becomes time-consuming and laborious as the number of fragments increases. There are many subordinate associations between fragments and categories and objects, and cultural relic restorers need to check the correctness of each subordinate association. Although there are currently methods of embedded node-link graphs that indicate the correctness of matching by visualizing the certainty of fragment matching relationships within a single object, how to show certainty between various objects and categories to assist in inspection is still rarely studied.

[0007] Secondly, there is a lack of solutions to correct erroneous fragments. After separating the erroneous fragments, there is still a huge search space between the erroneous fragments and the object. Restorers usually give priority to fragments that are easy to correct for splicing attempts, and these splicing results will serve as a breakthrough for subsequent restoration. However, evaluating the difficulty of restoring fragments requires multiple comparisons of many potential solutions and their subsequent results, and most methods rely on previewing all splicing solutions. It is very time-consuming to find the correct splicing method among a large number of splicing solutions. Summary of the invention

[0008] The purpose of the present invention is to provide a method, system and device for visual restoration of a set of cultural relic fragments based on determinism, which evaluates incorrectly spliced ​​fragments based on the size, category certainty and matching certainty indicators of the cultural relic fragments, and explores the splicing schemes and splicing results of the fragments in combination with interactive visual analysis operations, solves the problem of splicing multiple fragments and helps users find the correct splicing method among a large number of splicing schemes.

[0009] To achieve the above-mentioned purpose of the invention, an embodiment provides a method for visually repairing a set of cultural relic fragments based on determinism, comprising the following steps:

[0010] Select the data set and restoration model of the cultural relic fragments to be restored, and generate the initial restoration results of the cultural relic fragments;

[0011] Based on the initial restoration results of the cultural relic fragments, the possibility of fragment errors is identified by calculating the certainty indicators of categories, objects and fragments, and the possibility of fragment errors is visualized;

[0012] Based on deterministic indicators, explore the splicing relationship of fragments, compare and evaluate splicing plans, correct errors in fragment splicing, and complete the restoration of the cultural relic fragment set.

[0013] In one embodiment, the dataset of cultural relic fragments to be restored is a three-dimensional point cloud dataset of cultural relic fragments, and the three-dimensional point cloud dataset is derived from a three-dimensional scanning point cloud of cultural relic fragments at an archaeological excavation site.

[0014] In one embodiment, the restoration model includes a classification model and a splicing model, which are used to automatically classify and splice the cultural relic fragment set;

[0015] The classification is based on a three-dimensional point cloud data set of cultural relic fragments, using a classification model to predict the probability of each cultural relic fragment belonging to a category, and classifying each cultural relic fragment with the maximum value of the predicted probability;

[0016] The splicing is based on the classification result. The fragments of the same category are assembled into an object using a splicing model, and the alignment result is obtained using the chamfer distance, and the fragment pairs are iteratively formed.

[0017] In one embodiment, the iterative formation of fragment pairs includes: iteratively forming fragment pairs through a greedy algorithm and satisfying constraints; wherein the constraints include a closed loop of position transformation and non-overlapping spatial positions.

[0018] In one embodiment, the certainty indicators of the categories, objects and fragments include: size, category certainty and matching certainty;

[0019] The size is obtained by evaluating the size of the fragments of the cultural relic;

[0020] The category certainty is obtained by using a classification model to predict the probability of each cultural relic fragment belonging to a category;

[0021] The matching certainty is obtained by evaluating the matching quality between the fragments based on the chamfer distance between the matching surfaces.

[0022] In one embodiment, the correction of fragment splicing errors includes: the user visualizes the results based on the possibility of fragment errors, identifies and selects easy-to-splice cultural relic fragments, uses a splicing model to generate the best candidate splicing results for the selected easy-to-splice cultural relic fragments, and the user manually adjusts the position of the cultural relic fragments or accepts the splicing result.

