CN107389697B - A kind of crack detection method based on half interactive mode - Google Patents
A kind of crack detection method based on half interactive mode Download PDFInfo
- Publication number
- CN107389697B CN107389697B CN201710558215.9A CN201710558215A CN107389697B CN 107389697 B CN107389697 B CN 107389697B CN 201710558215 A CN201710558215 A CN 201710558215A CN 107389697 B CN107389697 B CN 107389697B
- Authority
- CN
- China
- Prior art keywords
- point
- crack
- pixel
- seed
- coordinate
- 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
Classifications
-
- 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
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
Abstract
The invention proposes a kind of based on half interactive crack detection method, primarily to solving the robustness problem and requirements for high precision of the crack detection method in current Crack Detection field, and is effectively applied to various actual scenes.This method is based on interactive seed point and chooses and crack forward prediction progress Crack Detection, noise, illumination, the influence of the unfavorable factors such as background complexity can not only be effectively eliminated, and image image quality is required low, there is adaptive ability to image capture environment, there is universality.The present invention can also simple extension to other similar features occasion, such as in medical domain capillary detection etc..
Description
Technical field
It is specifically a kind of based on half interactive crack detection method the invention belongs to technical field of machine vision.
Background technique
The every field of socio-economic development has been goed deep into the presence in crack, for example the highway pavement in traffic and transport field is split
Seam, freeway tunnel wall crack etc.；Optical glass crack in industrial products field, wood surface crack etc.；Its safety
Property detection be directly related to economic development and safety etc. significant problems.Although crack is mostly less than 1mm, but one kind is most common
The form of expression is endangered with most important, on the one hand it implies the appearance of harm；On the other hand potential serious safety is implied
The factors such as hidden danger, such as infiltration, which induce profound remitted its fury, will cause immeasurable loss；Meanwhile if crack not in time
Processing will shorten the service life of industrial products or highway etc..Therefore, the quick detection in crack, efficient identification and reparation in time
It is the effective way for promoting service of goods quality and prolonging its service life.
The width detection in crack is most important index in Crack Detection.The crack in a very long time, each field
Detection depends on experienced professional and completes.Firstly, finding crack by human eye, then with vernier caliper measurement, it is wide
Degree.It not only causes serious waste of human resource, and the poor in timeliness detected, and precision is low, and subjectivity is strong.In general,
The complicated multiplicity of the appearance form in crack, traditional professional's detection have been unable to meet actual application demand, this to be based on
Automatic (or semi-automatic) detection technique of machine vision becomes current hot research field.Automatic (or semi-automatic) detection herein
In method, target image is acquired first, then, computer automatic (or semi-automatic) analyzes the content of image and identifies crack,
Finally calculate width index.In the detection method based on digital picture, the analysis of target image and crack identification are crucial.Mesh
Before, common processing method is mostly traditional image processing algorithm, such as Threshold segmentation, Fusion Features method and model learning method,
Such method accuracy of identification is high, but more sensitive to the characteristics of image of extraction, and needs a large amount of training data, artificial to mark
It is at high cost.
In recent years, the detector technologies of the subpixel accuracy based on Hessian matrix are good etc. with its precision height and robustness
Feature is extracted in light strip center of structured light and is applied in curvilinear structure detection.Although the feature of Hessian matrix to
Amount can characterize the normal direction and tangential direction of curvilinear structures well, but this method seeks extreme value by Taylor series expansion
Point comes the center and boundary of Detection curve structure, often is influenced to obtain Local Extremum by noise spot and be judged by accident.
Usually in practical projects, the influence for the external environments such as image capturing system is often illuminated by the light is bigger, is difficult to protect
Card collects the crack image of clean mark；It is influenced by crack local width, depth, direction, micro-cracks and background contrast
It spends weak and is difficult to show complete geometric shape, but performance is intermittent, it is indistinct；The target image of actual acquisition
The uncertainty that middle background is shown, complexity and diversity make existing algorithm all be difficult to meet practical application request, or
It is only limited to specific application scenarios.Such as in Tunnel testing field, industry practical application shows both at home and abroad without any one
Crack Detection algorithm is able to satisfy practical engineering application demand.
Summary of the invention
The embodiment provides a kind of based on half interactive crack detection method, primarily to solving current
The robustness problem and requirements for high precision of crack detection method in Crack Detection field.
To achieve the goals above, this invention takes following technical solutions.
