CN117495674A - Point cloud splicing method for cultural relic fragments - Google Patents

Point cloud splicing method for cultural relic fragments Download PDF

Info

Publication number
CN117495674A
CN117495674A CN202311428559.XA CN202311428559A CN117495674A CN 117495674 A CN117495674 A CN 117495674A CN 202311428559 A CN202311428559 A CN 202311428559A CN 117495674 A CN117495674 A CN 117495674A
Authority
CN
China
Prior art keywords
point cloud
fragments
cultural relic
cloud
wudian
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311428559.XA
Other languages
Chinese (zh)
Inventor
祝磊
刘兴奇
兰德省
易伟同
胡云岗
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Civil Engineering and Architecture
Original Assignee
Beijing University of Civil Engineering and Architecture
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Civil Engineering and Architecture filed Critical Beijing University of Civil Engineering and Architecture
Priority to CN202311428559.XA priority Critical patent/CN117495674A/en
Publication of CN117495674A publication Critical patent/CN117495674A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the technical field of point cloud data splicing, in particular to a point cloud splicing method for cultural relic fragments, which comprises the following steps: acquiring and optimizing data of Wen Wudian cloud fragments, and establishing Wen Wudian cloud fragment data sets; training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model; performing first numbering on the cultural relic point cloud fragments, performing polygon fitting on the cultural relic point cloud fragments, performing second numbering, and randomly generating initial position matrixes of the cultural relic point cloud fragments and the cultural relic point cloud polygons; and calculating Wen Wudian cloud fragment intervals and overlapping volumes among the cultural relic point cloud polygons by adopting a peak inundation search algorithm, updating an initial position matrix if a calculation result has a better solution, iterating for a plurality of times to obtain an optimal position matrix, and splicing the global Wen Wudian cloud fragments. By adopting the splicing method, the optimal solution is obtained according to multiple iterations, and the accuracy of the cultural relic point cloud splicing result is improved.

