CN108166362A - A kind of automatic identifying method of asphalt pavement crack type - Google Patents
A kind of automatic identifying method of asphalt pavement crack type Download PDFInfo
- Publication number
- CN108166362A CN108166362A CN201711411862.3A CN201711411862A CN108166362A CN 108166362 A CN108166362 A CN 108166362A CN 201711411862 A CN201711411862 A CN 201711411862A CN 108166362 A CN108166362 A CN 108166362A
- Authority
- CN
- China
- Prior art keywords
- crack
- grid
- length
- cracking
- section
- 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.)
- Granted
Links
Classifications
-
- E—FIXED CONSTRUCTIONS
- E01—CONSTRUCTION OF ROADS, RAILWAYS, OR BRIDGES
- E01C—CONSTRUCTION OF, OR SURFACES FOR, ROADS, SPORTS GROUNDS, OR THE LIKE; MACHINES OR AUXILIARY TOOLS FOR CONSTRUCTION OR REPAIR
- E01C23/00—Auxiliary devices or arrangements for constructing, repairing, reconditioning, or taking-up road or like surfaces
- E01C23/01—Devices or auxiliary means for setting-out or checking the configuration of new surfacing, e.g. templates, screed or reference line supports; Applications of apparatus for measuring, indicating, or recording the surface configuration of existing surfacing, e.g. profilographs
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Architecture (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Road Repair (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a kind of automatic identifying methods of asphalt pavement crack type, pass through labor measurement crack information first, crack data are detected by 3D, the direction of be cracked block and crack is divided to the characteristic index of computing unit grid after respective size of mesh opening again, secondly direction that the is fracture length respectively arranged according to row each in grid and judging crack, it finally calculates and is cracked in each unit section, transverse joint, the distribution of longitudinal joint;This method determines that the judgement of fractuer direction determines to provide set of systemization theory, and the identification for its automation provides basic procedure in the method specifically quantified as cracking degree;This flow is laid a good foundation for further discussion Crack failure characteristic information and the rule of development simultaneously, is of great significance to the scientific maintenance in joint filling joint seal section, is beneficial to the exploitation that three-dimensional laser detection technique is further utilized to carry out intelligent maintaining system.
Description
Technical field
The invention belongs to Asphalt Pavement Damage detection evaluation fields, and in particular to a kind of asphalt pavement crack type it is automatic
Recognition methods.
Background technology
Crack is as one of most important Damage Types of bituminous paving, by traffic load and the repeated action of temperature stress institute
It generates, is the principal mode of early damage to pavement, almost along with the entire lifetime of bituminous paving.According to《Highway technology shape
Condition evaluation criteria》On-site investigation to China drug in some provinces highway road pavement damage type, crack account for the ratio of pavement damage
60% is often exceeded, if conserving not in time, foundation base intension and the rapid decline stablized is easily caused, forms the structures such as cracking, pit slot
Venereal disease is done harm to, and is reduced the service life and is increased maintenance cost.And to different characteristic Causes of Cracking development research and analyse, it is timely and effective
Crack treatment and each crack to the adaptability of joint filling, to preventing it from further developing, it is ensured that pavement quality and structure are steady
It is fixed particularly significant.These are built upon on the basis of classification of rifts, and therefore, quantization identification classification is carried out to each section crack is
The essential condition of maintenance performance study and effective foundation of maintenance section selection are carried out, there is important meaning to promoting Pavement Performance
Justice.
Due to weather conditions and the difference of traffic, the existence form in domestic and international different regions crack is not exactly the same,
In addition the maintenance management system demand for cooperation each department, the classification in crack are also not quite similar.《Highway technology status assessment mark
It is accurate》Four major class (cracking, block are split, transverse joint, longitudinal joint) are divided into as according to by crack using lumpiness and direction, although with typical appearance
Feature, but do not have the origin cause of formation and position etc. and be related to the factor of Analysis on Mechanism and consider that such class object is unfavorable for as filling out
The evaluation criterion of selection is stitched, in addition combined with the classification results of other specifications, the division for summarizing current fracture type is as follows:
(1) transverse joint, longitudinal joint, irregular cracks are divided into according to direction;
(2) wheel path longitudinal joint is divided into according to position and non-wheel path longitudinal joint, edge crack, fatigue (load) crack, block is split;
(3) load Crack/fatigue crack/cracking, reflection crack, slumping crack, blocky crack are divided into according to the origin cause of formation;
(4) load Crack/fatigue crack/cracking is divided into according to lumpiness size, block is split.Therefore, it is necessary to combine cracking initiation
Mechanism and distribution characteristics again rationally clearly define class object progress.