[0023] In order to clearly demonstrate the deterministic visual restoration method for cultural relic fragments, the embodiment provides a deterministic visual restoration system for cultural relic fragments, which is used to complete the visual restoration method for cultural relic fragments. The visual restoration system for cultural relic fragments includes: a user panel module, a quality view module, a relationship view module, and an instance view module;

[0024] The user panel module is used to display the initial splicing result of the cultural relic fragment data set and the error correction of the splicing of the cultural relic fragment set through data information, restoration model, restoration result, correction information and step control; wherein the data information displays the selected cultural relic fragment data set and the number of fragments, the restoration model displays the selected classification model and splicing model, the restoration result displays the number of initial restoration categories and the number of restored objects based on the selected restoration model, the correction information displays the error information identified in each stage, and the step control is used to switch the inspection stage or the error correction stage;

[0025] The quality view module provides identification symbols, thumbnails of the appearance of objects, indicator histograms and indicator stacked bar graphs, which are used to display the quantitative indicators of the current splicing and restoration results and alternative solutions, wherein the identification symbols display the selected categories, objects or fragments, the indicator histograms display the size of the cultural relic fragment set, the distribution of category certainty and matching certainty indicators, and the indicator stacked bar graphs display the average size of the fragments, category certainty and matching certainty indicators of the single fragments or categories and objects corresponding to each identification symbol;

[0026] The relationship view module is used to display the appearance of the spliced ​​objects and categories and the certainty of the alternative repair solutions through a relationship diagram, wherein the certainty of the alternative repair solutions is presented on a two-dimensional plane through a force-directed layout algorithm;

[0027] The instance view module provides an information panel and a preview of the spliced ​​object to display the number of fragments contained in the spliced ​​object and the matching certainty, helping the user to judge splicing errors or confirm the correct splicing method.

[0028] In one embodiment, the indicator stacked bar chart arranges the artifact fragments in descending order based on the selected certainty indicator, wherein the matching certainty indicator sets an error bar to display the matching certainty indicator range of the fragments in the current category or object and other fragments in the same category.

[0029] In one embodiment, the certainty of the alternative repair solutions is presented on a two-dimensional plane through a force-directed layout algorithm, including: locating the position of the object or category node on the two-dimensional plane through the force-directed layout algorithm, and exploring alternative repair solutions through categories and matching deterministic encoding distance parameters.

[0030] An embodiment provides a deterministic visual restoration device for cultural relic fragments, including a memory and a processor, wherein the memory is used to store a computer program, fragment category certainty and matching certainty values, and classification and splicing results; and the processor is used to implement a deterministic visual restoration method for cultural relic fragments when executing the computer program.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] Compared with the prior art, the deterministic visual restoration method for cultural relic fragments provided by the present invention is based on a data set of cultural relic fragments to be restored. By restoring a model and combining size, category certainty and matching certainty indicators, the method enables users to identify errors in the restoration process at three levels: category, object and fragment, thereby reducing the difficulty of splicing and restoring mixed fragments from multiple categories of objects. The present invention also provides a deterministic visual restoration system for cultural relic fragments, which visualizes the deterministic indicators of cultural relic fragment restoration through bar charts and relationship diagrams, and displays the splicing process and potential splicing schemes between fragments and objects, compares the deterministic indicators of different splicing schemes and combines them with visual interactive analysis. Users can control and intervene in the restoration process, making the restoration work more precise and systematic, and can be extended to the splicing work of different types of objects or fragments of different sizes, thereby improving the accuracy of splicing work and the restoration efficiency of cultural relic fragments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0034] Figure 1 This is a flow chart of the visual restoration method of cultural relic fragments based on determinism;

[0035] Figure 2 This is a schematic diagram of the interface of the visual restoration system for cultural relics fragments based on determinism;

[0036] Figure 3 This is a schematic diagram of the quality view module in the visual restoration system for cultural relics fragments;

[0037] Figure 4 This is a schematic diagram of the quality view module being expanded into three levels: category, object, and fragment;

[0038] Figure 5 A schematic diagram of the stitching scheme is provided for the quality view-based module;

[0039] Figure 6 It is a schematic diagram of the relationship view module in the visual restoration system of cultural relics fragments;

[0040] Figure 7 This is a schematic diagram of the instance view module in the visual restoration system for cultural relic fragments. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0042] In order to identify erroneous fragments in fragment splicing and restoration and improve the accuracy and efficiency of multi-fragment restoration work, the embodiment provides a deterministic visual restoration method for cultural relic fragments. The method generates an initial restoration result based on the size, category certainty and matching certainty indicators of the cultural relic fragments, evaluates the erroneous splicing fragments based on the initial restoration result, corrects the errors in fragment splicing, and completes the restoration of the cultural relic fragment set. Figure 1 As shown, the deterministic visual restoration method for cultural relic fragments provided in the embodiment includes the following steps:

[0043] S1. Select a data set of cultural relic fragments to be restored and a restoration model to generate an initial restoration result of the cultural relic fragments.