Description

Point cloud splicing method for cultural relic fragments
Technical Field
The invention relates to the technical field of point cloud data splicing, in particular to a point cloud splicing method for cultural relic fragments.
Background
In the fields of cultural relic protection and archaeological excavation in China, many cultural relics are usually unearthed in a fragment form due to factors such as brittleness of materials, underground erosion and weathering, survey exploitation means limitation and the like, and the situation is a very important loss for historical inheritance in China and is also a very important issue in the field of cultural relic protection in China. The traditional cultural relic fragment restoration work usually adopts manual splicing by workers depending on experience, and the method is generally low in efficiency, seriously influenced by practical conditions and needs a large amount of manpower and material resources.
In recent years, due to the rapid development of the three-dimensional laser scanning technology and the artificial intelligence technology, a solid model is converted into a point cloud model through a three-dimensional laser scanner, and many works by means of manual operation can be solved through reverse digitization and automation, so that automation and intellectualization of the recovery work of terracotta figures are possible. The artificial intelligence is an emerging computer research field in recent years, and based on a targeted point cloud artificial intelligence depth model network, the work of identifying, classifying, dividing and the like of the point cloud is realized by training different network models, and meanwhile, the splicing work of the digital model can be realized by matching with corresponding algorithms.
The common point cloud splicing algorithm is to find a specific type of point cloud characteristics and then realize the splicing of the corresponding models by a point cloud characteristic matching method. The heuristic algorithm is usually used for carrying out optimization solution in an iterative mode, has the characteristics of wide applicability and strong stability, and therefore has strong practical significance in carrying out point cloud matching operation through the heuristic algorithm.
At present, the research on point cloud splicing is focused on the work of more accurate registration of specified point clouds, and a few near blanks are researched on a global and large-quantity splicing method; because the point cloud data volume is large and the difference of the point cloud characteristics of different kinds of fragments is large, a single characteristic matching algorithm is difficult to achieve a splicing result with high global point cloud fitness and high automation degree; the traditional heuristic algorithm has the problems of large calculation amount, large quantity of adverse results in the calculation process, easy occurrence of local convergence, easy occurrence of convergence near an optimal solution and the like, so that the point cloud splicing method which has strong adaptability to point cloud data, is suitable for large quantity of data calculation, is not easy to occur local convergence and has high automation degree has extremely high practical value for the splicing working tool of the point cloud data in the field of cultural relics protection in China.
Disclosure of Invention
The embodiment of the invention provides a point cloud splicing method for cultural relic fragments, which aims to solve the technical problem of low accuracy of point cloud splicing results.
The point cloud splicing method for the cultural relic fragments provided by the embodiment of the invention comprises the following steps:
acquiring and optimizing data of Wen Wudian cloud fragments, and establishing Wen Wudian cloud fragment data sets;
training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model;
performing first numbering on the cultural relic point cloud fragments, performing polygon fitting on the cultural relic point cloud fragments, performing second numbering, and randomly generating initial position matrixes of the cultural relic point cloud fragments and the cultural relic point cloud polygons;
and calculating Wen Wudian cloud fragment intervals and overlapping volumes among the cultural relic point cloud polygons by adopting a peak inundation search algorithm, updating an initial position matrix if a calculation result has a better solution, iterating for a plurality of times to obtain an optimal position matrix, and splicing the global Wen Wudian cloud fragments.
Preferably, the acquiring and optimizing Wen Wudian cloud fragment data establishes Wen Wudian cloud fragment data sets, and includes the following steps:
acquiring Wen Wudian cloud fragment data;
removing data noise of Wen Wudian cloud fragments by using point cloud processing software, deleting data in-vitro orphan points of Wen Wudian cloud fragments, unifying data density and format of Wen Wudian cloud fragments, calibrating data incomplete surfaces and non-incomplete surfaces of Wen Wudian cloud fragments, and generating new cultural relic point cloud fragments according to Wen Wudian cloud fragment enhancement;
the Wen Wudian cloud data is randomly extracted, and a training set, a testing set and a verification set for deep learning are manufactured.
Preferably, training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model includes the following steps:
setting up a Pointernet++ network model frame;
training the data training set of the cultural relic point cloud fragments by using the Pointernet++ network to obtain a Wen Wudian cloud fragment deep learning segmentation model.
Preferably, the training of the data training set of the cultural relic point cloud fragments by using the Pointernet++ network further comprises the following steps after obtaining the Wen Wudian cloud fragment deep learning segmentation model:
and evaluating the deep learning segmentation model by adopting semantic segmentation evaluation indexes and accuracy.
Preferably, the first numbering is performed on the cultural relic point cloud fragments, the polygonal fitting is performed on the cultural relic point cloud fragments, the second numbering is performed, and the initial position matrix of the numbered cultural relic point cloud fragments and the cultural relic point cloud polygons is randomly generated, and the method comprises the following steps:
the global Wen Wudian cloud fragments to be matched are numbered first and are P respectively 1 、P 2 、P 3 、…、P N Judging the distance between any two cloud fragments to be matched Wen Wudian and a preset fragment distance threshold F, if the distance between any two cloud fragments to be matched is smaller than the preset fragment distance threshold F, performing polygon fitting on the two fragments, co-fitting N Wen Wudian cloud polygons, and performing second numbering on cultural relic point cloud polygons, wherein the distances between any two cloud fragments to be matched and the preset fragment distance threshold F are D respectively 1 、D 2 、D 3 、…、D N ；
And randomly generating G N multiplied by 6 initial position matrixes after the translation and rotation of N cultural relic point cloud fragments in the x, y and z directions are changed.
Preferably, the peak inundation search algorithm is adopted to calculate Wen Wudian intervals of cloud fragments and overlapping volumes among the cultural relic point cloud polygons, if a calculation result has a better solution, an initial position matrix is updated, the optimal position matrix is obtained through multiple iterations, and global Wen Wudian cloud fragments are spliced, including the following steps:
step of solving a stepping matrix, namely taking G initial position matrixes generated randomly as initial data of a peak inundation searching method, calculating stepping distances of the initial position matrixes, and moving each initial position matrix along positive and negative directions by the stepping distances to obtain a stepping position matrix;
updating the matrix, namely respectively calculating the point cloud distance and the overlapping volume according to the initial position matrix and the stepping matrix, and updating the initial position matrix into the stepping position matrix if the point cloud distance and the overlapping volume calculated according to the stepping matrix are better;
and (3) a loop step, namely, a step matrix step and a matrix updating step are obtained through loop iteration until the initial position matrix is updated.