The description of country's specification fracture classification foundation is still that qualitatively, these rules are mainly used for manually examining at present
It surveys, inefficiency too strong there are subjectivity, it is difficult to the problems such as being accurately positioned and carry out section division, evaluating severity.But with
The development of detection technique, based on accurate careful crack data, can realize and it is quantified, the classification of rifts method of automation
It is more conducive to scientific crack progressing evaluation and joint filling regional choice.And current specifications is not to conserve performance study as mesh
Mark carries out the considerations of classification detailed rules and regulations and sorting technique, can not preferably reflect the hair of the crack under different role (load, temperature etc.)
Open up feature, it is difficult to meet practical joint filling demand, and often occur that the section of joint filling is needed in time not located in practical application
Controlling leads to the generation of more serious plant disease, causes the increase of maintenance cost.Different types of crack carries out supporting for crack filling acquisition
It is also different to protect benefit.Crack according to mode of appearance difference be broadly divided into single crack and cracking (the larger block of lumpiness, which is split, to be divided in
In single crack, smaller piece of lumpiness, which is split, to be divided in cracking, and block is not split progress and individually classified by the present invention), it is cracked due to depositing
In the possibility that structure is destroyed, and skid resistance can be caused too poor using crack filling when density is too big, be generally possible to by splitting
It is all single crack that seam, which fills out the crack that envelope is punished, to meet practical joint filling demand, a current skill in the urgent need to address
Art problem is to determine the method that the crack selected suitable for joint filling is classified automatically.
Invention content
The purpose of the present invention is to overcome the above shortcomings and to provide a kind of automatic identifying method of asphalt pavement crack type,
Method can accurately and rapidly distinguish the types of fractures of the different characteristic suitable for joint filling demand, and the severity of fracture
It is divided.
In order to achieve the above object, the present invention includes the following steps:
Step 1, the crack that the unit crack road segment classification and 3D laser Crack Detection demarcated according to artificial detection obtain are total
By Decision tree classification threshold value, total crack length and threshold value are compared for length, and unit section is made to be divided into cracking
With non-cracking two types;
Step 2 influences the size of width according to wheel path, cracking lumpiness and crack, respectively for the identification of cracking block
Judgement with the direction in single crack divides respective size of mesh opening, and the characteristic index of computing unit grid;
Step 3, according to the result of calculation of four characteristic indexs of grid at wheel path, by linear SVM into
Row classification, and filter out the cracking block in non-cracking section with reference to neighbours' domain method;
Step 4 in the mesh generation size of the cracking block in filtering out non-cracking section, is occupied by single crack
The difference of grid range sums respectively to the fracture length that each row respectively arranges, according to the location determination crack of maximum value after summation
Length and the direction for judging crack;
Step 5 repeats step 2 to step 4, calculates and be cracked in each unit section, transverse joint, the distribution of longitudinal joint is cracked
Ratio, transverse joint length and longitudinal joint length complete the automatic identification of asphalt pavement crack type.
In step 1, threshold value determination method is as follows:Several samples are randomly selected, and are bisected into two groups, are respectively used to determine
The training and test of plan tree are trained decision tree by first group of sample, and the mark of cracking is judged according to practical artificial detection
Standard carries out sample the calibration in cracking section, and decision tree is judged by second group of sample, and threshold is completed after judging qualification
Value determines.
In step 2, mesh generation size and unit grid characteristic index calculating process are as follows:
The first step is compared in external specification to the different definition of wheel path, selection and the block range that is cracked in non-cracking section
Consistent scheme carries out the division of wheel path;
Section with cracking is divided into grid of the lumpiness for minimum dimension in specification, and determine wheel path by second step
With non-wheel path;
Road surface with transverse joint is divided into grid of the lumpiness for minimum dimension in specification;
Third walks, and calculates the length of each grid internal fissure, represents the fracture spacing of road surface local location, the length in crack
It by the air line distance of adjacent slits points all on crack and obtains, when a crack is in different grids, connects quilt
The separated two neighboring crack point of net boundary, seeks the intersection point of line and Grid Edge boundary line, line is divided into two sections by intersection point,
Two sections of length is contributed to respectively in affiliated grid, finally obtains the total length L of each grid internal fissurek,m,n。
The characteristic index of unit grid includes grid internal fissure length and Lk,m,n, laterally adjacent grid internal fissure length mean value
TLk,m,n, longitudinally adjacent grid internal fissure length moving average LLk,m,nWith wheel path internal fissure length mean value PLk,m,n;
xi, yi- represent the crack point set coordinate after asking for boundary intersection as boundary using grid;
K-represent section sample unit number;
M-represent grid y directions number at wheel path, m=1,2 ..., M, M is the maximum occurrences of m;
N-represent grid x directions number at wheel path, n=1,2 ..., N, N is the maximum occurrences of n.