[0044] By loading 3D point cloud datasets containing cultural relic fragments, these datasets can come from 3D scanning point clouds of cultural relic fragments at archaeological excavation sites. Select appropriate restoration models to process these cultural relic fragments. The restoration model can be based on deep learning, geometric matching or other computational geometry technologies. The restoration model includes splicing models and classification models.

[0045] Classification is performed based on the classification model of fragments, training and testing are performed through the three-dimensional point cloud data set, the probability of each category to which each cultural relic fragment belongs is predicted, and each cultural relic fragment is classified according to the maximum predicted probability.

[0046] The splicing model of fragments includes the following steps: first, the fracture surface area is identified using surface segmentation technology; second, each fragment is aligned through its fracture area using a pairing model to form a fragment pair, the chamfer distance between the areas is calculated and normalized; finally, the fragment pairs are iteratively formed through a greedy algorithm to maximize the total matching score while satisfying certain constraints, where the constraints include position transformation closure and spatial position non-overlap to ensure correct splicing.

[0047] Based on the selected restoration model, the fragments of the cultural relic are classified and spliced ​​to generate preliminary restoration results, including the classified fragment categories and spliced ​​cultural relics, providing a basis for the user's subsequent evaluation and correction.

[0048] S2. Based on the initial restoration results of the cultural relic fragments, the possibility of fragment errors is identified by calculating the deterministic indicators of categories, objects and fragments, and the possibility of fragment errors is visualized.

[0049] At this point, the error checking phase begins, and the embodiment calculates a certainty index for each fragment, object, and category by calculating the category certainty and matching certainty of the fragment. The purpose of this step is to identify and display certainty and potential errors in the repair process.

[0050] Certainty indicators include size, category certainty and matching certainty: size refers to the size of the cultural relic fragment; category certainty refers to the probability that the fragment is classified into a specific category, and the embodiment can use a classification model to predict this probability; matching certainty refers to the quality of matching between pairs of fragments, and the embodiment can use the chamfer distance value in the matching model to evaluate this indicator.

[0051] The fragment-level indicators reflect the certainty that the fragments belong to a specific category and match; the object-level indicators reflect the overall certainty of the fragments in the object, which is obtained by the average certainty at the fragment level; the category-level indicators reflect the overall certainty of a class of objects, which is obtained by the average certainty at the object level.

[0052] Based on these certainty indicators, users identify possible errors, such as classification errors or matching errors, which may lead to inaccurate restoration of the artifact and determine the fragments that need restoration.

[0053] S3. Based on deterministic indicators, explore the splicing relationship of fragments, compare and evaluate splicing plans, correct errors in splicing fragments, and complete the restoration of the cultural relic fragment set.

[0054] At this point, the error correction stage is entered. The embodiment uses a visualization method to display potential candidate matching relationships and alternative certainty indicators between fragments, so that users can intuitively compare and evaluate different splicing schemes; users understand factors such as the proximity of fragments, color coding, and the distance between fragments to evaluate the difficulty of splicing, identify and select easy-to-splice fragments, select one of the cultural relic fragments, and the splicing model generates 5 best candidate splicing results for the alternative fragments. The user interactively selects and compares fragments, and carefully checks the alignment and overall appearance of each fragment before final confirmation. The user confirms the splicing of fragments by manually adjusting the fragment position or accepting the splicing scheme recommended by the splicing model, performs error correction, and completes the restoration of the cultural relic fragment set.

[0055] Through the deterministic visual restoration method for cultural relic fragments of this embodiment, experts can effectively identify and correct errors in the model during the restoration process, thereby improving the accuracy and efficiency of cultural relic restoration. This method is particularly suitable for restoration work involving a large number of fragments and complex cultural relics. It provides experts with a powerful method to support their decision-making and restoration strategies through visual deterministic indicators and splicing relationships.

[0056] In order to clearly demonstrate the method for visual restoration of cultural relics fragments based on determinism, the embodiment provides a system for visual restoration of cultural relics fragments based on determinism, such as Figure 2 As shown, the visual repair system includes a user panel module, a quality view module, a relationship view module and an instance view module;

[0057] In the embodiment, the user panel module is used to display the initial splicing result of the cultural relic fragment data set and the error correction of the splicing of the cultural relic fragment set through data information, restoration model, restoration result, correction information and step control; wherein the data information displays the selected cultural relic fragment data set and the number of fragments, the restoration model displays the selected classification model and splicing model, the restoration result displays the initial restoration category number and the number of restored objects based on the selected restoration model, the correction information displays the error information identified in each stage, and the step control is used to switch the inspection stage or the error correction stage;

[0058] In an embodiment, the quality view module can present quantitative indicators of the restoration of cultural relic fragments and alternative solutions in a hierarchical structure, summarize the certainty of cultural relic fragments, and allow comparison and identification of potential errors at multiple levels. In this embodiment, the quality view module provides a list of splicing results and three indicators of size, category certainty, and matching certainty to support error checking.