A kind of crack detection method based on half interactive mode, comprising the following steps:
Step 1: a series of seed points are chosen at crack on target image, obtain the coordinate of seed point；
Step 2: extracting crack area image to the target image, pre-process to the crack area image,
Increase the contrast of crack area image and background；
Step 3: the tangential direction and normal direction of each pixel are calculated pretreated crack area image；
Step 4: being numbered the seed point of selection, enables i for seed point index, i is natural number, i > 0, and is initialized
For i=1；
Step 5: by crack from tracking and potential crack point extended method, the tangential direction and normal of pixel are utilized
The crack between i-th of seed point and i+1 seed point is detected in direction；
Step 6: judging whether all seed points detect and finish, if it is judged that be it is no, then enable i=i+1, hold again
Row step 5；If it is judged that be it is yes, then extract a complete crack, and obtain the width in crack.
Further, include: to the target image extraction crack area image in the step 2
If the sum for the seed point chosen is N, the abscissa of the horizontal direction of i-th of seed point is xiVertical direction is indulged
Coordinate is yi
The smallest horizontal direction coordinate value x in N number of seed point is sought respectivelymin, maximum horizontal direction coordinate value xmax, minimum
Vertical direction coordinate value ymin, maximum vertical direction coordinate value ymax, the left upper apex coordinate for extracting crack area is (xmin,
ymin), and bottom right vertex coordinate is (xmax,ymax), it is extracted according to the left upper apex coordinate and bottom right vertex coordinate rectangular
Crack area image.
Further, pre-processing in the step 2 to the crack area image increases crack area image
Contrast with background includes:
Bilateral filtering operation is carried out to the crack area image, logarithm behaviour is carried out to filtered crack area image
Make, and remove mean normalization, increases the contrast of crack area image and background.
Further, the tangent line for calculating pretreated crack area image each pixel in the step 3
Direction and normal direction, comprising:
The tangential direction and normal direction of pixel correspond to the feature vector of the Hessian matrix of the pixel, consider to split
Stitch some pixel p (x, y) in area image, Hessian matrix are as follows:
Wherein:
Wherein I is enhanced crack area image in step 2, and g (x, y) is dimensional Gaussian convolution mask,For convolution
Operation, I (x, y) are the matrix centered on pixel p (x, y) and with the sizes such as Gaussian convolution template, Hessian matrix H (x,
Y) the corresponding feature vector n (p) of the absolute value of maximum eigenvalue=[nx,ny], for the normal direction of pixel p (x, y),
And the corresponding feature vector v (p) of absolute value=[v of minimal eigenvaluex,vy], for the tangential direction of pixel p (x, y).
Further, being tracked certainly in the step five by crack and potential crack point extended method, utilize pixel
The tangential direction and normal direction of point detect the crack between i-th of seed point and i+1 seed point, comprising the following steps:
Step 5-1: crack is carried out from tracking from starting seed point, and is being worked as constantly calculating crack from tracing process
The width of preceding point, and mark the crack pixel detected；From tracer technique, specific step is as follows in crack:
1) enabling p is the starting point of tracking certainly, and starts the tracking in progress crack with this starting point；
2) using current point p as zero point, the two of crack where determining the point in [0,30] pixel coverage in the direction of its normal
A boundary point, judgment criterion are gradient and tangential direction similarity；All pixels point between two boundary points is to detect
Crack, and the crack is marked；Final updating current point p is the central point between two boundary points；
3) increase a pixel step length in the tangential direction of current point p, and coordinate is rounded to obtain next point p1；?
Processing method is identical on the opposite direction of tangent line；
4) judge whether tangential direction and the tangential direction of next point p1 of current point p are similar；Whether point p1 is detecting
Inside crack, i.e., whether between two seed points；Whether p1 point is not labeled；
If 5) p1 point meets above three condition, updating current point p is p1 point, step 2) is executed, if being unsatisfactory for
Execute step 6)；
6) increase by two pixel step lengths in the tangential direction of current point p, and coordinate is rounded to obtain next point p2；?