In step 3, classified by linear SVM and be as follows:
The first step, selecting linear kernel function Φ () that sample is switched to the linear of Hilbert spaces from the former input space can
Divide problem,
The inner product space Φ () of feature vector=K (Xj, X) and=XjX, (5)
Xj=(Lk,m,n,TLk,m,n,LLk,m,n,PLk,m,n) (6)
Second step determines optimal hyperlane condition, if sample set (Xj,yj), wherein j=1,2 ..., 1600, X ∈ Rn, yj
∈ { -1,1 }, then optimal separating hyper plane is:
WTΦ (X)+b=0 (7)
Slack variable ξj>=0, determining for hyperplane parameter (W, b) is carried out, that is, is converted into constrained optimization problem
Introduce Lagrange multiplierConstrained optimization problem is converted into
Third walks, and SVM model parameters determine;
According to Kuhn-Tucker theorems, aiIt is not 0, corresponding training sample is known as supporting vector, is denoted as
It is carried out by SMO algorithmsDetermine, using heuritic approach setting two of which aiFor variable, fixed other ai
Multicomponent planning is converted into One- place 2-th Order planning problem, operation is carried out until meeting constraints, according to formula (8) and formula
(9) constraints in derives and calculating parameter:
b*=y0-W*TX0, (X0, y0) for the sample point set (11) corresponding to supporting vector
4th step, function interval calculation;
According to determining optimal hyperlane, determine different sample points to the function interval of the plane
Complete the preliminary classification of cracking block.
In step 4, the difference of single crack occupancy grid range, the tool summed respectively to the fracture length that each row respectively arranges
Body method is as follows,
The first step after mesh generation, judges in section the fracture length in the grid and grid where per crack;
Second step calculates the often total length RL of row and each column grid internal fissure belonging to per crackk,No,pAnd CLk,No,q;
P- represents the x directions number of each column grid comprising single crack Minimum Area, and P is the maximum value of row p;
Q- represents the y directions number of the often row grid comprising single crack Minimum Area, and Q is the maximum value of row q;
Third walks, and finds the row or column where total crack length peak value;
4th step, when peak value occur being expert at and when, judge crack for transverse joint, when peak value appear in row and when, judge crack
For longitudinal joint, and numbered according to corresponding crack and determine that corresponding total length is fracture length.
In step 5, be cracked ratio=cracking block meshes number/80;
Crack average length=∑ same type total crack length/same type crack number.
Compared with prior art, the present invention detects crack data, then will first by labor measurement crack information by 3D
The direction in cracking block and crack is divided to the characteristic index of computing unit grid after respective size of mesh opening, secondly according in grid
Direction that the is fracture length that each row respectively arranges and judging crack, finally calculates and is cracked in each unit section, transverse joint, point of longitudinal joint
Cloth;This method determines that the judgement of fractuer direction determines to provide set of system as cracking degree in the method specifically quantified
Theory, the identification for its automation provide basic procedure;This flow is further inquires into Crack failure characteristic information simultaneously
And the rule of development is laid a good foundation, and is of great significance to joint filling joint seal section, is beneficial to further examine using three-dimensional laser
Survey technology carries out the exploitation of intelligent maintaining system.
Description of the drawings
Fig. 1 is flow chart of the present invention;
Fig. 2 is non-cracking section and cracking section comparison of classification figure;Wherein (a) is non-cracking section, and (b) is cracking road
Section;
Fig. 3 is two kinds of wheel path and defines schematic diagram;Wherein (a) is the first wheel path definition mode, and (b) is second
Wheel path definition mode;
Fig. 4 is two kinds of wheel path division result contrast schematic diagrams;Wherein (a) is the division of the first wheel path definition mode
As a result, (b) is the division result of second of wheel path definition mode;
Fig. 5 is wheel path division result schematic diagram of the present invention;Wherein, (a) distinguishes schematic diagram for track, and (b) is wheel path
Position view;
Fig. 6 identifies mesh generation result schematic diagram for cracking block;
Fig. 7 judges mesh generation result schematic diagram for single fractuer direction;
Fig. 8 is crack by the computational methods schematic diagram of mesh segmentation;
Fig. 9 is unit grid internal fissure total length result schematic diagram;
Figure 10 is four neighborhood exclusive method example schematic diagrams;
Figure 11 judges schematic diagram for transverse and longitudinal seam identification.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and examples.
Embodiment:
Referring to Fig. 1, embodiment is by taking the detection data in the section of 1.6km as an example.
First, non-cracking section screening;
1st, artificial detection crack information;
The main means that artificial detection is detected as pavement behavior, by having testing staff's road pavement crack of certain experiences
Type judged that record crack location, direction, the information that the present invention is obtained using artificial detection is classified as classification of rifts
Actual standard, for being compared with the result classified automatically using 3D detection datas.
2nd, 3D detects crack data;
The present invention obtains accurate crack data using 3D laser detection datas, randomly selects the wide 3.66m in track, unit is long
50 unit sections of 5m are spent for sample, 0,1,2 are numbered to all slits in each section ..., with 5m sections towards inspection
The left side graticule surveyed into line direction initial section is origin, records the coordinate of all section difference number cracks point data.