Preferably, the step of solving the stepping matrix comprises the following steps:
the initial stepping distance of the initial position matrix is calculated as follows:
l p ＝L/2；
wherein L is the distance between adjacent initial position matrixes, and L p An initial stepping distance for a p-th initial position matrix;
and moving each initial position matrix along the positive and negative directions of the matrix by an initial stepping distance to obtain a stepping position matrix.
Preferably, the step of updating the matrix includes the steps of:
calculating according to the initial position matrix to obtain a first point cloud distance L of the nearest point between each cultural relic point cloud fragment and other cultural relic point cloud fragments ab Calculating to obtain a first overlapped volume C of each cultural relic point cloud polygon and other cultural relic point cloud polygons mn ；
Calculating a second point cloud distance L 'of the nearest point between each cultural relic point cloud fragment and other cultural relic point cloud fragments according to the stepping position matrix' ab Calculating to obtain a second overlapping volume C 'of each cultural relic point cloud polygon and other cultural relic point cloud polygons' mn ；
Comparing absolute value L of first point cloud distance by adopting non-dominant sorting method ab Absolute value of L and second point cloud distance L' ab Absolute value of first overlap volume |c mn Absolute value of i and second overlap volume |c' mn The magnitude of the i is such that,
if any |L is present ab I is greater than its corresponding L' ab I, and there is any |C mn I is greater than its corresponding |c' mn I, the initial position matrix is updated to be L' ab I and C' mn A step position matrix corresponding to the minimum;
if any |L is present ab The I is smaller than its corresponding I L' ab |, and/or there is any |C mn The I is smaller than its corresponding I C' mn Setting a step-down coefficient alpha in the next cycle, wherein the step-down coefficient alpha set each time is smaller than the step-down coefficient alpha of the previous cycle, the step-down coefficient alpha gradually approaches 0, and updating the step distance l 'of the next cycle of the initial position matrix' p The method comprises the following steps:
l’ p ＝α×l p ；
wherein (0 < alpha < 1).
Preferably, the step of updating the matrix further includes the steps of:
distance L between first point clouds ab Compared with a preset interval threshold H, if L ab And if the number is greater than H, the two part of the object point cloud fragments are considered to be non-adjacent Wen Wudian cloud fragments, the number of the two part of the object point cloud fragments is changed to be ≡, and the two part of the object point cloud fragments are not spliced.
Preferably, the circulating step further comprises the steps of:
setting a first water line W 1 And a second water line W 2 The method comprises the following steps:
W 1 ＝(L abmax -L abmin )×i/iter；W 2 ＝(C mnmax -C mnmin )×i/iter；
wherein L is abmax And L is equal to abmin C is the maximum value and the minimum value of the first point cloud distance in the current iteration respectively mnmax And C mnmin Respectively obtaining the maximum value and the minimum value of the first overlapped volume in the current iteration, wherein i is the current iteration number, and iter is the total iteration number;
step of solving stepping matrix and step of updating matrix by cyclic iteration, and setting floodingWithout interval J, a flooding operation is performed every iteration of the generation J, the flooding operation comprising: deleting the first point cloud space smaller than the first water line W 1 Deleting the corresponding initial position matrix, wherein the first overlapping volume is smaller than the second water level line W 2 Setting water level warning quantity A according to the corresponding initial position matrix, and randomly generating the initial position matrix if the quantity of the initial position matrix is smaller than A;
and after the loop iteration is completed, an optimal position matrix is obtained, and global Wen Wudian cloud fragments are spliced.
The embodiment of the invention has the beneficial effects that: according to the point cloud splicing method for the cultural relic fragments, provided by the invention, the data of Wen Wudian cloud fragments are obtained and optimized, the data of the cultural relic point cloud fragments are trained to obtain a deep learning segmentation model, automatic segmentation of the point cloud fragments is realized, the splicing difficulty is greatly reduced, and non-spliced point cloud computing is eliminated; an initial position matrix is randomly generated, so that the calculation efficiency is improved; the peak inundation search algorithm is introduced into the global splicing of the cultural relic point cloud fragments, the initial position matrix is updated according to the calculation result by calculating the distance between Wen Wudian cloud fragments and the overlapping volume between the cultural relic point cloud polygons, the optimal solution is obtained through multiple iterations, the splicing of the global Wen Wudian cloud fragments is completed, the optimal solution for the Wen Wudian cloud fragment splicing is effectively obtained, and the accuracy of the cultural relic point cloud splicing result is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic step diagram of a point cloud splicing method for cultural relic fragments according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a point cloud space curve of a point cloud fragment of a certain cultural relic in an iterative process according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of a coordinate point substitution peak flooding search algorithm provided in an embodiment of the present invention.
Fig. 4A is a schematic diagram of a portion a of the split point cloud fragments of the cultural relics by using the Pointnet++ module according to the embodiment of the present invention.
Fig. 4B is a schematic diagram of a portion B of the split point cloud fragments of the cultural relics by using the Pointnet++ module according to the embodiment of the present invention.
Fig. 5 is a schematic diagram of a splicing result of a part of object point cloud fragments according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of a splicing result of global Wen Wudian cloud fragments according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made apparent and fully in view of the accompanying drawings, in which some, but not all embodiments of the invention are shown. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention will now be described in further detail by way of specific examples of embodiments in connection with the accompanying drawings.
The embodiment of the invention provides a point cloud splicing method for cultural relic fragments, which is shown in fig. 1 and comprises the following steps:
s110, acquiring and optimizing data of Wen Wudian cloud fragments, and establishing Wen Wudian cloud fragment data sets;
s120, training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model;
s130, carrying out first numbering on the cultural relic point cloud fragments, carrying out polygon fitting on the cultural relic point cloud fragments, carrying out second numbering, and randomly generating initial position matrixes of the cultural relic point cloud fragments and the cultural relic point cloud polygons;
and S140, calculating Wen Wudian cloud fragment intervals and overlapping volumes among the cultural relic point cloud polygons by adopting a peak inundation search algorithm, updating an initial position matrix if a calculation result has a better solution, iterating for a plurality of times to obtain an optimal position matrix, and splicing the global Wen Wudian cloud fragments.