Processing method is identical on the opposite direction of tangent line；
7) same to step 4), judges whether p2 point meets following three conditions: whether point p2 is inside fracture detection, i.e.,
Between two seed points；Whether p2 point is not labeled；
If 8) p2 point meets above three condition, updating current point p is p2 point, step 2) is executed, if not satisfied, then
Certainly termination is tracked in crack；
Step 5-2: enabling q is current tracking point, and is initialized as starting seed point；
Step 5-3: increase a pixel step length on the primary tangent direction of current point q, and coordinate is rounded to obtain next
A point q1, primary tangent direction definition are to originate seed point to the L2 norm normalization direction for terminating seed point, and principal normal direction
It is defined as the vertical L2 norm normalization direction in primary tangent direction；
Step 5-4: judging current point q, whether has marked pixel on its principal normal direction；
Step 5-5: on q point principal normal direction, constraining according to gray value, tangential direction similitude and taken point it
Whether preceding point is labeled the selection that these three conditions carry out series crack point；
Step 5-6: judge whether the crack point number chosen is zero；If it is zero, show to examine at the crack with front
The point is then skipped compared to being mutated in the crack measured, and update current point q is q1, executes step 5-3；If not zero, then it holds
Row step 5-7；
Step 5-7: a series of median coordinate points of potential points are calculated, i.e., these are on principal normal direction according to coordinate
The coordinate at the midpoint of the point of sequence；
Step 5-8: judge whether the median point meets the condition of gray value constraint and tangential direction similitude；
Step 5-9: this series of potential crack pixel of label is crack pixel；
Step 5-10: judge to need whether the crack detected is detected completely；If the Crack Detection between two seed points is also not
It finishes, then updating current point q is that q1 executes step 5-3；If detection finishes, entire crack is finally identified.
The embodiment of the present invention based on half interactive crack detection method, the seed point taken dependent on user and crack from
Body tendency detects crack and calculates its width.On the one hand, a series of seed points that user takes can guarantee the continuous of Crack Detection
Property；On the other hand, the tendency in crack is portrayed according to the feature vector of Hessian matrix, and crack is extracted according to this trend,
It can guarantee the accuracy of Crack Detection.Noise, illumination, the influence of the unfavorable factors such as background complexity can be effectively eliminated；And according to
Gradient information and directional information criterion extract entire crack pixel；Guarantee the complete of crack tracking according to the selection of potential crack point
Property and continuity, detection accuracy it is high.
The additional aspect of the present invention and advantage will be set forth in part in the description, these will become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment
Attached drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill of field, without any creative labor, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of crack schematic diagram provided in an embodiment of the present invention；
Fig. 2 is that a kind of seed point provided in an embodiment of the present invention chooses schematic diagram；
Fig. 3 is that schematic diagram is tracked in a kind of crack provided in an embodiment of the present invention；
Fig. 4-Fig. 7 is the Crack Detection result schematic diagram under a kind of various application scenarios provided in an embodiment of the present invention；
Fig. 8 is provided in an embodiment of the present invention a kind of based on half interactive Crack Detection block diagram；
Fig. 9 is Crack Detection block diagram between a kind of neighboring seeds point provided in an embodiment of the present invention；
Figure 10 is a kind of crack provided in an embodiment of the present invention from trace detection block diagram.
Specific embodiment
Embodiments of the present invention are described below in detail, the example of the embodiment is shown in the accompanying drawings, wherein from beginning
Same or similar element or element with the same or similar functions are indicated to same or similar label eventually.Below by ginseng
The embodiment for examining attached drawing description is exemplary, and for explaining only the invention, and is not construed as limiting the claims.
Those skilled in the art of the present technique are appreciated that unless expressly stated, singular " one " used herein, " one
It is a ", " described " and "the" may also comprise plural form.It is to be further understood that being arranged used in specification of the invention
Diction " comprising " refer to that there are the feature, integer, step, operation, element and/or component, but it is not excluded that in the presence of or addition
Other one or more features, integer, step, operation, element, component and/or their group.It should be understood that when we claim member
Part is " connected " or when " coupled " to another element, it can be directly connected or coupled to other elements, or there may also be
Intermediary element.In addition, " connection " used herein or " coupling " may include being wirelessly connected or coupling.Wording used herein
"and/or" includes one or more associated any cells for listing item and all combinations.
Those skilled in the art of the present technique are appreciated that unless otherwise defined, all terms used herein (including technology art
Language and scientific term) there is meaning identical with the general understanding of those of ordinary skill in fields of the present invention.Should also
Understand, those terms such as defined in the general dictionary, which should be understood that, to be had and the meaning in the context of the prior art
The consistent meaning of justice, and unless defined as here, it will not be explained in an idealized or overly formal meaning.
In order to facilitate understanding of embodiments of the present invention, it is done by taking several specific embodiments as an example below in conjunction with attached drawing further
Explanation, and each embodiment does not constitute the restriction to the embodiment of the present invention.
The embodiment of the invention provides a kind of based on half interactive crack detection method, and specific implementation flow chart is as schemed
Shown in 8, specifically includes the following steps:
Step 1: seed point is chosen.Fig. 2 is that a kind of seed point provided in an embodiment of the present invention chooses schematic diagram, Yong Huyong
Mouse chooses series of points coordinate on target image and is denoted as (x at cracki,yi)i N=1, wherein x is horizontal direction coordinate, and y is
Vertical direction coordinate.Notice that seed point is chosen to choose as far as possible at the big inflection point in crack, as shown in Figure 2.