1 3D of table detects crack data
According to the difference in crack progressing stage, section is divided into cracking section and non-two, section of cracking by the present invention first
Point, cracking section, that is, cracking degree is high, is substantially not present the section of other types of fractures;Non- cracking section situation is complex,
Comprising transverse joint and longitudinal joint, therefore the division of two kinds of road segment classifications is first passed through, non-cracking section is screened.
In order to carry out precision judgement to training and test sample, the standard pair of cracking is judged according to practical artificial detection first
Sample carries out the calibration in cracking section.Then the sum of section wheel path internal fissure length is utilized, selects decision tree to road segment classification
It is divided, obtains the sample unit section division result such as Fig. 2.
100 samples are randomly selected from table 1, is divided into 2 groups, every group 50, is respectively used to the training and test of decision tree.
In order to accurately be judged training and test sample, judge that the standard of cracking carries out sample according to practical artificial detection first
The calibration in cracking section.
Decision tree is trained using 50 training samples, the judgement rate for obtaining cracking section when threshold value is 2552cm is
94%, it is believed that be reliable.Therefore, it sets when single sample section internal fissure length reaches more than 2552cm as the section that is cracked,
By classifying to 50 test samples, specific result of calculation is as shown in table 2, obtains the accuracy judged cracking section
Up to 98%.
2 test sample of table cracking section classification results
2nd, mesh generation and calculation process;
1st, wheel path location determination;
Caused by cracking is by load, occur mainly in the range of wheel path, the cracking block in non-wheel path is main
It is formed by the cracking extension in wheel path, and when non-wheel path has cracking, then the cracking in wheel path is very serious, should
Entire section is defined as cracking section by this, therefore the present invention calculates to the identification of cracking and with ratio and only considers wheel path model
Enclose interior crack.
It is defined firstly the need of to wheel path, according to two kind definition of the external specification to wheel path, respectively Fig. 3 (a)
With the form shown in Fig. 3 (b).The two is slightly different the definition of wheelmark bandwidth, according to two definition to collected crack
Section is divided, and obtains result shown in Fig. 4.It can be seen that second of definition and the position in actual loading crack more meet, because
This present invention using wheel path in scheme 2 definition, first by track subregion, as shown in Figure 5.Entire track passes through two wheelmarks
Band and lane line are divided into 5 regions, and two wheel paths are respectively 0.915m, wheel path and lane line spacing 0.4575m, and two are taken turns
Mark band is at a distance of 0.915m.
2nd, global grid size determines;
1) mesh generation, based on cracking block identification;
Mesh generation is the fracture spacing in order to calculate road surface part, so as to which positioning be identified to cracking block, to ensure
Mesh scale can recognize that cracking definition lumpiness, grid need to approach it is rectangular, according to it is non-cracking section on cracking block distribution spy
Property, the size of mesh opening about cracking block identification determines to carry out on the basis of wheel path and non-wheel path divide.It is specific to divide
Scale can need progress value according to classification of rifts, for single crack is avoided to obscure with cracking block generation, according to China《It is public
Road technique status assessment standard》Middle cracking lumpiness is less than the definition of 0.5m, and section is divided into the grid of lumpiness approximation 0.5m, by
In the section that 3D laser detection equipments obtain for 5m long, therefore be longitudinally divided into 10 sections, every section of 0.5m, laterally due to design have a lot of social connections for
3.66m with reference to the wheel path division result of step 2, is laterally divided into 8 sections, obtains mesh generation example as shown in Figure 6, will
5m long, the wide 3.66m in track section be divided into 10*8 grid, wherein longitudinal direction the 2nd, 3,6,7 is classified as wheel path, the 1st, 4,5,8
It is classified as non-wheel path.
2) mesh generation based on transverse and longitudinal seam identification;
Due to the uncertainty in single fractue spacing direction, if mesh generation scale is too big, less than the crack of mesh scale
A grid may only be accounted for, it is difficult to carry out the judgement in transverse and longitudinal direction, therefore in order to which more accurately fracture direction is identified, carry
The accuracy rate that height judges needs to carry out grid thinner division.
Currently without the definition of fracture minimum length, according to《Highway technology status assessment standard》, calculating crack
During impaired area, taking influences width 0.2m, it is therefore assumed that fracture length should be greater than influencing width, it, will for the accuracy rate of judgement
Section is divided into the grid of lumpiness approximation 0.2m, since unit section is 5m long in the present invention, is longitudinally divided into 24 sections, every section
0.2083m, laterally since design is had a lot of social connections for 3.66m, lateral rounding is divided into 18 sections, every section of 0.2033m, entire 5m long vehicles
Road is divided into 432 grids, it is this divide scale on the one hand meet fracture minimum length it is assumed that on the other hand utmostly
Raising judge accuracy rate, as shown in Figure 7.
