CN114092433B - Apparent defect detection method based on triangle centroid optimization - Google Patents
Apparent defect detection method based on triangle centroid optimization Download PDFInfo
- Publication number
- CN114092433B CN114092433B CN202111355733.3A CN202111355733A CN114092433B CN 114092433 B CN114092433 B CN 114092433B CN 202111355733 A CN202111355733 A CN 202111355733A CN 114092433 B CN114092433 B CN 114092433B
- Authority
- CN
- China
- Prior art keywords
- point
- triangle
- defect
- centroid
- points
- 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.)
- Active
Links
- 230000007547 defect Effects 0.000 title claims abstract description 90
- 238000001514 detection method Methods 0.000 title claims abstract description 34
- 238000005457 optimization Methods 0.000 title claims abstract description 18
- 238000000034 method Methods 0.000 claims abstract description 43
- 238000004364 calculation method Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 9
- 238000004458 analytical method Methods 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 230000002950 deficient Effects 0.000 claims description 5
- 238000012847 principal component analysis method Methods 0.000 claims description 2
- 230000001133 acceleration Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 230000002431 foraging effect Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000001303 quality assessment method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30132—Masonry; Concrete
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2210/00—Indexing scheme for image generation or computer graphics
- G06T2210/12—Bounding box
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an apparent defect detection method of a civil engineering structure, in particular to an apparent defect detection method based on triangle centroid optimization, which uses centroid points of triangle surface elements as equivalent substitutes of the triangle surface elements, and the distance calculation between a point cloud model and a grid model is equivalent to the distance calculation between the point cloud model and centroid point clouds, so that the search of neighbor points can be completed by using a K-d Tree and other search acceleration structures, the time complexity of an algorithm is effectively reduced on the premise of not losing detection precision, and the improved C2M method is more feasible.
Description
Technical Field
The invention relates to an apparent defect detection method of a civil engineering structure, in particular to an apparent defect detection method based on triangle centroid optimization.
Background
The apparent defect detection (surface defect detection) is one of important means for evaluating the quality (quality assessment, QA) of civil engineering structures such as bridges, tunnels and the like, is widely applied to various scenes such as engineering acceptance and inspection maintenance, and aims to detect apparent defects of the civil engineering structures so as to predict potential safety accident risks. Many scholars have conducted intensive exploration in the field, and traditional apparent defect detection means such as two-dimensional images, infrared rays and ultrasonic waves have a mature system flow, and the apparent defect detection flow using an emerging laser point cloud as a data medium is in progress, and a great improvement space still exists in a plurality of links.
The main method of apparent defect detection using laser point cloud as data medium is to compare the building information model (building information modeling, BIM) of the target civil engineering structure with the currently collected laser point cloud model. BIM often comes from a design model (such as a CAD model) of a civil engineering structure, or uses an early-collected laser point cloud model as an approximate substitute, and the core method is to compare and analyze digital models obtained in different periods to obtain differences among the models, so as to determine the deformation area of the target civil engineering structure, namely the area with potential safety hazards.
In the aspect of comparing two-stage laser point clouds and determining deformation areas, a method for calculating a corresponding distance (corresponding distance) between a point cloud model and a mesh model (mesh) obtained by a three-dimensional surface reconstruction technology is proposed by a scholars, namely a traditional point cloud-to-mesh (C2M) distance calculation method, so that higher detection precision is obtained. However, the method adopts a double-layer circulation structure to respectively traverse each coordinate point in the point cloud model and each triangular surface element in the grid model, the time complexity is the product of the scale of the point cloud model and the grid model, the calculation cost is high, and the detection result is difficult to be given in ideal time, so that the method is not applied much in the actual production environment.
Disclosure of Invention
In order to solve the problems that the traditional C2M method is insufficient in timeliness and difficult to apply to practice, the invention provides the method which uses the centroid point of the triangular surface element as the equivalent substitute of the triangular surface element, and the distance calculation between the point cloud model and the grid model is equivalent to the distance calculation between the point cloud model and the centroid point cloud, so that the K-d Tree and other search acceleration structures can be used for completing the neighbor point search, the time complexity of the algorithm is effectively reduced on the premise of not losing the detection precision, and the improved C2M method is more feasible.