Step 2: region of interesting extraction and image preprocessing.In order to accelerate the treatment effeciency of algorithm, crack need to be only extracted
Area image.Specific extracting method is to seek the horizontal direction coordinate value { x of minimax in N number of coordinate points respectivelymin,xmax,
And the vertical direction coordinate value { y of minimaxmin,ymax}.So the left upper apex coordinate of area-of-interest is (xmin,
ymin), and bottom right vertex coordinate is (xmax,ymax).Notice that crack area can also carry out extending out for appropriate pixel, according to the left side
Upper apex coordinate and bottom right vertex coordinate are by drawing the rectangular crack area image of straight-line segment extraction.
Further, crack area image is subjected to bilateral filtering operation, enhances the edge in crack while crossing noise filtering；
Then filtered image is subjected to log operations, and removes mean normalization, increase the contrast in crack and background.
Step 3: the tangential direction and normal direction of pixel are calculated.By taking crack area pixel as an example, the crack area
The normal direction of pixel corresponds to the most violent direction of gray-value variation, and to correspond to gray-value variation gentle for tangential direction
Direction.The tangential direction of pixel in crack area is exactly the tendency in crack, and normal direction vector sum tangential direction vector is just
The feature vector of the Hessian matrix of the pixel is corresponded to well, and basic principle is as follows:
Consider that some pixel p (x, y) in region of interest area image, Hessian matrix are
Wherein:
Wherein I is enhanced region of interest area image in step 2, and g (x, y) is dimensional Gaussian convolution mask,For volume
Product operation, I (x, y) are the matrix centered on pixel p (x, y) and with the sizes such as Gaussian convolution template.Hessian matrix H
The corresponding feature vector n (p) of the absolute value of the maximum eigenvalue of (x, y)=[nx,ny], for the normal direction of the point, and it is minimum special
The corresponding feature vector v (p) of the absolute value of value indicative=[vx,vy], for the tangential direction of the point.If in crack area image
Each pixel goes construction Hessian matrix further to seek feature vector, and operand is especially big.Simultaneously because each picture
The Hessian matrix of vegetarian refreshments is 2 dimension matrixes, and feature vector can perform mathematical calculations to obtain a simple closed solution.Cause
This is extended to matrix operation in practical engineering applications, by the closed solution of single pixel point, and then acquires each pixel
Tangential direction and normal direction.Final direction vector carries out the normalization of L2 norm.
Step 4: enable i for seed point index, i is natural number, i > 0, and is initialized as i=1.
Step 5: by crack from tracking and potential crack point extended method, the tangential direction and normal of pixel are utilized
Direction starts to detect the crack between i-th of seed point and i+1 seed point, and testing process is as shown in Figure 9.Specific steps are such as
Under:
Step 5-1: crack is carried out from tracking from starting seed point, and is being worked as constantly calculating crack from tracing process
The width of preceding point, and mark the crack pixel detected.Crack is as shown in Figure 10 from tracer technique detailed step flow chart, in detail
Carefully it is described as follows:
1) enabling p is the starting point of tracking certainly, and starts the tracking in progress crack with this starting point；
2) using current point p as zero point, the two of crack where determining the point in [0,30] pixel coverage in the direction of its normal
A boundary point, judgment criterion are gradient and tangential direction similarity.The tangential direction for remembering current point p is v (p), then in its method
The directional similarity measurement of any point z and current point p on line direction in [0,30] pixel coverage are s=v (p)T v(z)。
So the response of point z is G (z)=T*s, and wherein T is the gradient of the point.According to the response that these are put, determine that maximum value is corresponding
Coordinate points are the side boundary point in crack where current point.Similarly, in the anti-normal direction of current point p [- 30,0] pixel coverage
Interior determining other side boundary point, Fig. 3 are that schematic diagram is tracked in a kind of crack provided in an embodiment of the present invention, as shown in figure 3, wherein as
Plain range can be set according to actual conditions, and for Crack Detection, it is reasonable that 30 pixels, which are arranged,.The Europe of two boundary point coordinates
Family name's distance is width of the crack in the point；And all pixels point between two boundary points is the crack that detects and to it
It is marked；Final updating current point p is the central point between two boundary points.
Wherein p point is the variable being arranged from tracing process；And q point is to start seed point to set to seed point tracing process is terminated
The variable set；Pay attention to starting seed point to terminating in seed point tracing process, can constantly call from tracing process, such as step
Five -8, and p point is each time from the starting point of tracking.