In the process of the point cloud splicing method of the cultural relic fragments, acquiring and optimizing data of Wen Wudian cloud fragments, and establishing Wen Wudian cloud fragment data sets; training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model, realizing automatic segmentation of Wen Wudian cloud fragments, greatly reducing the splicing difficulty and eliminating calculation of non-spliced Wen Wudian cloud fragments; performing first numbering on the cultural relic point cloud fragments, performing polygon fitting on the cultural relic point cloud fragments, performing second numbering, randomly generating initial position matrixes of the cultural relic point cloud fragments and the cultural relic point cloud polygons, and improving calculation efficiency; and calculating the distance between Wen Wudian cloud fragments and the overlapping volume between the cultural relic point cloud polygons by adopting a peak inundation search algorithm, updating an initial position matrix if a calculation result has a better solution, repeatedly obtaining an optimal position matrix, splicing global Wen Wudian cloud fragments, introducing the peak inundation search algorithm into the global splicing of the cultural relic point cloud fragments, updating the initial position matrix according to the calculation result by calculating the point cloud distance of Wen Wudian cloud fragments and the overlapping volume between the cultural relic point cloud polygons, repeatedly obtaining an optimal solution, completing the splicing of global Wen Wudian cloud fragments, effectively obtaining the optimal solution of Wen Wudian cloud fragment splicing, and improving the accuracy of the cultural relic point cloud splicing result.
In the embodiment of the present invention, step S110 includes the following steps:
s1102, acquiring data of Wen Wudian cloud fragments by using a handheld three-dimensional laser scanner;
s1104, removing data noise of Wen Wudian cloud fragments by using point cloud processing software, deleting data in-vitro orphan points of Wen Wudian cloud fragments, unifying data density and format of Wen Wudian cloud fragments, calibrating data residual surfaces and non-residual surfaces of Wen Wudian cloud fragments, generating new cultural relic point cloud fragments by enhancing Wen Wudian cloud fragments, obtaining cultural relic point cloud fragments with reduced noise and data quantity, enabling the reserved cultural relic point cloud fragments to be closer to the surface of a real object, and improving the splicing efficiency of Wen Wudian cloud fragments;
because the method has high adaptability to sparse Wen Wudian cloud fragments, downsampling can be properly performed when Wen Wudian cloud data are randomly extracted, a training set, a test set and a verification set for deep learning are manufactured, and the calculated amount is reduced to the greatest extent on the basis of guaranteeing Wen Wudian cloud fragment characteristics.
In the embodiment of the present invention, step S120 includes the following steps:
s1202, setting up a Pointernet++ network model framework;
and S1204, training a data training set of the cultural relic point cloud fragments by using a Pointernet++ network to obtain a Wen Wudian cloud fragment deep learning segmentation model.
When the Pointet++ network model is used for training, the incomplete surface and the non-incomplete surface of the Wen Wudian cloud fragments can be effectively identified, the influence of the non-incomplete surface of the Wen Wudian cloud fragments on the splicing can be effectively reduced in the splicing process, and the calculated amount is greatly reduced.
In the embodiment of the present invention, after step S1204, the method further includes the following steps:
s1206, evaluating the deep learning segmentation model by using semantic segmentation evaluation index (miou) and accuracy (actuacy), and visually analyzing the effect of the deep learning segmentation model by evaluating the miou and the actuality.
In the embodiment of the present invention, step S130 includes the following steps:
s1302, carrying out first numbering on global to-be-matched Wen Wudian cloud fragments, wherein the first numbering is P respectively 1 、P 2 、P 3 、…、P N Judging the distance between any two cloud fragments to be matched Wen Wudian and a preset fragment distance threshold F, if the distance between any two cloud fragments to be matched is smaller than the preset fragment distance threshold F, performing polygon fitting on the two fragments, co-fitting N Wen Wudian cloud polygons, and performing second numbering on cultural relic point cloud polygons, wherein the distances between any two cloud fragments to be matched and the preset fragment distance threshold F are D respectively 1 、D 2 、D 3 、…、D N ；
S1304, generating G N multiplied by 6 initial position matrixes randomly after translational and rotational changes of N cultural relic point cloud fragments in the directions of x, y and z.
In the embodiment of the present invention, step S140 includes the following steps:
step S1401, a step matrix step is obtained, G initial position matrixes generated randomly are used as initial data of a peak inundation searching method, step distances of the initial position matrixes are calculated, and the step distances are moved along positive and negative directions by the initial position matrixes to obtain step position matrixes;
step S1403, updating the matrix, namely respectively calculating the point cloud distance and the overlapping volume according to the initial position matrix and the step matrix, and updating the initial position matrix into the step position matrix if the point cloud distance and the overlapping volume calculated according to the step matrix are better;
step S1405, a loop step, and a loop iteration step of solving the step matrix and the update matrix step until the initial position matrix is updated.
And randomly generating an initial position matrix, calculating the stepping distance of the initial position matrix, moving the stepping distance to obtain the stepping position matrix, determining whether to update the initial position matrix by judging whether the cloud space and the overlapping volume are better, iterating for a plurality of times until the initial position matrix is updated, and completing global point cloud splicing.
In the embodiment of the present invention, step S1401 includes the following steps:
the initial stepping distance of the initial position matrix is calculated as follows:
l p ＝L/2；
wherein L is the distance between adjacent initial position matrixes, and L p An initial stepping distance for a p-th initial position matrix;
and moving each initial position matrix along the positive and negative directions of the matrix by an initial stepping distance to obtain a stepping position matrix.
In the embodiment of the present invention, step S1403 includes the following steps:
calculating according to the initial position matrix to obtain a first point cloud distance L of the nearest point between each cultural relic point cloud fragment and other cultural relic point cloud fragments ab Calculating to obtain a first overlapped volume C of each cultural relic point cloud polygon and other cultural relic point cloud polygons mn ；
According to a step position matrixCalculating to obtain a second point cloud distance L 'of the closest point between each cultural relic point cloud fragment and other cultural relic point cloud fragments' ab Calculating to obtain a second overlapping volume C 'of each cultural relic point cloud polygon and other cultural relic point cloud polygons' mn ；
Comparing absolute value L of first point cloud distance by adopting non-dominant sorting method ab Absolute value of L and second point cloud distance L' ab Absolute value of first overlap volume |c mn Absolute value of i and second overlap volume |c' mn The magnitude of the i is such that,
if any |L is present ab I is greater than its corresponding L' ab I, and there is any |C mn I is greater than its corresponding |c' mn I, the initial position matrix is updated to be L' ab I and C' mn A step position matrix corresponding to the minimum;
if any |L is present ab The I is smaller than its corresponding I L' ab |, and/or there is any |C mn The I is smaller than its corresponding I C' mn Setting a step-down coefficient alpha in the next cycle, wherein the step-down coefficient alpha set each time is smaller than the step-down coefficient alpha of the previous cycle, the step-down coefficient alpha gradually approaches 0, and updating the step distance l 'of the next cycle of the initial position matrix' p The method comprises the following steps:
l’ p ＝α×l p ；
wherein (0 < alpha < 1). The initial position matrix needs to be updated according to the point cloud distance and the overlapping volume judgment, so that the stepping distance is continuously reduced to obtain the optimal stepping position matrix, updating of the stepping position matrix is guaranteed, and meanwhile the globally and finely of searching nearby the local optimal stepping position matrix in the searching process are guaranteed.
In the embodiment of the present invention, step S1403 further includes the following steps:
distance L between first point clouds ab Compared with a preset interval threshold H, if L ab If the number is greater than H, the two part of object point cloud fragments are considered to be non-adjacent Wen Wudian cloud fragments, the number is changed into ≡, and the splicing is not performed any more, so that the non-adjacent Wen Wudian cloud fragments are effectively reducedThe impression of the tile to the stitching process reduces the amount of computation.
In the embodiment of the present invention, step S1405 further includes the following steps:
setting a first water line W 1 And a second water line W 2 The method comprises the following steps:
W 1 ＝(L abmax -L abmin )×i/iter；W 2 ＝(C mnmax -C mnmin )×i/iter；
wherein L is abmax And L is equal to abmin C is the maximum value and the minimum value of the first point cloud distance in the current iteration respectively mnmax And C mnmin Respectively obtaining the maximum value and the minimum value of the first overlapped volume in the current iteration, wherein i is the current iteration number, and iter is the total iteration number;
the steps 1401 and 1403 are iterated, a flooding interval J is set, and each iteration J is performed with one flooding operation, wherein the flooding operation comprises: deleting the first point cloud space smaller than the first water line W 1 Deleting the corresponding initial position matrix, wherein the first overlapping volume is smaller than the second water level line W 2 Setting water level warning quantity A according to the corresponding initial position matrix, and randomly generating the initial position matrix if the quantity of the initial position matrix is smaller than A; setting a first water line, a second water line and a submerged interval, and gradually removing a disadvantaged solution and a locally optimal solution in an iterative process, so that the calculation process is greatly simplified, and the workload of a computer is reduced; setting the number of water level warning lines, adding random solutions into the coordinate set in a random generation mode, and ensuring the easy convergence, global performance, stability and diversity of the iterative process;
and after the loop iteration is completed, an optimal position matrix is obtained, and global Wen Wudian cloud fragments are spliced.
The principles of operation or methods of use of embodiments of the present invention are further described below with reference to a specific example:
the embodiment of the invention provides a peak inundation search algorithm, which has the advantages of high solving speed, good convergence, good fault tolerance and difficult local convergence, can play a good solving effect in the large-scale data solving process, has extremely high adaptability to sparse point cloud splicing, and comprises the following steps:
s210, acquiring a multi-element function coordinate set to be solved;
s220, calculating the stepping distance of each coordinate point in the multi-function coordinate set, and moving each coordinate point by the stepping distance along the positive and negative directions to obtain the stepping point;
s230, setting up a solving function, obtaining a first solution according to the coordinate points and the solving function, obtaining a second solution according to the stepping points and the solving function, comparing the first solution with the second solution, selecting points corresponding to more optimal solutions to replace points in the coordinate set, and updating the coordinate set;
s240, circularly iterating to obtain a stepping point step and an updating step until the updating of the coordinate set is completed.
In the embodiment of the present invention, S210 includes the following steps:
acquiring an N-dimensional to-be-solved multiple function; selecting a coordinate point on each dimension at intervals of a preset distance, and selecting M coordinate points on each dimension to establish a coordinate set P, P= { P ij I e N, j e M; wherein p is ij A j coordinate point of the i-th dimension; the method for equidistantly generating the coordinate points expands the searching range of the function to be solved to the greatest extent, ensures the global property of the searching process and prevents the problem of local convergence of the algorithm.
In the embodiment of the present invention, S220 includes the following steps:
the initial step distance of each coordinate point is calculated as,
l ij ＝L/2，(i∈N，j∈M，)
wherein L is the distance between adjacent points in the same dimension, L ij An initial stepping distance of a j coordinate point of an i-th dimension;
moving each coordinate point in the coordinate set along positive and negative directions of the dimension of the multiple functions by a stepping distance to obtain a stepping point set corresponding to each coordinate point,
S＝{s ijk ，i∈N，j∈M，k∈(-N，N)}
wherein s is ijk Is the stepping point corresponding to the j coordinate point of the i-th dimension.
In the embodiment of the present invention, S230 includes the following steps:
setting up a solving function, obtaining a first solution according to the coordinate points and the solving function, setting up a coordinate point solution set Y, obtaining a second solution from the stepping points and the solving function, and setting up a stepping point solution set Y S ；
If solve set Y S The second solution of a certain stepping point in the solution set Y is larger than the first solution of the corresponding coordinate point of the stepping point to be a better solution, the coordinate set is updated,
wherein Y is ij Substituting the j coordinate point of the i-th dimension into a first solution obtained by solving a function, Y Sij Substituting a step point corresponding to the j coordinate point of the i dimension into a second solution obtained by the solving function;
if Y is present Sij Greater than Y ij Replacing the coordinate point in the coordinate set with the corresponding step point s when the second solution is maximum in the corresponding step points ije ；
If there is no Y Sij Greater than Y ij Setting the step-down factor a in the next cycle, updating the step-down distance of the point,
l ij ’＝α×l ij (α＜1)；
if solve set Y S The second solution of a certain stepping point in the solution set Y is smaller than the first solution of the coordinate point corresponding to the stepping point in the solution set Y, the coordinate set is updated,
wherein, if Y is present Sij Less than Y ij Replacing the coordinate point in the coordinate set with the corresponding step point s when the second solution is the smallest in the corresponding step points ijf ；
If there is no Y Sij Less than Y ij Setting the step-down factor alpha in the next cycle, updating the step-down distance of the point,
l ij ’＝α×l ij (α＜1)；
the method is suitable for two situations that the larger and the smaller the solution are, the better.
In the embodiment of the present invention, S240 includes the following steps:
the coordinate water level line is set as a water level line,
W＝(Ymax-Ymin)×i/iter，
wherein Ymax and Ymin are maximum and minimum values of a first solution obtained by a coordinate point and a solving function in the current iteration, i is the current iteration number, and iter is the total iteration number;
and (3) performing loop iteration, setting a flooding interval J, and performing one flooding operation every iteration J generation, wherein the flooding operation is to delete coordinate points corresponding to the first solution smaller than the water level line W, set the water level warning quantity A, and add new coordinate points to the coordinate set through random generation if the quantity of the coordinate points is smaller than the A.
The invention provides a peak inundation search algorithm application case, in particular to a method for splicing object point cloud fragments by using coordinate points through the algorithm, which comprises the following steps:
step one, analyzing the solving function and generating a coordinate set
After translational and rotational operations are performed on the cloud fragments Cs to be spliced Wen Wudian, the distance between the cloud fragments Cs to be spliced Wen Wudian and the target Wen Wudian cloud fragments Ct is minimized, the dimension n=6, specifically, translational independent variables x, y and z in a three-dimensional space and rotational independent variables rx, ry and rz in the three-dimensional space, and then the independent variable ranges are as follows:
wherein:
x s.max is the maximum value of x in Cs;
x s.min is the minimum value of x in Cs;
x t.max is the maximum value of x in Ct;
x t.min is the minimum value of x in Ct;
y s.max is the most y in CsA large value;