The invention aims to provide an apparent defect detection method based on triangle centroid optimization, which is realized by the following technical scheme:
(1) Data input, respectively acquiring a period of point cloud model at different time points for a given target civil engineering structure to be analyzed Second-phase point cloud model/>Grid model/>, obtained by three-dimensional surface reconstructionWherein/>Respectively, is a single data point in the point cloud,/>And (3) withThe number of the points contained in the point cloud is respectively;
(2) Deformation distance calculation, pair Each point P i in (1) is determined/>Triangle element M r,/>, with closest Euclidean distance to P i And this procedure is denoted as map/>Distance is denoted as/>Namely/>Each point and/>Euclidean distance between, i.e. representing/>A deformation distance of each point and its vicinity;
(3) Defect point set judgment, artificial setting of deformation distance threshold D, and determination of Satisfies/>Items of (2)Consider/>Corresponding Point set/>Defects exist near each point in (1), wherein N' 1<N1 is/>The number of points determined to be defective;
(4) Defect connected domain analysis using Euclidean clustering of point-to-point sets Performing connected domain analysis, and dividing the defect point set into a plurality of connected regions/>, according to Euclidean distanceWherein/>And mutually exclusive of each other, each/>All the points included in the map are a defect area where the civil engineering structure actually exists;
(5) Bounding box generation, point-to-point set sequence And calculating the direction bounding box OBB of each element, namely the defect area, wherein the center point of the OBB is the center point of the defect, and the size of the OBB is the size of the defect.
Further, in the step (1), theThree-dimensional surface reconstruction using greedy projection triangulation algorithm to obtain grid model/>
Further, in the step (2), theEach point P i in (1) is determined/>Triangle bin/>, closest to P i Euclidean distanceThe process specifically comprises the following steps:
(21) Extraction of Centroid point cloud/>For triangle element M k, if three vertices are recorded asThe centroid G k is calculated as:
Thereby can establish And/>Mapping f between; where f is both bijective and homojective, because each triangular bin M k and centroid G k are in one-to-one correspondence, and/>Each triangular surface element of the model has a corresponding centroid; by the above calculation, it can be known that f (M k)=Gk, and M k=f-1(Gk);
(22) Determining an alternative to the nearest triangle element, pair Establishing K-d Tree, pair/>At/>, P i K neighbor searching is performed to obtain K neighbor points/>Wherein/>A subscript sequence for the searched points; combining the mapping/>, in step (21)It can be seen that AND/>Corresponding triangle bin sequence/>This is the K triangular bins nearer to P i;
(23) Determining the nearest triangle surface element, and sequentially calculating P i and P i Euclidean distance between the two, and the distance obtains the minimum value/>Namely M r.
Further, in step (4), euclidean clustering is used for point-to-point setPerforming connected domain analysis, and dividing the defect point set into a plurality of connected regions/>, according to Euclidean distanceThe method specifically comprises the following steps:
(41) For a point T in space, a point set is established At/>, to be clusteredNeighbor searching is performed on a plurality of T neighbor points, and all points with the distance smaller than a certain threshold epsilon in the neighbor points are classified as/>
(42) SelectingAnd repeating the step 4.1 at the other point R' different from R until/>No further increases.
Further, in the step (22), K is more than or equal to 100.
Further, the deformation distance threshold D in the step (3) and the connected domain analysis threshold epsilon in the step (41) are adjusted according to the density degree and the size of the actually obtained point cloud model of the civil engineering structure, and D > zeta and epsilon > 2zeta are required to be met, wherein zeta is the point cloud resolution of the laser radar used for acquisition.
Further, the method comprises the steps of,
Further, in step (5), a principal component analysis method is used to calculate the directional bounding box of the defective region.
Compared with the prior art, the invention has the following advantages and beneficial effects:
(1) The invention introduces centroid point cloud as equivalent substitute of grid model, so that K-d Tree and other search acceleration structures can be used, and the process of traversing the point cloud model and the grid model in sequence in the traditional C2M method requiring a double-layer circulation structure is equivalently converted into a rapid search process using K-d Tree, thereby greatly reducing algorithm time consumption on the premise of not losing detection precision.