3rd, unit grid characteristic index calculates;
After section is divided into grid, as shown in figure 8, calculating the length of each grid internal fissure, road surface part position is represented
The fracture spacing put.The air line distance that the length in crack passes through adjacent slits points all on crack and obtain, when a crack
When in different grids, the intersection point for by the separated two neighboring crack point of net boundary, asking line and Grid Edge boundary line is connected,
Line is divided into two sections by intersection point, two sections of length is contributed to respectively in affiliated grid, finally obtains each grid internal fissure
Total length Lk,m,n, as shown in Figure 9.
The present invention reflects the density in crack, and then according to the formation mechenism of cracking by calculating the length of grid internal fissure
And distribution characteristics, propose 4 Symbiotic relationships (grid internal fissure length and Lk,m,n, laterally adjacent grid internal fissure length it is equal
Value TLk,m,n, longitudinally adjacent grid internal fissure length moving average LLk,m,nWith wheel path internal fissure length mean value PLk,m,n)
For the identification for the block that is cracked in non-cracking section.The present invention takes 40 grids in two wheel paths to be analyzed.For being cracked
The identification of block finds the cracking block in non-cracking section by linear SVM (SVM), determines the position that cracking occurs,
Calculation specifications are as follows:
xi, yi- represent the crack point set coordinate after asking for boundary intersection as boundary using grid, i=1,2 ..., I;
K-represent section sample unit number;
M-represent is numbered in grid y directions at wheel path, and (M is the maximum occurrences of m, M=in the present invention by m=1,2 ..., M
10);
N-represent is numbered in grid x directions at wheel path, and (N is the maximum occurrences of n, N=in the present invention by n=1,2 ..., N
4)。
3rd, the cracking block identification based on grid location;
1st, training linear vector machine model;
In order to carry out precision judgement to training and test sample, the standard pair of cracking is judged according to practical artificial detection first
Sample carries out the calibration of cracking block.Then 4 characteristic indexs proposed using step 3 select linear SVM (SVM)
Cracking block is identified, 40 non-cracking section samples, each two, sample wheel path 40 are randomly selected in sample database
Grid, 1600 grids altogether, each grid forms a data by 4 characteristic indexs, as training set and the sample of test set
This information, as shown in Table 3 and Table 4:
3 training set sample information of table (amounts to 40 sections)
Note:0 represents non-cracking block, and 1 represents cracking block
4 test set sample information of table (amounts to 40 sections)
1) selection linear kernel function Φ () first switchs to sample from the former input space linear separability in Hilbert spaces
Problem;
Φ ()=K (Xj, X) and=XjX (i.e. the inner product space of feature vector) (5)
Xj=(Lk,m,n,TLk,m,n,LLk,m,n,PLk,m,n) (6)
2) optimal hyperlane condition is determined;
WTΦ (X)+b=0 (7)
Slack variable ξj>=0, determining for hyperplane parameter (W, b) is carried out, that is, is converted into constrained optimization problem
Introduce Lagrange multiplierConstrained optimization problem is converted into
3) SVM model parameters determine;
According to Kuhn-Tucker theorems, aiIt is not 0, corresponding training sample is known as supporting vector, is denoted as
It is carried out by SMO algorithmsDetermine, using heuritic approach setting two of which aiFor variable, fixed other ai
Multicomponent planning is converted into One- place 2-th Order planning problem, operation is carried out until meeting constraints, according in (8) (9)
Constraints derives and calculating parameter:
b*=y0-W*TX0, (X0, y0) for the sample point set (11) corresponding to supporting vector
4) function interval calculation;
According to determining optimal hyperlane, determine different sample points to the function interval of the plane;
The present invention using training set as sample, is carried out 1) to 3) calculating, according to trained model, 4) progress is fallen into a trap first
It calculates, and then obtains the preliminary recognition result of cracking block.
2nd, four neighborhood exclusive methods determine cracking block;
The present invention excludes the independent grid that grid crack total length reaches a certain level when identification is cracked block,
As shown in Figure 10, when many cracks intersect, the length of grid internal fissure also can be larger, but these cracks can't be formed it is small
Block, therefore cannot be judged as being cracked.According to the formation mechenism of cracking, cracking is as caused by the damage of base, and therefore, cracking is not
It can individually generate, must exist in a certain range in a grid.Therefore, present invention definition, which only has in four neighborhoods, grid
When being also judged as cracking block, just judge it for cracking.
1600 datas are randomly divided into 2 groups, every group 800, are respectively used to the training and test of SVM.Pass through 800 instructions
What white silk data were cracked judges rate for 95.1%, it is believed that is reliable.Using trained SVM to 800 test datas into
Row classification, after being rejected according to four neighborhood exclusive methods, obtains that specific result of calculation is as shown in table 5, and cracking block is judged
Accuracy is up to 93.5%.