3) increase a pixel step length in the tangential direction of current point p, and coordinate is rounded to obtain next point p1；Note
It is identical to be intended to processing method on the opposite direction of tangent line, as shown in Figure 3.Point tracking is carried out on the both forward and reverse directions of tangent line to guarantee to split
Stitch the integrality of tracking.
4) judge whether tangential direction and the tangential direction of next point p1 of current point p are similar；Whether point p1 is detecting
Inside crack, i.e., whether between two seed points；Whether p1 point is not labeled.Notice that similarity measurement here is equally adopted
It is similar if it is greater than 0.9 two o'clock with the inner product of the tangential direction vector of two o'clock.Basic principle is any curve very adjacent
(two o'clock here is two adjacent pixels) is straight line between close two o'clock, and the tangential direction of all the points is consistent in straight line.
A point is judged whether between two seed points, is the direction by judgement starting seed point to the point and the point to termination kind
Whether the angle between the direction of son point is less than 90 ° or whether the inner product in direction is greater than 0.
If 5) p1 point meets above three condition, updating current point p is p1 point, step 2) is executed, if being unsatisfactory for holding
Row step 6).
6) increase by two pixel step lengths in the tangential direction of current point p, and coordinate is rounded to obtain next point p2；Note
It is identical to be intended to processing method on the opposite direction of tangent line.If p1 is unsatisfactory for condition, p1 point may be noise spot.
7) similarly judge whether p2 point meets above three condition.
If 8) p2 point meets above three condition, updating current point p is p2 point, step 2) is executed, if not satisfied, then
Certainly termination is tracked in crack.If p2 is unsatisfactory for condition, all there is tangential direction mutation, shows song in latter two point of current point p
Line tracking terminates；If increasing by three pixel step lengths by force, crack may be deviateed from tracking.
Step 5-2: enabling q is current tracking point, and is initialized as starting seed point.
Step 5-3: increase a pixel step length on the primary tangent direction of current point q, and coordinate is rounded to obtain next
A point q1.Primary tangent direction definition is to originate seed point to the L2 norm normalization direction for terminating seed point, and principal normal direction
It is defined as the vertical L2 norm normalization direction in primary tangent direction, as shown in Figure 3.
Step 5-4: judging current point q, whether has marked pixel on its principal normal direction.In current point q
Principal normal direction in [- 30,30] pixel coverage, if having the crack pixel being detected.If it has, then where the point
Crack be q1 from trace detection to, update current point q in step 5-1, execute step 5-3；If it has not, the then point
Place crack area is not detected also, as the terminal point from tracking, thens follow the steps five -5.
Step 5-5: on q point, principal normal direction in a certain range, constraining, tangential direction similitude according to gray value,
The selection that these three conditions carry out series crack point whether is labeled with point before taken point.If showing to split at this point
Seam is terminated from tracking, i.e., the tangential direction at the point is mutated compared with the tangential direction of tracking point before, so at this time
Tangential direction similitude is judged using the tangential direction of candidate point with primary tangent direction.Therefore, seed point selection exists as far as possible
It is chosen at the big inflection point in crack, it is more tight from tracing process condition although crack is from tracking energy automatic tracing knee of curve
Lattice, it is likely that there is noise spot interference, the factors such as direction mutation and track to go down.Gray value is constrained to the gray value of selected point
It is less than given threshold value with from the difference between the average gray value of the crack point of tracking.Last condition is to judge that candidate point is cut in master
Whether labeled point is had in the certain pixel coverage in line direction, and the condition is the most key, is directly related to the complete of Crack Detection
Property and continuity.Use the basic thought of this condition for the successional priori in crack, which will obtain a series of crack
Point.
Step 5-6: judge whether the crack point number chosen is zero.If it is zero, show at the crack at least with it is preceding
The point is then skipped compared to being mutated in the crack that face detects, update current point q is q1, executes step 5-3.If not zero,
Then follow the steps five -7.
Step 5-7: a series of median coordinate points of potential points are calculated, i.e., these are on principal normal direction according to coordinate
The coordinate at the midpoint of the point of sequence.
Step 5-8: judge whether the median point meets gray value constraint and tangential direction similitude.Here tangent line side
Still judged to similitude using with primary tangent direction.If meeting condition, show that the point is reliable inside crack
Point carries out tracking certainly for crack using the point as starting point.It is worth noting that, being direction tracing from tracking, step can overcome the disadvantages that
The series of points skipped in five -6.It has executed from after tracking (flow chart of steps is as shown in Figure 10), and has updated current point q and held for q1
Row step 5-3.If being unsatisfactory for condition, five -9 are thened follow the steps.