(2) The invention provides a complete civil engineering structure apparent defect detection and defect information extraction flow from data acquisition to final result output. The traditional method is usually stopped at a deformation distance calculation link, and the invention provides a follow-up connected domain analysis and bounding box generation link for defect information extraction, so that the method has completeness in a practical environment.
(3) The parameters (including the deformation distance threshold D (unit: m) in the step 3 and the connected domain analysis threshold epsilon in the step 4.1) related in the invention can be flexibly adjusted according to actual data conditions, so that the invention can cope with various data processing works and has better robustness.
Drawings
FIG. 1 is a flow chart of an apparent defect detection method based on triangle centroid optimization.
Fig. 2 is a schematic diagram of a mesh model (mesh) and a centroid point cloud structure thereof according to the present invention.
Fig. 3 is a schematic diagram showing the effect of calculating the deformation distance of the defect according to the present invention, wherein red represents that the deformation distance is the largest, and blue represents that the deformation distance is close to 0.
Fig. 4 is a schematic diagram showing the analysis effect of the defect connected domain according to the present invention, and different connected defect regions are marked by using different colors.
FIG. 5 is a schematic diagram showing the effect of generating a defect bounding box according to the present invention, and determining a direction bounding box (object-oriented minimum bounding box, OBB) for a plurality of defect connected regions.
Detailed Description
The present invention will be described in further detail with reference to examples, but embodiments of the present invention are not limited thereto.
As shown in fig. 1, the apparent defect detection method based on triangle centroid optimization comprises the following steps:
And step 1, inputting data. For a given target civil engineering structure to be analyzed, respectively acquiring a one-period point cloud model at different time points And second-phase point cloud model/>Considering that the laser point cloud is a set of points, the two-phase point cloud model can be further expressed as/>And/>Wherein/>Respectively, is a single data point in the point cloud,/>And/>The number of points included in the point cloud is respectively. In practical applications, the time interval between the two collection operations is generally long, so that the target object may generate an apparent defect during the time, for example, the target object is obtained by two periodic inspection collection of the same structure. Pair/>Three-dimensional surface reconstruction using a greedy projection triangulation algorithm (greedy projection) to obtain a mesh model/>Wherein/>For a single triangular surface element in the mesh model,/>Is the number of triangle primitives contained in the mesh model.
And 2, calculating the deformation distance. For a pair ofEach point P i in (1) is determined/>Triangle element M r,/>, closest to P i Euclidean distance (Euclidean distance)And this procedure is denoted as map/>Distance is recorded as Namely/>Each point and/>Euclidean distance between, i.e. representing/>And deformation distances in the vicinity thereof.
And 3, judging a defect point set. Manually setting a deformation distance threshold D, and determiningItem in (D i > D) is satisfiedConsider/>Corresponding Point set/>Defects exist near each point in (1), wherein N' 1<N1 is/>The number of points determined to be defective.
And 4, analyzing the defect connected domain. Clustering method for sample points in high-dimensional feature space by simulating pattern recognition field and using European cluster (Euclidean clustering) to point setPerforming connected domain analysis to divide the defect point set into a plurality of connected regions (point set sequence)/>, according to Euclidean distanceWherein/>And mutually exclusive of each other, each/>All points included in the map are a defect area where the civil engineering structure actually exists.
And 5, generating a bounding box. Sequences of pairs of point setsThe direction bounding box (object-oriented minimum bounding box, OBB) of each element (i.e., defect region) is calculated by using methods such as principal component analysis, and the center point of the OBB is regarded as the center point of the defect, and the size of the OBB is regarded as the size of the defect. Thus, the invention completes the work of apparent defect detection and defect information extraction of the civil engineering structure.
Further, in step2, the pairEach point P i in (1) is determined/>Triangle surface element with nearest middle Euclidean distance with P i The process specifically comprises the following steps:
Step 2.1, extraction Centroid point cloud/>For triangle element M k, if three vertices are recorded asThe centroid G k is calculated as:
Thereby can establish And/>Mapping f between. Obviously, f is both bijective and homojective, since each triangular bin M k and centroid G k are in one-to-one correspondence, and/>Each of the triangular primitives has a corresponding centroid. By the above calculation, f (M k)=Gk, and M k=f-1(Gk) can be found.
Step 2.2, determining an alternative to the nearest triangle element. For a pair ofBuild K-dTree, pair/>At/>, P i K neighbor searching is performed to obtain K neighbor points/>Wherein/>For searching the subscript sequence of the resulting points. Combining the mapping/>, in step 2.1It can be seen that AND/>Corresponding triangle bin sequence/>This is the K triangular bins closer to P i.
And 2.3, determining the nearest triangle element. Sequentially calculating P i andEuclidean distance between the two, and the distance obtains the minimum value/>Namely M r.
Further, in step 4, euclidean clustering is used for point-to-point setPerforming connected domain analysis to divide the defect point set into a plurality of connected regions (point set sequence)/>, according to Euclidean distanceThe method specifically comprises the following steps:
step 4.1, for a point T in space, establishing a point set At/>, to be clusteredNeighbor searching is performed on a plurality of T neighbor points, and all points with the distance smaller than a certain threshold epsilon in the neighbor points are classified as/>
Step 4.2, selectingAnother point T' than T repeats step 4.1 until/>No further increases.
Further, the accuracy of the defect detection and information extraction results can be ensured when the parameter K value of the K neighbor search in the step 2 is sufficiently large. Generally, the detection effect of the invention can be ensured to be highly similar to that of the traditional C2M method by selecting K to be more than or equal to 100.
Further, the deformation distance threshold D (unit: m) in the step 3 and the connected domain analysis threshold epsilon (unit: m) in the step 4.1 should be adjusted according to the density degree, the size and the like of the actually obtained point cloud model of the civil engineering structure, and D > ζ, epsilon >2ζ should be satisfied, wherein ζ is the point cloud resolution (unit: m) of the laser radar used for acquisition. In general, d=5ζ and ε=15ζ are preferable. When the accuracy required for defect detection is high, D and ε can be small.
The embodiment is an apparent defect detection method based on triangle centroid optimization, wherein the main hardware configuration used is Intel (R) Core (TM) i5-8600CPU and 32GBRAM, and the software environment is Ubuntu 18.04LTS 64-bit operating system; in the aspect of data acquisition, a point cloud model is obtained by scanning a foraging Mid-40 laser radar. The embodiment can be divided into four stages (as in fig. 1): the method comprises a data acquisition preprocessing stage, a deformation distance calculating stage, a defect information extracting stage and a testing stage.
The data acquisition preprocessing stage comprises the following steps:
Step 1, using a cardboard box (44.00 cm multiplied by 27.50cm multiplied by 18.50 cm) as a collection target object, simulating the civil engineering structure of which apparent defects are to be detected. A laser radar (such as the hybrid solid state laser radar foraging Mid-40 manufactured by Livox company) is used to collect the original surface point cloud model part of the target object at multiple angles. Due to the influence of external factors such as ambient illumination and temperature and humidity and internal factors such as jitter generated during laser radar working, noise points and outliers exist in the acquired local point cloud model, the point cloud model is downsampled by using a point cloud simplifying and denoising algorithm, the data volume is reduced, and the noise points and the outliers are removed.
Step 2, secondly, locally splicing the original surface point cloud model into a whole by using a point cloud registration and splicing technology, and recording asAnd acquiring a complete point cloud model of the target object.
Step 3, in order to simulate apparent defects existing in civil engineering and architectural engineering structures in actual engineering, artificially destroying the surfaces of the acquisition target objects, creating 8 apparent defects with various sizes and shapes, and repeating the steps 1 and 2 to obtain a defect point cloud model of the acquisition target objects
Step 4, for the defect point cloud modelThree-dimensional surface reconstruction is carried out by using a greedy projection triangulation algorithm, and a corresponding grid model/>, is obtained
Step 5, in order to facilitate comparison of experimental data in the subsequent test stage, the size and range of the apparent defect actually generated in step 3 are manually measured, and the results are shown in table 1:
TABLE 1 acquisition of apparent defect data for target objects
The deformation distance calculation comprises the following steps:
Step 6, for Each point P i in (1) is determined/>Triangle element nearest to P i Euclidean distance (Euclidean distance)Distance is denoted as/>Namely/>Each point and/>Euclidean distance between, i.e. representing/>Deformation distance of each point and its vicinity, thus array/>Element number and/>The number of elements is equal.
Step 7, manually setting a deformation distance threshold D, and determiningItem/>, satisfying D i > DConsider/>Corresponding Point set/>Defects exist near each point in the (c). Through multiple pre-experiments, when d=0.00005 m is determined, the determined defect area is ideal.
Further, in step 6, the pairEach point P i in (1) is determined/>Triangle surface element with nearest middle Euclidean distance with P i The process specifically comprises the following steps:
Step 6.1, extraction Centroid point cloud/>For triangle element M k, if three vertices are recorded asThe centroid G k is calculated as:
Thereby can establish And/>Mapping f between.
Step 6.2, determining an alternative to the nearest triangle element. For a pair ofBuild K-dTree, pair/>At each point P i inK neighbor searching is performed to obtain K neighbor points/>Wherein/>For searching the subscript sequence of the resulting points. Combining the mapping/>, in step 6.1It can be seen that AND/>Corresponding triangle bin sequence/>Let k=80 take into account N 3 =12591.
And 6.3, determining the nearest triangle element. Sequentially calculating P i andEuclidean distance between the two, and the distance obtains the minimum value/>Namely M r.
The defect information extraction stage includes the steps of:
And 8, analyzing the defect connected domain. Clustering method for sample points in high-dimensional feature space by simulating pattern recognition field, and determining defect point set in step 7 by using Euler cluster (Euclidean clustering) Performing connected domain analysis, and dividing the defect point set into a plurality of connected regions/>, according to Euclidean distance
And 9, generating a bounding box. Sequences of pairs of point setsAnd (3) calculating a direction bounding box (OBie-oriented minimum bounding box, OBB) of the defect region, wherein the center point of the OBB is regarded as the center point of the defect, and the size of the OBB is regarded as the size of the defect.
Further, in step 8, euclidean clustering is used for point-to-point setPerforming connected domain analysis to divide the defect point set into a plurality of connected regions (point set sequence)/>, according to Euclidean distanceThe method specifically comprises the following steps:
Step 8.1, for a point T in space, establishing a point set At/>, to be clusteredNeighbor searching is performed on a plurality of T neighbor points, and all points with the distance smaller than a threshold epsilon in the neighbor points are classified as/>Epsilon=0.015 m was chosen for euclidean clustering considering the distribution of apparent defects of the acquisition target object.
Step 8.2, selectingAnother point T' than T repeats step 8.1 until/>No further increases.
The test phase comprises the following steps:
Step 10, calculating the axis alignment bounding box (axis-aligned minimum bounding Box, AABB) of each defect area obtained in step 8, and calculating the intersection ratio (intersection over union, ioU) of the axis alignment bounding box and the defect range data (table 1) manually marked in the step 5.
Step 11, testing the method of the present invention using a confusion matrix (confusion matrix). Setting a threshold value eta for the IoU calculation result, considering that the detected result of IoU larger than eta is the experimental value which is consistent with the true value, using a True Positive (TP) to represent the number of defects of which the experimental value is consistent with the true value, a False Positive (FP) to represent the number of defects which are falsely detected as non-defects, and a false negative (FALSE NEGATIVE, FN) to represent the number of defects which are not detected. In combination with an evaluation system (benchmark) in the traditional target detection field, setting η=25%, and obtaining results by using the C2M method for centroid point cloud optimization and the traditional C2M method provided by the invention are shown in table 2.
Table 2c 2m method comparison of defect detection results before and after improvement
Accuracy rate of | Recall rate of recall | F1 score | Algorithm time consuming | |
Traditional C2M method | 0.71429 | 0.62500 | 0.66667 | 3.09853s |
C2M method of centroid Point cloud optimization (invention) | 0.71429 | 0.62500 | 0.66667 | 0.14152s |
Therefore, the influence of the improved method provided by the invention on the detection effect is negligible, and compared with the traditional C2M method, the method disclosed by the invention has the advantages that the time consumption is greatly reduced, and the method is improved by more than 90%.
The embodiments described above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the embodiments described above, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principle of the present invention should be made in the equivalent manner, and the embodiments are included in the protection scope of the present invention.
Claims (8)
1. The apparent defect detection method based on triangle centroid optimization is characterized by comprising the following steps of:
(1) Data input, respectively acquiring a period of point cloud model at different time points for a given target civil engineering structure to be analyzed Second-phase point cloud model/>Grid model obtained by three-dimensional surface reconstructionWherein/>Respectively, is a single data point in the point cloud,/>And (3) withThe number of the points contained in the point cloud is respectively;
(2) Deformation distance calculation, pair Each point P i in (1) is determined/>The triangle element M r closest to P i Euclidean distance,And this procedure is denoted as map/>Distance is denoted as/> Namely/>Each point and/>Euclidean distance between, i.e. representing/>A deformation distance of each point and its vicinity;
(3) Defect point set judgment, artificial setting of deformation distance threshold D, and determination of Item/>, satisfying D i > DConsider/>Corresponding Point set/>Defects exist near each point in (1), wherein N' i<N1 is/>The number of points determined to be defective;
(4) Defect connected domain analysis using Euclidean clustering of point-to-point sets Performing connected domain analysis, and dividing the defect point set into a plurality of connected regions/>, according to Euclidean distanceWherein/>And mutually exclusive of each other, each/>All the points included in the map are a defect area where the civil engineering structure actually exists;
(5) Bounding box generation, point-to-point set sequence And calculating the direction bounding box OBB of each element, namely the defect area, wherein the center point of the OBB is the center point of the defect, and the size of the OBB is the size of the defect.
2. The apparent defect detection method based on triangle centroid optimization of claim 1, wherein: in step (1)Three-dimensional surface reconstruction using greedy projection triangulation algorithm to obtain grid model/>
3. The apparent defect detection method based on triangle centroid optimization of claim 1, wherein: in the step (2)Each point P i in (1) is determined/>Triangle bin/>, closest to P i Euclidean distanceThe process specifically comprises the following steps:
(21) Extraction of Centroid point cloud/>For triangle element M k, if three vertices are recorded as/>The centroid G k is calculated as:
Thereby can establish And/>Mapping f between; where f is both bijective and homojective, because each triangular bin M k and centroid G k are in one-to-one correspondence, and/>Each triangular surface element of the model has a corresponding centroid; by the above calculation, it can be known that f (M k)=Gk, and M k=f-1(Gk);
(22) Determining an alternative to the nearest triangle element, pair Build K-dTree, pair/>At/>, P i K neighbor searching is performed to obtain K neighbor points/>Wherein/>A subscript sequence for the searched points; combining the mapping/>, in step (21)It can be seen that AND/>Corresponding triangle bin sequence/>This is the K triangular bins nearer to P i;
(23) Determining the nearest triangle surface element, and sequentially calculating P i and P i Euclidean distance between them, the distance obtaining minimum valueNamely M r.
4. The apparent defect detection method based on triangle centroid optimization of claim 1, wherein: in step (4), european clustering is used for point-to-point setPerforming connected domain analysis, and dividing the defect point set into a plurality of connected regions/>, according to Euclidean distanceThe method specifically comprises the following steps:
(41) For a point T in space, a point set is established At/>, to be clusteredNeighbor searching is performed on a plurality of T neighbor points, and all points with the distance smaller than a certain threshold epsilon in the neighbor points are classified as/>
(42) SelectingAnother point T' than T repeats step 4.1 until/>No further increases.
5. The apparent defect detection method based on triangle centroid optimization according to claim 3, wherein: in the step (22), K is more than or equal to 100.
6. The method for detecting apparent defects based on triangle centroid optimization as recited in claim 4, wherein: and (3) adjusting a deformation distance threshold D in the step (3) and a connected domain analysis threshold epsilon in the step (41) according to the density degree and the size of the actually obtained point cloud model of the civil engineering structure, wherein D > zeta and epsilon > 2 zeta are required to be met, and zeta is the point cloud resolution of the laser radar used for acquisition.
7. The method for detecting apparent defects based on triangle centroid optimization as recited in claim 6, wherein: d=5ζ, epsilon=15ζ.
8. The apparent defect detection method based on triangle centroid optimization of claim 1, wherein: in step (5), a principal component analysis method is used to calculate the directional bounding box of the defective region.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111355733.3A CN114092433B (en) | 2021-11-16 | 2021-11-16 | Apparent defect detection method based on triangle centroid optimization |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111355733.3A CN114092433B (en) | 2021-11-16 | 2021-11-16 | Apparent defect detection method based on triangle centroid optimization |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114092433A CN114092433A (en) | 2022-02-25 |
CN114092433B true CN114092433B (en) | 2024-06-25 |
Family
ID=80300944
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111355733.3A Active CN114092433B (en) | 2021-11-16 | 2021-11-16 | Apparent defect detection method based on triangle centroid optimization |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114092433B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107389701A (en) * | 2017-08-22 | 2017-11-24 | 西北工业大学 | A kind of PCB visual defects automatic checkout system and method based on image |
CN109345523A (en) * | 2018-09-21 | 2019-02-15 | 中国科学院苏州生物医学工程技术研究所 | Surface defects detection and three-dimensional modeling method |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7162073B1 (en) * | 2001-11-30 | 2007-01-09 | Cognex Technology And Investment Corporation | Methods and apparatuses for detecting classifying and measuring spot defects in an image of an object |
CN112345534B (en) * | 2020-10-30 | 2023-08-04 | 上海电机学院 | Defect detection method and system for particles in bubble plate based on vision |
CN112798648B (en) * | 2020-12-09 | 2022-06-03 | 浙江工业大学 | Composite material defect detection method based on generated kernel principal component thermal image analysis |
-
2021
- 2021-11-16 CN CN202111355733.3A patent/CN114092433B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107389701A (en) * | 2017-08-22 | 2017-11-24 | 西北工业大学 | A kind of PCB visual defects automatic checkout system and method based on image |
CN109345523A (en) * | 2018-09-21 | 2019-02-15 | 中国科学院苏州生物医学工程技术研究所 | Surface defects detection and three-dimensional modeling method |
Also Published As
Publication number | Publication date |
---|---|
CN114092433A (en) | 2022-02-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Xu et al. | Reconstruction of scaffolds from a photogrammetric point cloud of construction sites using a novel 3D local feature descriptor | |
CN111811420B (en) | Tunnel three-dimensional contour integral absolute deformation monitoring method and system | |
CN116167668B (en) | BIM-based green energy-saving building construction quality evaluation method and system | |
CN112347550B (en) | Coupling type indoor three-dimensional semantic graph building and modeling method | |
Bremer et al. | Eigenvalue and graph-based object extraction from mobile laser scanning point clouds | |
CN111932669A (en) | Deformation monitoring method based on slope rock mass characteristic object | |
Kim et al. | 3D as-built modeling from incomplete point clouds using connectivity relations | |
Alidoost et al. | Y-shaped convolutional neural network for 3d roof elements extraction to reconstruct building models from a single aerial image | |
CN104268940A (en) | MEMS structure reconstruction and detection method based on CT scanned images | |
Huo et al. | Research on product surface quality inspection technology based on 3D point cloud | |
CN112561989B (en) | Recognition method for hoisting object in construction scene | |
CN117371949B (en) | Three-dimensional visual model-based power transmission line construction safety monitoring method and system | |
CN113763529B (en) | Substation modeling method based on three-dimensional scanning | |
CN114092433B (en) | Apparent defect detection method based on triangle centroid optimization | |
CN115937149A (en) | Wall surface local deviation automatic detection method based on triangular gridding | |
KR101808958B1 (en) | Method for obtaining shape information of structure and method for measuring deformation of structure | |
CN111402256B (en) | Three-dimensional point cloud target detection and attitude estimation method based on template | |
JPH05280941A (en) | Inputting apparatus of three-dimensional shape | |
CN113743483A (en) | Road point cloud error scene analysis method based on spatial plane offset analysis model | |
Yastikli et al. | Automatic 3D building model generations with airborne LiDAR data | |
CN112069445A (en) | 2D SLAM algorithm evaluation and quantification method | |
CN115620278B (en) | Method for identifying and measuring materials | |
CN113393450B (en) | Data quality inspection method and system for digital line drawing and readable storage medium | |
JP7333870B2 (en) | Equipment data processing device, equipment data processing system, equipment data processing method, and equipment data processing program | |
CN113269207B (en) | Image feature point extraction method for grid structure light vision measurement |
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 | ||
GR01 | Patent grant |