y s.min is the minimum value of y in Cs;
y t.max is the maximum value of y in Ct;
y t.min is the minimum value of y in Ct;
z s.max is the maximum value of z in Cs;
z s.min is the minimum value of z in Cs;
z t.max is the maximum value of z in Ct;
z t.min is the minimum value of z in Ct;
the variable ranges of the translational independent variables x, y and z are obtained by subtracting the minimum values of x, y and z in Ct from the maximum values of x, y and z in Cs respectively, and comparing the absolute values of the values obtained by subtracting the minimum values of x, y and z in Cs from the maximum values of x, y and z in Ct respectively, the corresponding values of Cs and Ct with large absolute values are selected as the translational independent variable ranges, the variable ranges of the rotational independent variables rx, ry and rz are 0-360 degrees, and the coordinate set P of the function to be solved is based on the characteristics of a splicing algorithm:
a＝1，...，D x ，b＝1，...，D y ，k＝1，...，D z ，o＝1，...，D rx ，m＝1，...，D ry ，n＝1，...，D rz }
wherein D is x ，D y ，D z ，D rx ，D ry ，D rz The case selects D for the dividing number of each independent variable x ＝D y ＝D z ＝D rx ＝D ry ＝D rz =3, the coordinate set has 729 coordinate points in total.
Step two objective function iterative solution
Setting a solving function to be the Euclidean distance between the target Wen Wudian cloud fragment searched by the kdtree and the transformed spliced Wen Wudian cloud fragment, taking the coordinate points as transformed coordinates into the solving function to obtain a first solution, and establishing a coordinate point solution set to finish the coordinate set solving process.
After the coordinate set is solved, a stepping module is entered, an initial stepping distance is firstly set and calculated, a stepping reduction coefficient alpha=0.85 is set, coordinate points are further generated into corresponding stepping points, the dimension N=6 of the corresponding stepping points, each coordinate point corresponds to 12 stepping points, and a second solution is obtained according to a solving function and the stepping points. If the first solution of a certain coordinate point is larger than the second solution after solving all corresponding stepping points, reserving the points in the coordinate set and reducing the stepping coefficient to alpha multiplied by l ij The method comprises the steps of carrying out a first treatment on the surface of the If the first solution of a certain coordinate point is smaller than the second solution of one or more corresponding stepping points, replacing the coordinate point in the coordinate set with the stepping point with the smallest corresponding second solution.
And setting a submerged water line W= (Ymax-Ymin) multiplied by i/iter in the iteration process, wherein the submerged interval J=5, the water level warning line is 100, and the iteration number is 500. After the iteration is completed, an optimal coordinate set is obtained.
Fig. 2 is a schematic diagram of a point cloud space curve of a point cloud fragment of an cultural relic in an iterative process, where the point cloud space of the point cloud fragment of the cultural relic gradually decreases and gradually approaches to an optimal solution as the number of iterations increases.
Fig. 3 is a schematic diagram of a peak inundation search algorithm substituted by coordinate points, specifically, a and B are first solutions corresponding to the peak inundation search algorithm carried by an example coordinate point, and a shaded portion is a inundation portion in an iterative process, where the example coordinate point a searches for a corresponding step point a 1 、A 2 、A 3 And A 4 And the solution of the stepping point is not submerged, which proves that the stepping point is searched for a better stepping point, and the example coordinate point B is searched for a corresponding stepping point B 1 、B 2 、B 3 、B 4 And its step point B after the step distance is reduced 1 ’、B 2 ’、B 3 ’、B 4 ' but the solutions of their corresponding stepping points are all submerged, proving that they have not searched for better stepping points.
The embodiment of the invention provides a method for splicing character point cloud fragments by taking certain terracotta warriors and horses fragments as character point cloud fragments, which comprises the following steps:
step one: setting initial parameters
Pointet++ module: batch size=50, iteration number=150, initial learning rate set=0.0001, training set number 12576, test set number 2000, validation set 2000; peak flooding search algorithm part: the independent variable range is divided into 3 parts, the iteration number iter is 500 generations, the initial step size= (adjacent point spacing/2), the step size reduction coefficient = 0.8, the flooding interval J is 3 generations, and the number of flooding guard lines a is 100.
Training the Pointernet++ model,
the training set is imported into a Pointernet++ model training module for training, the hardware environment is a Win 1064-bit system, and the training set is configured into an RTX A4000 display card (16 GB) and E5-2686 [email protected],32GB memory. The development environment was python 3.7.13 and the pytorch version was 1.12.1.
The training results of the terracotta warriors and horses are shown in table 1, the point cloud fragments of the cultural relics to be spliced are passed through the split model trained by Pointernet++, and the split results of a certain Wen Wudian cloud fragment of the split model are shown in fig. 4A and fig. 4B.
TABLE 1 Pointet++ model training results
The accuracy of the training set reaches 0.91935, the accuracy of the testing set reaches 0.920045, the optimal accuracy reaches 0.92057, and the semantic evaluation segmentation index reaches 0.838543.
Third, cultural relic point cloud fragment splicing based on peak inundation search algorithm
The corresponding Wen Wudian cloud fragment splice is performed by Python,
# set rotation matrix function
def rotate(pointcloud)
Turn_matix＝[cosy*cosz,cosy*sinz,-siny],
[sinx*siny*cosz-cosx*sinz,sinx*siny*sinz+cosx*cosz,sinx*cosy],
[cosx*siny*cosz+sinx*sinz,cosx*siny*sinz-sinx*cosz,cosx*cosy]
An initial coordinate set is generated according to the independent variable dividing range, and 729 initial points are included in the coordinate set.
Iteration is carried out #
for iin range(iter_number)
# calculate coordinate set fitness
# generating coordinate set step points
Calculation of step Point fitness
# determine whether to update coordinate set
if initial_fitness>step_fitness:
Step size of # reduction
step＝step*step_lenth
else
initial_point＝step_point
# submerged operation
if iter_number％3＝＝0
Deleting points with fitness less than the water line
# submerged warning line judgment
if coordinate set number <100
Randomly generating initial positions (100-number of coordinate sets)
# derive the optimal rotation matrix.
And step four, iterating by adopting a peak inundation search algorithm to finish overall multi-Wen Wudian cloud fragment splicing, wherein fig. 5 is a schematic diagram of a splicing result of partial object point cloud fragments provided by the embodiment of the invention, and fig. 6 is a schematic diagram of a splicing result of overall Wen Wudian cloud fragments provided by the embodiment of the invention.
According to the point cloud splicing method for the cultural relic fragments, provided by the embodiment of the invention, the automatic segmentation of the point cloud fragments can be achieved by training an artificial intelligent deep learning model, so that the splicing difficulty is greatly reduced, and non-spliced point cloud computing is eliminated; the algorithm is introduced into the global splicing process of corresponding fragments and multiple fragments, so that the full-flow intellectualization and automation of point cloud splicing are finally realized, and the effect is remarkable.
In the description of the present invention, it should be noted that the azimuth or positional relationship indicated by the terms "upper", "lower", "front", "horizontal", etc. are based on the azimuth or positional relationship shown in the drawings, and are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or element referred to must have a specific azimuth, be constructed and operated in a specific azimuth, and thus should not be construed as limiting the present invention.
In the description of the present invention, it should be noted that, unless explicitly stated and limited otherwise, the term "mounted" should be interpreted broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.