5 test sample of table cracking block sort result
4th, the fractuer direction identification based on mesh refinement;
According to the cracking block recognition result determined, obtain the distributing position of non-cracking block, non-cracking block include transverse joint and
Longitudinal joint, below fracture direction judged, be as follows:
1) after mesh generation, the fracture length in the grid and grid in section where every crack is judged, such as Figure 11 (a)
It is shown;
2) the often total length RL of row and each column grid internal fissure belonging to per crack is calculatedk,No,pAnd CLk,No,q, such as Figure 11
(b) shown in;
P- represents the x directions number of each column grid comprising single crack Minimum Area, and P is the maximum value of row p;
Q- represents the y directions number of the often row grid comprising single crack Minimum Area, and Q is the maximum value of row q;
3) row or column where total crack length peak value is found, such as the number in Figure 11 (b) (c);
4) when being expert at occurs in peak value and when, judge crack for transverse joint, when peak value appear in row and when, it is vertical to judge crack
Seam, and numbered in table 1 according to corresponding crack and determine that corresponding total length is fracture length, as shown in Figure 11 (c).
5th, types of fractures is distributed;
It can identify to obtain in each unit section according to above step and be cracked, transverse joint, the distribution of longitudinal joint, that is, be cracked ratio,
Transverse joint length and longitudinal joint length, as shown in table 6:
Cracking ratio=cracking block meshes number/80 (the cracking ratio in present invention definition cracking section is 100%);
Crack average length=∑ same type total crack length/same type crack number.
6 types of fractures of table is distributed
Claims (7)
1. a kind of automatic identifying method of asphalt pavement crack type, which is characterized in that include the following steps:
Step 1, the crack overall length that the unit crack road segment classification and 3D laser Crack Detection demarcated according to artificial detection obtain
Degree, by Decision tree classification threshold value, total crack length and threshold value are compared, make unit section be divided into cracking and
Non- cracking two types;
Step 2 influences the size of width according to wheel path, cracking lumpiness and crack, respectively the identification for cracking block and list
The judgement in the direction in root crack divides respective size of mesh opening, and the characteristic index of computing unit grid;
Step 3 according to the result of calculation of four characteristic indexs of grid at wheel path, is divided by linear SVM
Class, and filter out the cracking block in non-cracking section with reference to neighbours' domain method;
Step 4 in the mesh generation size of the cracking block in filtering out non-cracking section, passes through single crack occupancy grid
The difference of range sums respectively to the fracture length that each row respectively arranges, according to the length in the location determination crack of maximum value after summation
And judge the direction in crack;
Step 5 repeats step 2 to step 4, calculates and be cracked in each unit section, transverse joint, the distribution of longitudinal joint, that is, be cracked ratio
Example, transverse joint length and longitudinal joint length complete the automatic identification of asphalt pavement crack type.
A kind of 2. automatic identifying method of asphalt pavement crack type according to claim 1, which is characterized in that step 1
In, threshold value determination method is as follows:Several samples are randomly selected, and are bisected into two groups, are respectively used to the training and survey of decision tree
Examination, is trained decision tree by first group of sample, and the standard for judging cracking according to practical artificial detection carries out tortoise to sample
The calibration in section is split, decision tree is judged by second group of sample, determining for threshold value is completed after judging qualification.
A kind of 3. automatic identifying method of asphalt pavement crack type according to claim 1, which is characterized in that step 2
In, mesh generation size and unit grid characteristic index calculating process are as follows:
The first step is compared in external specification to the different definition of wheel path, is selected consistent with cracking block range in non-cracking section
Scheme carry out wheel path division;
Section with cracking is divided into grid of the lumpiness for minimum dimension in specification by second step, and determines wheel path and non-
Wheel path;
Road surface with transverse joint is divided into grid of the lumpiness for minimum dimension in specification;
Third walks, and calculates the length of each grid internal fissure, represents the fracture spacing of road surface local location, and the length in crack passes through
On crack the air line distance of all adjacent slits points and obtain, when a crack is in different grids, connect by grid
The separated two neighboring crack point in boundary seeks the intersection point of line and Grid Edge boundary line, line is divided into two sections, two sections by intersection point
Length contribute to respectively belonging in grid, finally obtain the total length L of each grid internal fissurek,m,n。
A kind of 4. automatic identifying method of asphalt pavement crack type according to claim 1, which is characterized in that element mesh
The characteristic index of lattice includes grid internal fissure length and Lk,m,n, laterally adjacent grid internal fissure length mean value TLk,m,n, longitudinal phase
The moving average LL of adjacent grid internal fissure lengthk,m,nWith wheel path internal fissure length mean value PLk,m,n;
xi, yi- represent the crack point set coordinate after asking for boundary intersection as boundary using grid;
K-represent section sample unit number;
M-represent grid y directions number at wheel path, m=1,2 ..., M, M is the maximum occurrences of m;
N-represent grid x directions number at wheel path, n=1,2 ..., N, N is the maximum occurrences of n.