[0059] like Figure 3 As shown in the figure, in order to summarize each indicator, three indicator histograms are listed above the list, corresponding to the distribution of the three indicators of size, category certainty, and matching certainty. In each indicator histogram, the x-axis represents the indicator size, in which the mean of the indicator is highlighted; the y-axis represents the number of cultural relics fragments in each indicator value range, in which the highest number of values ​​is highlighted, so that a statistical overview of the corresponding indicator can be obtained.

[0060] Below the quality view module are identification symbols, a thumbnail of the object's appearance, and a stacked bar chart of indicators. Each row represents a category based on color coding. The identification symbols represent categories, objects, and fragments.

[0061] The leftmost bar in the indicator stacked bar chart represents the average size of the fragments in that category, helping users focus on more important fragments; the middle bar describes the category certainty; the rightmost bar identifies the matching certainty. All indicators are normalized and the indicator values ​​are mapped to the length of the bar chart. By default, categories are listed in descending order of the sum of the indicators (the total length of the three indicator bars) for category certainty comparison. Therefore, lower-ranked categories indicate a higher probability of error.

[0062] The quality view module also supports multi-level deterministic exploration, allowing users to drill down to the object and fragment level and identify erroneous fragments. In this embodiment, the splicing quality module models the category-object-fragment hierarchy as a nested structure, with child items (objects, fragments) nested and indented under their parent items (category, object), and indicator stacked bar charts display their corresponding deterministic indicators. Items at the same level are aligned to the left and highlighted with a vertical dashed line. By default, all sub-hierarchies are hidden and can be expanded as needed by double-clicking the corresponding row.

[0063] like Figure 4 As shown in the figure, when you double-click to select the parent item t3, the object's child items will be expanded accordingly, including: 15 , o 11 , o 13 , o 14 , o 12 , when double-clicking selects the object's sub-item o14 , the corresponding fragment sub-items will be expanded, including: 67 、f71 、f 76 .

[0064] The quality view module can also visualize the deterministic indicators of the fragments being assembled into candidate objects or belonging to other categories, providing valuable clues for fragment error assessment and correction. Figure 5 As shown in the figure, after clicking on the category certainty histogram, the category certainty indicator shows the certainty bar chart (t3, t2, t1, t4) of each fragment belonging to other categories through color coding. For the convenience of comparison, the fragments are arranged in descending order based on the category certainty indicator quality view module; the leftmost bar (dotted box) in the category certainty indicator bar chart corresponds to the current category certainty; Figure 5 The part selected by 1 in the figure represents object o 13 The probability of belonging to category t1.

[0065] In addition, in order to visualize the matching certainty of the specified solution, the quality view module adds error bars to the matching certainty bar chart to show the certainty range of matching the fragments of the current object to the fragments of other objects; when the matching certainty indicator is selected, the background color of the matching certainty indicator histogram is darkened, and the fragments are arranged in descending order based on the matching certainty quality view module.

[0066] In the embodiment, the relationship view module is used to display the appearance of the spliced ​​objects, categories and the certainty of the alternative repair solutions through the relationship diagram. Figure 6 As shown, each node represents the geometry or category of an object or fragment, and the category node is represented by the object with the highest total index value to help experts identify the category. The nodes are consistent with the items displayed in the splicing quality module, and the certainty of the alternative repair solutions is presented on a two-dimensional plane through a force-directed layout algorithm.

[0067] like Figure 6 1 in the figure shows a relationship diagram of categories. Each node represents a category of cultural relic fragments. Category nodes are represented by objects with the highest splicing quality and are coded by color. The distance between nodes and arrows encode the certainty of alternatives between categories. The higher the certainty of a node and an alternative category, the closer it is to the alternative node. Therefore, the relationship diagram can compare alternatives with the highest certainty.