Claims (10)

1. The point cloud splicing method for the cultural relic fragments is characterized by comprising the following steps of:
acquiring and optimizing data of Wen Wudian cloud fragments, and establishing Wen Wudian cloud fragment data sets;
training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model;
performing first numbering on the cultural relic point cloud fragments, performing polygon fitting on the cultural relic point cloud fragments, performing second numbering, and randomly generating initial position matrixes of the cultural relic point cloud fragments and the cultural relic point cloud polygons;
and calculating Wen Wudian cloud fragment intervals and overlapping volumes among the cultural relic point cloud polygons by adopting a peak inundation search algorithm, updating an initial position matrix if a calculation result has a better solution, iterating for a plurality of times to obtain an optimal position matrix, and splicing the global Wen Wudian cloud fragments.
2. The method for point cloud stitching of cultural relic fragments according to claim 1, wherein the steps of obtaining and optimizing data of Wen Wudian cloud fragments and establishing Wen Wudian cloud fragment data sets include the following steps:
acquiring Wen Wudian cloud fragment data;
removing data noise of Wen Wudian cloud fragments by using point cloud processing software, deleting data in-vitro orphan points of Wen Wudian cloud fragments, unifying data density and format of Wen Wudian cloud fragments, calibrating data incomplete surfaces and non-incomplete surfaces of Wen Wudian cloud fragments, and generating new cultural relic point cloud fragments according to Wen Wudian cloud fragment enhancement;
the Wen Wudian cloud data is randomly extracted, and a training set, a testing set and a verification set for deep learning are manufactured.
3. The method for point cloud splicing of cultural relic fragments according to claim 2, wherein training the data of the cultural relic point cloud fragments to obtain a deep learning segmentation model comprises the following steps:
setting up a Pointernet++ network model frame;
training the data training set of the cultural relic point cloud fragments by using the Pointernet++ network to obtain a Wen Wudian cloud fragment deep learning segmentation model.
4. The method for point cloud splicing of cultural relic fragments according to claim 3, wherein the training of the data training set of cultural relic point cloud fragments by using a Pointnet++ network, after obtaining a Wen Wudian cloud fragment deep learning segmentation model, further comprises the following steps:
and evaluating the deep learning segmentation model by adopting semantic segmentation evaluation indexes and accuracy.
5. The method for splicing the point clouds of the cultural relic fragments according to claim 1, wherein the method for carrying out first numbering on the cultural relic point cloud fragments, carrying out polygon fitting on the cultural relic point cloud fragments and carrying out second numbering, and randomly generating initial position matrixes of the numbered cultural relic point cloud fragments and the cultural relic point cloud polygons comprises the following steps:
the global Wen Wudian cloud fragments to be matched are numbered first and are P respectively 1 、P 2 、P 3 、…、P N Judging the distance between any two cloud fragments to be matched Wen Wudian and a preset fragment distance threshold F, if the distance between any two cloud fragments to be matched is smaller than the preset fragment distance threshold F, performing polygon fitting on the two fragments, co-fitting N Wen Wudian cloud polygons, and performing second numbering on cultural relic point cloud polygons, wherein the distances between any two cloud fragments to be matched and the preset fragment distance threshold F are D respectively 1 、D 2 、D 3 、…、D N ；
And randomly generating G N multiplied by 6 initial position matrixes after the translation and rotation of N cultural relic point cloud fragments in the x, y and z directions are changed.
6. The method for point cloud splicing of cultural relic fragments according to claim 5, wherein the peak inundation search algorithm is adopted to calculate Wen Wudian the distance between the cloud fragments and the overlapping volume between the point cloud polygons of the cultural relic, if the calculation result has a better solution, the initial position matrix is updated, the optimal position matrix is obtained by multiple iterations, and the global Wen Wudian cloud fragments are spliced, comprising the following steps:
step of solving a stepping matrix, namely taking G initial position matrixes generated randomly as initial data of a peak inundation searching method, calculating stepping distances of the initial position matrixes, and moving each initial position matrix along positive and negative directions by the stepping distances to obtain a stepping position matrix;
updating the matrix, namely respectively calculating the point cloud distance and the overlapping volume according to the initial position matrix and the stepping matrix, and updating the initial position matrix into the stepping position matrix if the point cloud distance and the overlapping volume calculated according to the stepping matrix are better;
and (3) a loop step, namely, a step matrix step and a matrix updating step are obtained through loop iteration until the initial position matrix is updated.
7. The method for point cloud stitching of cultural relic fragments according to claim 6, wherein the step of solving a step matrix comprises the steps of:
the initial stepping distance of the initial position matrix is calculated as follows:
l p ＝L/2；
wherein L is the distance between adjacent initial position matrixes, and L p An initial stepping distance for a p-th initial position matrix;
and moving each initial position matrix along the positive and negative directions of the matrix by an initial stepping distance to obtain a stepping position matrix.
8. The method for point cloud stitching of cultural relic fragments according to claim 7, wherein the step of updating the matrix comprises the steps of:
calculating according to the initial position matrix to obtain a first point cloud distance L of the nearest point between each cultural relic point cloud fragment and other cultural relic point cloud fragments ab Calculating to obtain a first overlapped volume C of each cultural relic point cloud polygon and other cultural relic point cloud polygons mn ；
Calculating a second point cloud distance L 'of the nearest point between each cultural relic point cloud fragment and other cultural relic point cloud fragments according to the stepping position matrix' ab Calculating to obtain a second overlapping volume C 'of each cultural relic point cloud polygon and other cultural relic point cloud polygons' mn ；
Comparing absolute value L of first point cloud distance by adopting non-dominant sorting method ab Absolute value of L and second point cloud distance L' ab Absolute value of first overlap volume |c mn Absolute value of i and second overlap volume |c' mn The magnitude of the i is such that,
if any |L is present ab I is greater than its corresponding L' ab I, and there is any |C mn I is greater than its corresponding |c' mn I, thenThe initial position matrix is updated to be |L '' ab I and C' mn A step position matrix corresponding to the minimum;
if any |L is present ab The I is smaller than its corresponding I L' ab |, and/or there is any |C mn The I is smaller than its corresponding I C' mn Setting a step-down coefficient alpha in the next cycle, wherein the step-down coefficient alpha set each time is smaller than the step-down coefficient alpha of the previous cycle, the step-down coefficient alpha gradually approaches 0, and updating the step distance l 'of the next cycle of the initial position matrix' p The method comprises the following steps:
l’ p ＝α×l p ；
wherein (0 < alpha < 1).
9. The method for point cloud stitching of cultural relic fragments according to claim 8, wherein the step of updating the matrix further comprises the steps of:
distance L between first point clouds ab Compared with a preset interval threshold H, if L ab And if the number is greater than H, the two part of the object point cloud fragments are considered to be non-adjacent Wen Wudian cloud fragments, the number of the two part of the object point cloud fragments is changed to be ≡, and the two part of the object point cloud fragments are not spliced.
10. The method for point cloud stitching of cultural relic fragments according to claim 8, wherein the cycling step further comprises the steps of:
setting a first water line W 1 And a second water line W 2 The method comprises the following steps:
W 1 ＝(L abmax -L abmin )×i/iter；W 2 ＝(C mnmax -C mnmin )×i/iter；
wherein L is abmax And L is equal to abmin C is the maximum value and the minimum value of the first point cloud distance in the current iteration respectively mnmax And C mnmin Respectively obtaining the maximum value and the minimum value of the first overlapped volume in the current iteration, wherein i is the current iteration number, and iter is the total iteration number;
step matrix step and update matrix step are calculated by loop iteration, a flooding interval J is set, and each iteration J is carried out for one generationA secondary flooding operation, the flooding operation comprising: deleting the first point cloud space smaller than the first water line W 1 Deleting the corresponding initial position matrix, wherein the first overlapping volume is smaller than the second water level line W 2 Setting water level warning quantity A according to the corresponding initial position matrix, and randomly generating the initial position matrix if the quantity of the initial position matrix is smaller than A;
and after the loop iteration is completed, an optimal position matrix is obtained, and global Wen Wudian cloud fragments are spliced.
CN202311428559.XA 2023-10-31 2023-10-31 Point cloud splicing method for cultural relic fragments Pending CN117495674A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311428559.XA CN117495674A (en) 2023-10-31 2023-10-31 Point cloud splicing method for cultural relic fragments