A kind of 5. automatic identifying method of asphalt pavement crack type according to claim 1, which is characterized in that step 3
In, classified by linear SVM and be as follows:
The first step selects the linear separability that linear kernel function Φ () switchs to sample from the former input space Hilbert spaces to ask
Topic,
The inner product space Φ () of feature vector=K (Xj, X) and=XjX, (5)
Xj=(Lk,m,n,TLk,m,n,LLk,m,n,PLk,m,n) (6)
Second step determines optimal hyperlane condition, if sample set (Xj,yj), wherein j=1,2 ..., 1600, X ∈ Rn, yj∈{-1,
1 }, then optimal separating hyper plane is:
WTΦ (X)+b=0 (7)
Slack variable ξj>=0, determining for hyperplane parameter (W, b) is carried out, that is, is converted into constrained optimization problem
Introduce Lagrange multiplierConstrained optimization problem is converted into
Third walks, and SVM model parameters determine;
According to Kuhn-Tucker theorems, aiIt is not 0, corresponding training sample is known as supporting vector, is denoted as
It is carried out by SMO algorithmsDetermine, using heuritic approach setting two of which aiFor variable, fixed other aiTo be more
Meta-planning is converted to One- place 2-th Order planning problem, operation is carried out until meeting constraints, according in formula (8) and formula (9)
Constraints derive and calculating parameter:
b*=y0-W*TX0, (X0, y0) for the sample point set (11) corresponding to supporting vector
4th step, function interval calculation;
According to determining optimal hyperlane, determine different sample points to the function interval of the plane
Complete the preliminary classification of cracking block.
A kind of 6. automatic identifying method of asphalt pavement crack type according to claim 1, which is characterized in that step 4
In, the difference of single crack occupancy grid range, the specific method summed respectively to the fracture length that each row respectively arranges is as follows,
The first step after mesh generation, judges in section the fracture length in the grid and grid where per crack;
Second step calculates the often total length RL of row and each column grid internal fissure belonging to per crackk,No,pAnd CLk,No,q;
P- represents the x directions number of each column grid comprising single crack Minimum Area, and P is the maximum value of row p;
Q- represents the y directions number of the often row grid comprising single crack Minimum Area, and Q is the maximum value of row q;
Third walks, and finds the row or column where total crack length peak value;
4th step, when peak value occur being expert at and when, judge crack for transverse joint, when peak value appear in row and when, it is vertical to judge crack
Seam, and numbered according to corresponding crack and determine that corresponding total length is fracture length.
A kind of 7. automatic identifying method of asphalt pavement crack type according to claim 1, which is characterized in that step 5
In, be cracked ratio=cracking block meshes number/80;
Crack average length=∑ same type total crack length/same type crack number.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711411862.3A CN108166362B (en) | 2017-12-23 | 2017-12-23 | Automatic identification method for asphalt pavement crack types |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711411862.3A CN108166362B (en) | 2017-12-23 | 2017-12-23 | Automatic identification method for asphalt pavement crack types |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108166362A true CN108166362A (en) | 2018-06-15 |
CN108166362B CN108166362B (en) | 2020-04-07 |
Family
ID=62523864
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711411862.3A Active CN108166362B (en) | 2017-12-23 | 2017-12-23 | Automatic identification method for asphalt pavement crack types |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108166362B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109002667A (en) * | 2018-09-21 | 2018-12-14 | 江苏中路工程技术研究院有限公司 | A kind of crack origin analysis method based on crack progressing form |
CN109870458A (en) * | 2019-02-19 | 2019-06-11 | 福建农林大学 | It is a kind of based on the pavement crack of three-dimensional laser sensor and bounding box detection and classification method |
CN111738371A (en) * | 2020-08-26 | 2020-10-02 | 西南石油大学 | Stratum fracture number prediction method based on random forest data mining |
CN117030730A (en) * | 2023-08-10 | 2023-11-10 | 苏州苏瑞膜纳米科技有限公司 | Rapid defect detection method based on three-dimensional laser scanning |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH04240510A (en) * | 1991-01-24 | 1992-08-27 | Hajime Sangyo Kk | Discriminating method of defect |
CN103323526A (en) * | 2013-05-30 | 2013-09-25 | 哈尔滨工业大学 | Welding line defect detection and identification method based on ultrasonic phased array and support vector machine |
CN103364410A (en) * | 2013-07-23 | 2013-10-23 | 三峡大学 | Crack detection method of hydraulic concrete structure underwater surface based on template search |