[0068] The relationship graph allows for drill-down analysis and is consistent with the multi-level exploration in the Quality View module. When a category or object entry is expanded in the Quality View module, the relationship graph replaces the corresponding node with the geometry of its child objects or fragments and lays them out according to their candidate matching certainty. Figure 6 As shown in 2, for example, Figure 6The category node is connected to the object o14 with the highest category certainty. Each object node is connected to its alternative with the highest certainty, so the shorter edge indicates that the fragments in the current object are likely to be spliced ​​with the fragments of the alternative object, which also means that the fragments in the object are at a higher risk of category errors. These mechanisms allow users to compare suspicious objects in turn and infer the wrong category of the cultural relic fragments.

[0069] like Figure 6 As shown in Figure 3, after drilling down to the fragment level, the relationship diagram further supports fragment-level analysis. 14 ), the relationship graph will show the matching certainty or category certainty of the fragments inside this object with other objects or categories. Most of the edges are from object o 14 The bottle body fragments start to move to the bottle body parts of other objects of the same category, indicating that the splicing model may confuse such fragments.

[0070] like Figure 6 As shown in 4, when the user further selects a fragment (such as f 76 ), the relationship graph points all edges to the selected fragment and adjusts the node layout to show the matching certainty or category certainty of this fragment to other objects or categories through relative distance.

[0071] When the user has finished comparing and identifying the erroneous fragments, right-click on the fragment entry in the Quality View module to select the fragment to be corrected.

[0072] After the selection of the faulty fragments is completed and the error correction phase is entered through the step control of the user panel module, the relationship graph visualizes the potential stitching solutions. At this point, the relationship graph consists of nodes representing the objects and the faulty fragments selected from the inspection phase. Figure 2 As shown in the relationship view module in Figure 1, the fragment nodes are highlighted in darker tones, and these nodes are also laid out using force-directed layout, where the force between nodes is encoded as the matching certainty between nodes. Following the same design paradigm, the edges of the relationship graph connect each fragment to the node with the highest matching certainty, indicating the most likely splicing steps. Therefore, the splicing difficulty of the fragment can be evaluated by patterns such as the length of the link and the number of splicing steps.

[0073] The instance view module is used to display the spliced ​​object in detail, helping users to select incorrect fragments and confirm the correct splicing solution. Figure 7As shown in the figure, when the user selects an entry, the instance view module will display the entry and its spliced ​​object preview and the information panel accordingly. The information panel contains the object or category identifier to which the selected entry belongs, the number of fragments contained in the object, and the matching certainty. Among them, the fragments in the object preview are color-coded for matching certainty, and darker tones indicate lower matching certainty, indicating that careful observation is required.

[0074] The user can further enter the virtual reality space, put on the virtual reality headset and handle to view the three-dimensional fragmented objects, and use the handle to grab, move and interactively drag the fragments to explore and correct the alignment between the fragments. After confirming that the alignment is correct, return to the front-end interface and click the information panel to complete the confirmation of the splicing plan.

[0075] The relationship graph will then be re-arranged in a force-directed manner based on the determined splicing scheme and the matching certainty of the remaining fragments. The user repeats the above steps until the splicing scheme for all fragments is completed.

[0076] In order to clearly demonstrate the deterministic visual restoration method for cultural relic fragments, an embodiment further provides a deterministic visual restoration device for cultural relic fragments, including a memory and a processor, the memory being used to store a computer program, fragment category certainty and matching certainty values, and classification and splicing results, and the processor being used to implement the deterministic visual restoration method for cultural relic fragments when executing the computer program.

[0077] By utilizing the above-mentioned deterministic visual restoration method, system and device for a set of cultural relic fragments, the incorrectly spliced ​​fragments are evaluated through the size, category certainty and matching certainty indicators of the cultural relic fragments, and the splicing schemes and results of the fragments are explored in combination with interactive visual analysis operations, so as to solve the problem of splicing multiple fragments and help users find the correct splicing method among a large number of splicing schemes.

[0078] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for visual restoration of a set of cultural relic fragments based on determinism, characterized in that: The following steps are involved: Select the data set and restoration model of the cultural relic fragments to be restored, and generate the initial restoration results of the cultural relic fragments; Based on the initial restoration results of the cultural relic fragments, the possibility of fragment errors is identified by calculating the certainty indicators of categories, objects and fragments, and the possibility of fragment errors is visualized; Based on deterministic indicators, explore the splicing relationship of fragments, compare and evaluate splicing plans, correct errors in fragment splicing, and complete the restoration of the cultural relic fragment set.

2. The visual restoration method for a collection of cultural relics fragments according to claim 1, characterized in that: The data set of cultural relic fragments to be restored is a three-dimensional point cloud data set of cultural relic fragments, and the three-dimensional point cloud data set comes from the three-dimensional scanning point cloud of the cultural relic fragments at the archaeological excavation site.

3. The visual restoration method for a collection of cultural relics fragments according to claim 1, characterized in that: The restoration model includes a classification model and a splicing model, which are used to automatically classify and splice the cultural relic fragment set; The classification is based on a three-dimensional point cloud data set of cultural relic fragments, using a classification model to predict the probability of each cultural relic fragment belonging to a category, and classifying each cultural relic fragment with the maximum value of the predicted probability; The splicing is based on the classification result. The fragments of the same category are assembled into an object using a splicing model, and the alignment result is obtained using the chamfer distance, and the fragment pairs are iteratively formed.

4. The visual restoration method for a collection of cultural relics fragments according to claim 3 is characterized in that: The iterative formation of fragment pairs includes: iteratively forming fragment pairs through a greedy algorithm and satisfying constraints; wherein the constraints include a closed loop of position transformation and non-overlapping spatial positions.

5. The visual restoration method for a collection of cultural relics fragments according to claim 1, characterized in that: Certainty indicators of the categories, objects and fragments, including size, category certainty and matching certainty; The size is obtained by evaluating the size of the fragments of the cultural relic; The category certainty is obtained by using a classification model to predict the probability of each cultural relic fragment belonging to a category; The matching certainty is obtained by evaluating the matching quality between the fragments based on the chamfer distance between the matching surfaces.

6. The visual restoration method for a collection of cultural relics fragments according to claim 1, characterized in that: The correction of the fragment splicing error includes: the user visualizes the result based on the possibility of fragment error, identifies and selects the cultural relic fragments that are easy to splice, uses the splicing model to generate candidate splicing results for the selected cultural relic fragments that are easy to splice, and the user manually adjusts the position of the cultural relic fragment or accepts the splicing result.

7. A deterministic visual restoration system for cultural relics fragments, characterized in that: Used to complete the visual restoration method of cultural relic fragments according to any one of claims 1 to 6, the visual restoration system of cultural relic fragments comprises: a user panel module, a quality view module, a relationship view module and an instance view module; The user panel module is used to display the initial splicing result of the cultural relic fragment data set and the error correction of the splicing of the cultural relic fragment set through data information, restoration model, restoration result, correction information and step control; wherein the data information displays the selected cultural relic fragment data set and the number of fragments, the restoration model displays the selected classification model and splicing model, the restoration result displays the number of initial restoration categories and the number of restored objects based on the selected restoration model, the correction information displays the error information identified in each stage, and the step control is used to switch the inspection stage or the error correction stage; The quality view module provides identification symbols, thumbnails of the appearance of objects, indicator histograms and indicator stacked bar graphs, which are used to display the quantitative indicators of the current splicing and restoration results and alternative solutions, wherein the identification symbols display the selected categories, objects or fragments, the indicator histograms display the size of the cultural relic fragment set, the distribution of category certainty and matching certainty indicators, and the indicator stacked bar graphs display the average size of the fragments, category certainty and matching certainty indicators of the single fragments or categories and objects corresponding to each identification symbol; The relationship view module is used to display the appearance of the spliced ​​objects and categories and the certainty of the alternative repair solutions through a relationship diagram, wherein the certainty of the alternative repair solutions is presented on a two-dimensional plane through a force-directed layout algorithm; The instance view module provides an information panel and a preview of the spliced ​​object to display the number of fragments contained in the spliced ​​object and the matching certainty, helping the user to judge splicing errors or confirm the correct splicing method.

8. The visual restoration system for cultural relic fragments according to claim 7, characterized in that: The indicator stacked bar chart arranges the cultural relic fragments in descending order based on the selected certainty indicator, wherein the matching certainty indicator sets an error bar to display the matching certainty indicator range of the fragments in the current category or object and other fragments in the same category.

9. The visual restoration system for cultural relic fragments according to claim 7, characterized in that: The certainty of the alternative repair scheme is presented on a two-dimensional plane through a force-directed layout algorithm, including: locating the position of the object or category node on the two-dimensional plane through the force-directed layout algorithm, and exploring alternative repair schemes through category and matching deterministic encoding distance parameters.

10. A visual restoration device for cultural relic fragments based on certainty, comprising a memory and a processor, wherein the memory is used to store a computer program, certainty of fragment categories and matching certainty values, and classification and splicing results, characterized in that: The processor is used to implement the deterministic visual restoration method for cultural relic fragments according to any one of claims 1 to 6 when executing the computer program.

Images