Step 5-9: this series of potential crack pixel of label is crack pixel.Due to crack this step mark not
Be it is very stringent, point out and do not calculate the width in crack at this.
Step 5-10: judge to need whether the crack detected is detected completely.If the Crack Detection between two seed points is also not
It finishes, then updating current point q is that q1 executes step 5-3.If detection finishes, entire crack is finally identified.
Step 6: judge whether seed point detects and finish.If NO, then i=i+1 is enabled, executes step 5 again；If
Judging result be it is yes, then extract a complete crack, and obtain the width in crack.
In conclusion crack detection method of the embodiment of the present invention based on half interactive mode, the seed point taken dependent on user
To detect crack and its width is calculated with itself tendency in crack.On the one hand, a series of seed points that user takes can guarantee crack
The continuity of detection；On the other hand, the tendency in crack is portrayed according to the feature vector of Hessian matrix, and according to this trend
Crack is extracted, can guarantee the accuracy of Crack Detection.Noise, illumination, the shadow of the unfavorable factors such as background complexity can be effectively eliminated
It rings；And entire crack pixel is extracted according to gradient information and directional information criterion；It is split according to the selection of potential crack point guarantee
The integrality and continuity of tracking are stitched, detection accuracy is high.
The present invention is perfectly combined half interactive mode of user with according to the tendency in crack, can not only fight variability
The interference of background, and the application scenarios of various complexity can be widely used in, and can guarantee the requirements for high precision of Crack Detection.
The method of the present invention is based on half interaction techniques, low to the requirement of image image quality, has adaptive ability to image capture environment；
With universality, it is suitable for tunnel slot, pavement crack, optical glass crack, wall cracks, a variety of actual fields such as plate crack
Scape, a kind of testing result provided in an embodiment of the present invention is as shown in figs. 4-7.
The present invention is similar with related work before, has also used the tendency that Hessian matrix portrays crack, but difference
It is in carrying out vee crack the invention proposes tangential direction similarity measurement and two-way principle, rather than cutting in current point
Unidirectional blindness expands to next point on line direction；It is sought according to gradient information and directional information criterion rather than by extreme point
Boundary is looked for, ensure that the accuracy on boundary, typically finds Local Extremum and band because differential method is affected by noise larger
Carry out detection error；In addition, going out in crack from tracking breakpoint, propose that a series of criterion carry out the selection of potential crack point, and guarantee
From the continuation of tracking, to realize the integrality and accuracy of Crack Detection.
Those of ordinary skill in the art will appreciate that: attached drawing is the schematic diagram of one embodiment, module in attached drawing or
Process is not necessarily implemented necessary to the present invention.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
Subject to.
Claims (3)
1. a kind of based on half interactive crack detection method, which comprises the following steps:
Step 1: a series of seed points are chosen at crack on target image, obtain the coordinate of seed point；
Step 2: crack area image is extracted to the target image, the crack area image is pre-processed, is increased
The contrast of crack area image and background；
Step 3: the tangential direction and normal direction of each pixel are calculated pretreated crack area image；
Step 4: being numbered the seed point of selection, enables i for seed point index, i is natural number, i > 0, and is initialized as i=
1；
Step 5: by crack from tracking and potential crack point extended method, the tangential direction and normal direction of pixel are utilized
To detect the crack between i-th of seed point and i+1 seed point；
Step 6: judging whether all seed points detect and finish, if it is judged that be it is no, then enable i=i+1, execute step again
Rapid five；If it is judged that be it is yes, then extract a complete crack, and obtain the width in crack；
Tangential direction and the normal side for calculating pretreated crack area image each pixel in the step 3
To, comprising:
The tangential direction and normal direction of pixel correspond to the feature vector of the Hessian matrix of the pixel, consider crack area
Some pixel p (x, y) in area image, Hessian matrix are as follows:
Wherein:
Wherein I is enhanced crack area image in step 2, and g (x, y) is dimensional Gaussian convolution mask,For convolution fortune
It calculates, I (x, y) is the matrix centered on pixel p (x, y) and with the sizes such as Gaussian convolution template, Hessian matrix H (x, y)
Maximum eigenvalue the corresponding feature vector n (p) of absolute value=[nx,ny], for the normal direction of pixel p (x, y), and
The corresponding feature vector v (p) of the absolute value of minimal eigenvalue=[vx,vy], for the tangential direction of pixel p (x, y)；
In the step five by crack from tracking and potential crack point extended method, using pixel tangential direction and
Normal direction detects the crack between i-th of seed point and i+1 seed point, comprising the following steps:
Step 5-1: crack is carried out from tracking from starting seed point, and is constantly calculating crack in current point from tracing process
Width, and mark the crack pixel that detects；From tracer technique, specific step is as follows in crack:
1) enabling p is the starting point of tracking certainly, and starts the tracking in progress crack with this starting point；
2) using current point p as zero point, two sides in crack where determining the point in [0,30] pixel coverage in the direction of its normal
Boundary's point, judgment criterion are gradient and tangential direction similarity；All pixels point between two boundary points is the crack detected,
And the crack is marked；Final updating current point p is the central point between two boundary points；
3) increase a pixel step length in the tangential direction of current point p, and coordinate is rounded to obtain next point p1；In tangent line
Opposite direction on processing method it is identical；
4) judge whether tangential direction and the tangential direction of next point p1 of current point p are similar；Whether point p1 is in fracture detection
Inside, i.e., whether between two seed points；Whether p1 point is not labeled；
If 5) p1 point meets above three condition, updating current point p is p1 point, executes step 2), executes if being unsatisfactory for
Step 6)；
6) increase by two pixel step lengths in the tangential direction of current point p, and coordinate is rounded to obtain next point p2；In tangent line
Opposite direction on processing method it is identical；
7) same to step 4), judges whether p2 point meets following three conditions: whether point p2 is inside fracture detection, i.e., two
Between a seed point；Whether p2 point is not labeled；
If 8) p2 point meets above three condition, updating current point p is p2 point, step 2) is executed, if not satisfied, then crack
It is terminated from tracking；
Step 5-2: enabling q is current tracking point, and is initialized as starting seed point；
Step 5-3: increase a pixel step length on the primary tangent direction of current point q, and coordinate is rounded to obtain next point
Q1, primary tangent direction definition are to originate seed point to the L2 norm normalization direction for terminating seed point, and principal normal direction definition
Direction is normalized for the vertical L2 norm of main tangential direction；
Step 5-4: judging current point q, whether has marked pixel on its principal normal direction；
Step 5-5: on q point principal normal direction, constraining according to gray value, the point before of tangential direction similitude and taken point
Whether selection that these three conditions carry out series crack point is labeled；
Step 5-6: judge whether the crack point number chosen is zero；If it is zero, show at the crack with it is previously detected
Crack compared to being mutated, then skip the point, updates current point q is q1, execution step 5-3；If not zero, then execute step
Rapid five -7；
Step 5-7: calculating a series of median coordinate points of potential points, i.e., these sort on principal normal direction according to coordinate
Point midpoint coordinate；
Step 5-8: judge whether the median point meets the condition of gray value constraint and tangential direction similitude；
Step 5-9: this series of potential crack pixel of label is crack pixel；
Step 5-10: judge to need whether the crack detected is detected completely；If the Crack Detection between two seed points is not also complete
Finish, then updating current point q is that q1 executes step 5-3；If detection finishes, entire crack is finally identified.
2. the method according to claim 1, wherein being split to the target image extraction in the step 2
Stitching area image includes:
If the sum for the seed point chosen is N, the abscissa of the horizontal direction of i-th of seed point is xiThe ordinate of vertical direction
For yi；
The smallest horizontal direction coordinate value x in N number of seed point is sought respectivelymin, maximum horizontal direction coordinate value xmax, the smallest perpendicular
Histogram is to coordinate value ymin, maximum vertical direction coordinate value ymax, the left upper apex coordinate for extracting crack area is (xmin,
ymin), and bottom right vertex coordinate is (xmax,ymax), it is extracted according to the left upper apex coordinate and bottom right vertex coordinate rectangular
Crack area image.
3. the method according to claim 1, wherein being carried out in the step 2 to the crack area image
Pretreatment, the contrast for increasing crack area image and background include:
Bilateral filtering operation is carried out to the crack area image, log operations are carried out to filtered crack area image, and
Mean normalization is removed, the contrast of crack area image and background is increased.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710558215.9A CN107389697B (en) | 2017-07-10 | 2017-07-10 | A kind of crack detection method based on half interactive mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710558215.9A CN107389697B (en) | 2017-07-10 | 2017-07-10 | A kind of crack detection method based on half interactive mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107389697A CN107389697A (en) | 2017-11-24 |
| CN107389697B true CN107389697B (en) | 2019-08-30 |
Family
ID=60339014
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710558215.9A Active CN107389697B (en) | 2017-07-10 | 2017-07-10 | A kind of crack detection method based on half interactive mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107389697B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109063597A (en) * | 2018-07-13 | 2018-12-21 | 北京科莱普云技术有限公司 | Method for detecting human face, device, computer equipment and storage medium |
| CN110060243B (en) * | 2019-04-11 | 2022-06-28 | 国家电网有限公司 | Dam surface crack framework extraction method |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100570274C (en) * | 2008-03-18 | 2009-12-16 | 飞秒光电科技(西安)有限公司 | The building tear detection method |
| CN101893580B (en) * | 2010-06-10 | 2012-01-11 | 北京交通大学 | Digital image based detection method of surface flaw of steel rail |
| US8488852B2 (en) * | 2011-02-15 | 2013-07-16 | Siemens Aktiengesellschaft | Medical imaging system for segementing blood vessel |
| DE102011105182A1 (en) * | 2011-06-17 | 2012-12-20 | Albert-Ludwigs-Universität Freiburg | A method of providing a predictive model for crack detection and a method of crack detection on a semiconductor structure |
| US20130088707A1 (en) * | 2011-10-05 | 2013-04-11 | Behzad Imani | Method and system for crack detection |
| CN102663391B (en) * | 2012-02-27 | 2015-03-25 | 安科智慧城市技术（中国）有限公司 | Image multifeature extraction and fusion method and system |
| CN103077512B (en) * | 2012-10-18 | 2015-09-09 | 北京工业大学 | Based on the feature extracting and matching method of the digital picture that major component is analysed |
| CN104504388B (en) * | 2014-12-17 | 2017-10-24 | 长安大学 | A kind of pavement crack identification and feature extraction algorithm and system |
| CN105719283A (en) * | 2016-01-18 | 2016-06-29 | 苏州科技学院 | Road surface crack image detection method based on Hessian matrix multi-scale filtering |
| CN106600600A (en) * | 2016-12-26 | 2017-04-26 | 华南理工大学 | Wafer defect detection method based on characteristic matching |
| CN106645205A (en) * | 2017-02-24 | 2017-05-10 | 武汉大学 | Unmanned aerial vehicle bridge bottom surface crack detection method and system |
-
2017
- 2017-07-10 CN CN201710558215.9A patent/CN107389697B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107389697A (en) | 2017-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Automatic pavement crack detection by multi-scale image fusion | |
| CN105761271B (en) | Nuclear containment surface defect automatic testing method and system | |
| US11580647B1 (en) | Global and local binary pattern image crack segmentation method based on robot vision | |
| Reid et al. | A semi-automated methodology for discontinuity trace detection in digital images of rock mass exposures | |
| CN105373135B (en) | A kind of method and system of aircraft docking guidance and plane type recognition based on machine vision | |
| CN103246883B (en) | A kind of underground coal mine thermal infrared images face identification method | |
| Siriborvornratanakul | An automatic road distress visual inspection system using an onboard in-car camera | |
| CN108596165B (en) | Road traffic marking detection method and system based on unmanned plane low latitude Aerial Images | |
| CN106447669A (en) | Circular masking-out area rate determination-based adhesive particle image concave point segmentation method | |
| CN101499168B (en) | Structured light strip center extraction method based on ridge line tracing and Hessian matrix | |
| CN104599286B (en) | A kind of characteristic tracking method and device based on light stream | |
| CN110246130A (en) | Based on infrared and visible images data fusion airfield pavement crack detection method | |
| CN108154159B (en) | A kind of method for tracking target with automatic recovery ability based on Multistage Detector | |
| CN102609934A (en) | Multi-target segmenting and tracking method based on depth image | |
| CN104809433A (en) | Zebra stripe detection method based on maximum stable region and random sampling | |
| CN104268872A (en) | Consistency-based edge detection method | |
| CN107729853A (en) | A kind of automatic identifying method suitable for the narrow tuning drive gear formula instrument of transformer station | |
| CN107085723A (en) | A kind of characters on license plate global recognition method based on deep learning model | |
| CN107389697B (en) | A kind of crack detection method based on half interactive mode | |
| CN105243356A (en) | Method of building pedestrian detection model and device and pedestrian detection method | |
| CN103914829B (en) | Method for detecting edge of noisy image | |
| CN110084830A (en) | A kind of detection of video frequency motion target and tracking | |
| CN115035141A (en) | Tunnel crack remote monitoring and early warning method based on image processing | |
| CN110210428A (en) | A kind of smog root node detection method under remote complex environment based on MSER | |
| Harvey | Performance evaluation for road extraction |
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 |