WO2014033643A1 (en) * | 2012-08-31 | 2014-03-06 | Systèmes Pavemetrics Inc. | Method and apparatus for detection of foreign object debris |
CN106087677A (en) * | 2016-06-02 | 2016-11-09 | 上海华城工程建设管理有限公司 | Asphalt pavement crack type automatic identifying method |
-
2017
- 2017-12-23 CN CN201711411862.3A patent/CN108166362B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH04240510A (en) * | 1991-01-24 | 1992-08-27 | Hajime Sangyo Kk | Discriminating method of defect |
WO2014033643A1 (en) * | 2012-08-31 | 2014-03-06 | Systèmes Pavemetrics Inc. | Method and apparatus for detection of foreign object debris |
CN103323526A (en) * | 2013-05-30 | 2013-09-25 | 哈尔滨工业大学 | Welding line defect detection and identification method based on ultrasonic phased array and support vector machine |
CN103364410A (en) * | 2013-07-23 | 2013-10-23 | 三峡大学 | Crack detection method of hydraulic concrete structure underwater surface based on template search |
CN106087677A (en) * | 2016-06-02 | 2016-11-09 | 上海华城工程建设管理有限公司 | Asphalt pavement crack type automatic identifying method |
Non-Patent Citations (2)
Title |
---|
惠冰: "沥青路面车辙破坏模式识别、多维度指标评价与预测研究", 《万方数据库》 * |
李伟 等: "基于3D数据和双尺度聚类算法的路面裂缝检测", 《华南理工大学学报(自然科学版)》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109002667A (en) * | 2018-09-21 | 2018-12-14 | 江苏中路工程技术研究院有限公司 | A kind of crack origin analysis method based on crack progressing form |
CN109002667B (en) * | 2018-09-21 | 2022-04-15 | 江苏中路工程技术研究院有限公司 | Crack formation cause analysis method based on crack development form |
CN109870458A (en) * | 2019-02-19 | 2019-06-11 | 福建农林大学 | It is a kind of based on the pavement crack of three-dimensional laser sensor and bounding box detection and classification method |
CN109870458B (en) * | 2019-02-19 | 2021-11-09 | 福建农林大学 | Pavement crack detection and classification method based on three-dimensional laser sensor and bounding box |
CN111738371A (en) * | 2020-08-26 | 2020-10-02 | 西南石油大学 | Stratum fracture number prediction method based on random forest data mining |
CN117030730A (en) * | 2023-08-10 | 2023-11-10 | 苏州苏瑞膜纳米科技有限公司 | Rapid defect detection method based on three-dimensional laser scanning |
Also Published As
Publication number | Publication date |
---|---|
CN108166362B (en) | 2020-04-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Peraka et al. | Pavement asset management systems and technologies: A review | |
CN110135351B (en) | Built-up area boundary identification method and equipment based on urban building space data | |
CN108166362A (en) | A kind of automatic identifying method of asphalt pavement crack type | |
CN107943880B (en) | Geological disaster susceptibility improvement and evaluation method based on analytic hierarchy process | |
Tran et al. | One stage detector (RetinaNet)-based crack detection for asphalt pavements considering pavement distresses and surface objects | |
CN106087677B (en) | Asphalt pavement crack type automatic identifying method | |
CN105701204B (en) | The extracting method and display methods of electronic map interest point based on road network | |
CN104614766B (en) | Seismic horizon skeletonization | |
CN110210776B (en) | Tunnel construction dynamic risk assessment method based on monitoring measurement | |
Thube | Artificial neural network (ANN) based pavement deterioration models for low volume roads in India | |
CN108596396A (en) | One kind is based on the modified pavement performance prediction of maintenance history and maintenance process and device | |
CN105825169A (en) | Road-image-based pavement crack identification method | |
CN110929939A (en) | Landslide hazard susceptibility spatial prediction method based on clustering-information coupling model | |
CN116797055A (en) | Urban road planning method and system based on Internet of things | |
CN111179592A (en) | Urban traffic prediction method and system based on spatio-temporal data flow fusion analysis | |
CN116739376A (en) | Highway pavement preventive maintenance decision method based on data mining | |
Cushman et al. | Landscape metrics, scales of resolution | |
CN105092451A (en) | Method for predicting void ratio of asphalt mixture | |
Bosurgi et al. | Optimizing artificial neural networks for the evaluation of asphalt pavement structural performance | |
Oni et al. | Analysis of accessibility Impact on commercial property values in Ikeja, Nigeria | |
CN111008730A (en) | Crowd concentration degree prediction model construction method and device based on urban space structure | |
CN115906669A (en) | Dense residual error network landslide susceptibility evaluation method considering negative sample selection strategy | |
Vanderhaegen et al. | Developing urban metrics to describe the morphology of urban areas at block level | |
CN113361854B (en) | Accident multi-point identification method based on Thiessen polygon and application thereof | |
Liu et al. | Road density analysis based on skeleton partitioning for road generalization |
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 | ||
GR01 | Patent grant |