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311428559.XA CN117495674A (en) 2023-10-31 2023-10-31 Point cloud splicing method for cultural relic fragments

Publications (1)

Publication Number Publication Date
CN117495674A true CN117495674A (en) 2024-02-02

Family

ID=89684023

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311428559.XA Pending CN117495674A (en) 2023-10-31 2023-10-31 Point cloud splicing method for cultural relic fragments

Country Status (1)

Country Link
CN (1) CN117495674A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6398839A (en) * 1986-10-15 1988-04-30 Hitachi Ltd Method and device for recording and reproducing optical information
US20220254095A1 (en) * 2021-02-03 2022-08-11 Electronics And Telecommunications Research Institute Apparatus and method for searching for global minimum of point cloud registration error
CN114998103A (en) * 2022-05-27 2022-09-02 中国计量大学 Point cloud cultural relic fragment three-dimensional virtual splicing method based on twin network
CN115509239A (en) * 2022-11-19 2022-12-23 中国人民解放军陆军装甲兵学院 Unmanned vehicle route planning method based on air-ground information sharing

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6398839A (en) * 1986-10-15 1988-04-30 Hitachi Ltd Method and device for recording and reproducing optical information
US20220254095A1 (en) * 2021-02-03 2022-08-11 Electronics And Telecommunications Research Institute Apparatus and method for searching for global minimum of point cloud registration error
CN114998103A (en) * 2022-05-27 2022-09-02 中国计量大学 Point cloud cultural relic fragment three-dimensional virtual splicing method based on twin network
CN115509239A (en) * 2022-11-19 2022-12-23 中国人民解放军陆军装甲兵学院 Unmanned vehicle route planning method based on air-ground information sharing

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
YU ZHONG.ET AL: "Point Cloud Splicing Based on 3D-Harris Operator", IEEE, 29 June 2022 (2022-06-29) *
俞浩;高飞;: "基于特征向量的点云配准方法研究", 合肥工业大学学报(自然科学版), no. 07, 28 July 2020 (2020-07-28) *
孙海鹏;余伟巍;席平;: "基于Level Set的交互式快速分割算法", 工程图学学报, no. 03, 15 June 2011 (2011-06-15) *

Similar Documents

Publication Publication Date Title
CN108594816B (en) Method and system for realizing positioning and composition by improving ORB-SLAM algorithm
CN109903383B (en) Method for accurately positioning coal mining machine in three-dimensional model of working face coal seam
CN110992473B (en) Tree branch modeling method and system based on vehicle-mounted laser scanning point cloud
CN108765475B (en) Building three-dimensional point cloud registration method based on deep learning
Wang et al. A bayesian approach to building footprint extraction from aerial lidar data
CN104573705A (en) Clustering method for building laser scan point cloud data
CN106157309A (en) A kind of airborne LiDAR ground point cloud filtering method based on virtual Seed Points
CN103295239A (en) Laser-point cloud data automatic registration method based on plane base images
CN104867137A (en) Improved RANSAC algorithm-based image registration method
CN108537263B (en) Grid map fusion method based on maximum public subgraph
Wei et al. A concentric loop convolutional neural network for manual delineation-level building boundary segmentation from remote-sensing images
CN115546116A (en) Method and system for extracting and calculating spacing of discontinuous surface of fully-covered rock mass
CN116559928A (en) Pose information determining method, device and equipment of laser radar and storage medium
CN115880713A (en) Method for automatically searching and constructing building property body based on semantic improvement
CN115937552A (en) Image matching method based on fusion of manual features and depth features
Wang et al. Roof plane segmentation from lidar point cloud data using region expansion based l 0 gradient minimization and graph cut
CN116091706B (en) Three-dimensional reconstruction method for multi-mode remote sensing image deep learning matching
CN113409332A (en) Building plane segmentation method based on three-dimensional point cloud
Li et al. Exploring label probability sequence to robustly learn deep convolutional neural networks for road extraction with noisy datasets
CN117495674A (en) Point cloud splicing method for cultural relic fragments
Li et al. An improved Douglas-Peucker algorithm for fast curve approximation
Liu et al. Adaptive algorithm for automated polygonal approximation of high spatial resolution remote sensing imagery segmentation contours
CN116309026A (en) Point cloud registration method and system based on statistical local feature description and matching
Campagnolo et al. Fully Automated Scan-to-BIM Via Point Cloud Instance Segmentation
Zhao et al. An automatic density clustering segmentation method for laser scanning point cloud data of buildings

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination