CN110163842A - Building cracks detection method, device, computer equipment and storage medium - Google Patents
Building cracks detection method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110163842A CN110163842A CN201910298780.5A CN201910298780A CN110163842A CN 110163842 A CN110163842 A CN 110163842A CN 201910298780 A CN201910298780 A CN 201910298780A CN 110163842 A CN110163842 A CN 110163842A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- crack
- architectural drawing
- connected domain
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30132—Masonry; Concrete
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
This application involves a kind of building cracks detection method, device, computer equipment and storage mediums.The described method includes: obtaining the Architectural drawing of target construction；Architectural drawing input machine learning model is detected, FRACTURE CHARACTERISTICS image is obtained；Binary conversion treatment is carried out to the FRACTURE CHARACTERISTICS image, obtains binaryzation FRACTURE CHARACTERISTICS image；Crack connected domain is extracted from the binaryzation FRACTURE CHARACTERISTICS image；Determine the target construction with the presence or absence of crack by the crack connected domain.This method combines machine learning model and connected domain to be detected twice to reduce error detection in the detection process, improves accuracy in detection.
Description
Technical field
This application involves detection technique fields, more particularly to a kind of building cracks detection method, device, computer equipment
And storage medium.
Background technique
With the development of detection technique, more and more field of industrial production start quality awareness detection.Bridge, dykes and dams,
The buildings such as house and people are closely bound up, when these building service lives are too long or build guality is bad it is possible that building
Crack leaves hidden trouble to the safety of people, therefore particularly important to the detection of building cracks.
However, traditional building cracks detection scheme is the Architectural drawing of photographic subjects building, Architectural drawing is carried out
The processing of various view-based access control model technologies, when the ambient conditions complexity where target construction, whether detection target construction is deposited
Accuracy in crack is lower.
Summary of the invention
Based on this, it is necessary in view of the above technical problems, provide a kind of building cracks inspection that can be improved accuracy in detection
Survey method, apparatus, computer equipment and storage medium.
A kind of building cracks detection method, which comprises
Obtain the Architectural drawing of target construction；
Architectural drawing input machine learning model is detected, FRACTURE CHARACTERISTICS image is obtained；
Binary conversion treatment is carried out to the FRACTURE CHARACTERISTICS image, obtains binaryzation FRACTURE CHARACTERISTICS image；
Crack connected domain is extracted from the binaryzation FRACTURE CHARACTERISTICS image；
Determine the target construction with the presence or absence of crack by the crack connected domain.
A kind of building cracks detection device, described device include:
Image collection module, for obtaining the Architectural drawing of target construction；
Image detection module obtains FRACTURE CHARACTERISTICS for detecting Architectural drawing input machine learning model
Image；
Image processing module obtains binaryzation FRACTURE CHARACTERISTICS for carrying out binary conversion treatment to the FRACTURE CHARACTERISTICS image
Image；
Connected domain extraction module, for extracting crack connected domain from the binaryzation FRACTURE CHARACTERISTICS image；
Crack determining module, for determining the target construction with the presence or absence of crack by the crack connected domain.
A kind of computer equipment can be run on a memory and on a processor including memory, processor and storage
Computer program, the processor perform the steps of when executing the computer program
Obtain the Architectural drawing of target construction；
Architectural drawing input machine learning model is detected, FRACTURE CHARACTERISTICS image is obtained；
Binary conversion treatment is carried out to the FRACTURE CHARACTERISTICS image, obtains binaryzation FRACTURE CHARACTERISTICS image；
Crack connected domain is extracted from the binaryzation FRACTURE CHARACTERISTICS image；
Determine the target construction with the presence or absence of crack by the crack connected domain.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
It is performed the steps of when row
Obtain the Architectural drawing of target construction；
Architectural drawing input machine learning model is detected, FRACTURE CHARACTERISTICS image is obtained；
Binary conversion treatment is carried out to the FRACTURE CHARACTERISTICS image, obtains binaryzation FRACTURE CHARACTERISTICS image；
Crack connected domain is extracted from the binaryzation FRACTURE CHARACTERISTICS image；
Determine the target construction with the presence or absence of crack by the crack connected domain.
Above-mentioned building cracks detection method, device, computer equipment and storage medium, obtain the architectural drawing of target construction
As after, Architectural drawing is inputted into machine learning model, the is carried out to the FRACTURE CHARACTERISTICS in Architectural drawing by machine learning model
One-time detection obtains FRACTURE CHARACTERISTICS image；Fracture characteristic image carries out binary conversion treatment again, obtains binaryzation FRACTURE CHARACTERISTICS figure
Picture extracts crack connected domain from binaryzation FRACTURE CHARACTERISTICS image；Second is carried out by crack connected domain to detect, and determines target
Building whether there is crack, combine machine learning model and connected domain to be detected twice to reduce mistake in the detection process
Detection, improves accuracy in detection.
Detailed description of the invention
Fig. 1 is the applied environment figure of building cracks detection method in one embodiment；
Fig. 2 is the flow diagram of building cracks detection method in one embodiment；
Fig. 3 is flow diagram the step of extracting crack connected domain in one embodiment；
Fig. 4 is flow diagram the step of determining crack connected domain by seed filling in one embodiment；
Fig. 5 is flow diagram the step of determining whether there is crack by crack connected domain in one embodiment；
Fig. 6 is flow diagram the step of determining whether there is crack by area in one embodiment；
The flow diagram for the step of Fig. 7 is training machine learning model in one embodiment；
Fig. 8 is flow diagram the step of constructing multiple groups Architectural drawing sample in one embodiment；
Fig. 9 is the schematic diagram of the multiple groups Architectural drawing sample of server construction in one embodiment；
Figure 10 is the schematic diagram of building cracks detection in one embodiment；
Figure 11 is the structural block diagram of building cracks detection device in one embodiment；
Figure 12 is the internal structure chart of computer equipment in one embodiment.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not
For limiting the application.
Building cracks detection method provided by the present application, can be applied in application environment as shown in Figure 1.Wherein, eventually
End 102 can be communicated by network with server 104, can also be communicate through a serial port.Wherein, terminal 102 can be with
But it is not limited to various cameras, video camera, the unmanned plane equipped with camera, can also be various with image collecting device
Industrial computer, personal computer, laptop, smart phone, tablet computer and portable wearable device；Server
104 can be realized with the server cluster of the either multiple server compositions of independent server.
In one embodiment, as shown in Fig. 2, providing a kind of building cracks detection method, it is applied to Fig. 1 in this way
In server for be illustrated, comprising the following steps:
Step 202, the Architectural drawing of target construction is obtained.
Wherein, target construction can be detected building.Architectural drawing can be photographic subjects building and obtain
Image.
Specifically, terminal needs detected position to take pictures target construction, obtains Architectural drawing.Terminal can be with
Collected Architectural drawing is sent to server by network.The Architectural drawing that server receiving terminal is sent.
In one embodiment, target construction can be various types of buildings, such as bridge, dykes and dams and house
Deng.Testing staff acquires target construction surface (including outer surface and interior by the unmanned plane that camera is carried in artificial or manipulation
Surface) image obtain Architectural drawing.
Step 204, Architectural drawing input machine learning model is detected, obtains FRACTURE CHARACTERISTICS image.
Wherein, FRACTURE CHARACTERISTICS image can be the architectural drawing comprising FRACTURE CHARACTERISTICS detected by machine learning model
Picture.
Specifically, there are trained machine learning models in server.The Architectural drawing input that server will acquire
Machine learning model is handled Architectural drawing by machine learning model.Machine learning model detects in Architectural drawing
FRACTURE CHARACTERISTICS when, fracture feature is marked, and obtains FRACTURE CHARACTERISTICS image.After detection, machine learning model output
FRACTURE CHARACTERISTICS image.
In one embodiment, machine learning model can be convolutional neural networks (Convolutional Neural
Networks,CNN)。
Step 206, fracture characteristic image carries out binary conversion treatment, obtains binaryzation FRACTURE CHARACTERISTICS image.
Wherein, binaryzation FRACTURE CHARACTERISTICS image can be fracture characteristic image and carry out the image that binary conversion treatment obtains.
Specifically, after server obtains the FRACTURE CHARACTERISTICS image of machine learning model output, fracture characteristic image is carried out
Binary conversion treatment obtains binaryzation FRACTURE CHARACTERISTICS image.The effect for there was only black and white is presented in binaryzation FRACTURE CHARACTERISTICS image,
Wherein white portion can be the FRACTURE CHARACTERISTICS detected, and black portions can be background.
In one embodiment, when server carries out binary conversion treatment, Otsu algorithm, Kittler most mistake can be used
Any one in misclassification method, average gray method and threshold method based on bimodal average value.
Step 208, crack connected domain is extracted from binaryzation FRACTURE CHARACTERISTICS image.
Wherein, crack connected domain can be the connected domain being made of the pixel of characterization FRACTURE CHARACTERISTICS.
Specifically, in order to avoid picture noise leads to the erroneous judgement generated, server needs to detect machine learning model
FRACTURE CHARACTERISTICS carry out second and handle.When server carries out second of processing, characterized according in binaryzation FRACTURE CHARACTERISTICS image
The pixel extraction crack connected domain of FRACTURE CHARACTERISTICS.
Step 210, determine target construction with the presence or absence of crack by crack connected domain.
Specifically, when crack when the FRACTURE CHARACTERISTICS that machine learning model detects in target construction, crack is special
It levies corresponding crack connected domain and meets connected domain Rule of judgment.After server extracts crack connected domain, judged according to connected domain
Condition judges the crack connected domain extracted, and when crack, connected domain meets connected domain Rule of judgment, server is determined
There are cracks in target construction.
In the present embodiment, after the Architectural drawing for obtaining target construction, Architectural drawing is inputted into machine learning model, is passed through
Machine learning model carries out first time detection to the FRACTURE CHARACTERISTICS in Architectural drawing, obtains FRACTURE CHARACTERISTICS image；Fracture is special again
It levies image and carries out binary conversion treatment, obtain binaryzation FRACTURE CHARACTERISTICS image, crack is extracted from binaryzation FRACTURE CHARACTERISTICS image and is connected
Logical domain；Second is carried out by crack connected domain to detect, and determines that target construction with the presence or absence of crack, combines in the detection process
Machine learning model and connected domain are detected twice to reduce error detection, and accuracy in detection is improved.
As shown in figure 3, in one embodiment, step 208 specifically further includes the steps that extracting crack connected domain, the step
Specifically comprise the following steps:
Step 302, Morphological scale-space is carried out to binaryzation FRACTURE CHARACTERISTICS image.
Specifically, server first splits binaryzation before extracting crack connected domain in binaryzation FRACTURE CHARACTERISTICS image
It stitches characteristic image and carries out Morphological scale-space.Server first can carry out expansive working to binaryzation FRACTURE CHARACTERISTICS image, then carry out
Etching operation, to remove spotted noise spuious in binaryzation FRACTURE CHARACTERISTICS image.
Step 304, each pixel in binaryzation FRACTURE CHARACTERISTICS image after traversing through Morphological scale-space.
Specifically, after the processing of server completion morphology, in the binaryzation FRACTURE CHARACTERISTICS image after traversing Morphological scale-space
Each pixel.Server can establish image coordinate system to the binaryzation FRACTURE CHARACTERISTICS image after Morphological scale-space, from origin
Begin stepping through each pixel.
Step 306, true according to the pixel for meeting default fracture conditions when the pixel traversed meets default fracture conditions
Determine crack connected domain.
Wherein, default fracture conditions can be the condition met when pixel characterization FRACTURE CHARACTERISTICS.
Specifically, whether server characterizes FRACTURE CHARACTERISTICS according to the pixel that default fracture conditions judgement traverses, and specifically may be used
Whether the pixel value to be the pixel that judgement traverses meets default fracture conditions.When white portion in binaryzation FRACTURE CHARACTERISTICS image
When point (pixel value 1) is FRACTURE CHARACTERISTICS, black portions (pixel value 0) are background, default fracture conditions can be pixel value
It is 1.When the pixel value of the pixel traversed is 1, which characterizes FRACTURE CHARACTERISTICS.
When the pixel traversed meets default fracture conditions, server carries out seed filling according to the pixel, and seed is filled out
After filling, crack connected domain is obtained.
In one embodiment, server can extract one from the binaryzation FRACTURE CHARACTERISTICS image after Morphological scale-space
A crack connected domain can also extract multiple crack connected domains.
In the present embodiment, Morphological scale-space first is carried out with the influence of less noise to binaryzation FRACTURE CHARACTERISTICS image, is improved
The accuracy of image procossing.Each pixel in binaryzation FRACTURE CHARACTERISTICS image after traversing Morphological scale-space, it is pre- according to meeting
If the pixel of fracture conditions determines crack connected domain, avoids omission and extract crack connected domain.
As shown in figure 4, in one embodiment, step 306 further includes specifically determining crack connected domain by seed filling
The step of, which specifically comprises the following steps:
Step 402, the pixel for meeting default fracture conditions is entered as sub-pixel, and by the coordinate of sub-pixel
Stack.
Wherein, initial pixel needed for sub-pixel can be progress seed filling；Seed filling can be from seed picture
Element starts, and searches the process that more multiple coincidence presets the pixel of fracture conditions.
Specifically, when the pixel traversed meets default fracture conditions, which is labeled as sub-pixel by server,
It obtains the coordinate of sub-pixel and the coordinate of sub-pixel is subjected to stacking.
Stack (stack) is called storehouse, is a kind of linear list that operation is limited.One end that stack is only allowed in table carry out insertion and
Pop-up, this one end is stack top, and the other end is stack bottom.
Step 404, the pixels of default fracture conditions will be met in the neighborhood of sub-pixel as fillable pixel, and can
The coordinate of filler pixels carries out stacking.
Wherein, fillable pixel is the pixel that can be used as sub-pixel and carry out seed filling.
Specifically, server is searched the pixel for also corresponding to default fracture conditions, will be searched in the neighborhood of sub-pixel
To the pixel for meeting default fracture conditions as fillable pixel.Neighborhood when server is searched can be four neighborhoods and eight neighbours
Any one in domain.Server obtains the coordinate of fillable pixel, and coordinate is pressed into stack.
Step 406, pop-up is located at the coordinate of stack top.
Specifically, when server is inserted into new coordinate to a stack, it is that new coordinate is put into the upper surface of stack top coordinate, is allowed into
For new stack top coordinate；Coordinate is popped up from stack, is that stack top coordinate is deleted, and the stack top that the coordinate for keeping its adjacent becomes new is sat
Mark.For server by after the coordinate stacking of fillable pixel, pop-up is located at the coordinate of stack top.
Step 408, fillable pixel is continued to search using pixel corresponding to the coordinate of pop-up as sub-pixel, until stack
Interior is empty.
Specifically, server is using pixel corresponding to the coordinate of pop-up as new sub-pixel, in new sub-pixel
Neighborhood in search fillable pixel, and by the coordinate stacking of the fillable pixel found.Server constantly pops up stack top
Coordinate simultaneously searches fillable pixel, and when that can not find fillable pixel again, the coordinate quantity in stack is with stack top coordinate
Constantly pop-up gradually decreases, and when being there is no when coordinate in empty i.e. stack in stack, seed filling terminates.
Step 410, pixel region corresponding to each coordinate by pop-up is determined as crack connected domain.
Wherein, pixel region can be image-region corresponding to the coordinate popped up in stack.
Specifically, pixel corresponding to all coordinates popped up out of stack forms pixel region, and server is by the pixel region
Domain is determined as crack connected domain.
In one embodiment, server is using the pixel for meeting default fracture conditions traversed as sub-pixel, and
When by the coordinate stacking of sub-pixel, label (label) is added to coordinate；Server carries out seed filling according to sub-pixel
When, the coordinate of whole stackings is added into identical label.After a seed filling, server will have same label
Coordinate corresponding to pixel region be determined as crack connected domain.When server traverses pixel, if the pixel traversed has been deposited
In label, then no longer pixel is handled, to avoid repeating to extract connected domain.
In the present embodiment, the coordinate that will meet the pixel of default fracture conditions carries out stacking, pops up stack top coordinate simultaneously one by one
The pixel for meeting default fracture conditions is continued to search, by the linear list that this operation of stack is limited, can not be found with omitting
All meet the pixel of default fracture conditions, and crack connected domain is determined according to the pixel found, improves and extract crack company
The accuracy in logical domain.
As shown in figure 5, in one embodiment, step 210 further includes specifically being determined whether there is by crack connected domain
The step of crack, the step specifically comprise the following steps:
Step 502, the quantitative value of pixel in the connected domain of crack is counted.
Specifically, server is after extracting crack connected domain in binaryzation FRACTURE CHARACTERISTICS image, counts splitting of extracting
The quantity for stitching pixel in connected domain, obtains the quantitative value of pixel in the connected domain of crack.
Step 504, when quantitative value is greater than present count magnitude, the boundary rectangle of crack connected domain is determined.
Wherein, present count magnitude can be preset quantitative value.Boundary rectangle, which can be, is external in crack connected domain
Rectangle.
Specifically, server obtains the present count magnitude of storage, and present count magnitude is judged for fracture connected domain,
To remove lesser dotted, Speckle noise the influence of area in binaryzation FRACTURE CHARACTERISTICS image.When pixel in the connected domain of crack
When quantitative value is less than or equal to present count magnitude, determine that this crack connected domain is background.When the quantity of pixel in the connected domain of crack
When value is greater than present count magnitude, the boundary rectangle of crack connected domain is determined in binaryzation FRACTURE CHARACTERISTICS image.
In one embodiment, when quantitative value is greater than present count magnitude, server determines that the minimum of crack connected domain is outer
Meet rectangle (minimum bounding rectangle, MBR).
Step 506, the rectangular area and length-width ratio of boundary rectangle are calculated.
Wherein, length-width ratio can be the length of boundary rectangle and the ratio of width.
Specifically, after server determines the boundary rectangle of crack connected domain, using longer side as the length of boundary rectangle,
Using shorter side as the width of boundary rectangle, and the ratio of computational length and width, obtain length-width ratio.
Step 508, when length-width ratio is greater than or equal to the first default ratio, determining target construction, there are cracks.
Wherein, the first default ratio can be the ratio of preset length and width, for whether judging target construction
There are cracks.
Specifically, server obtains the first default ratio of storage, and length-width ratio and the first default ratio are compared.Outside
It is related to crack connected domain to connect rectangular shape, when length-width ratio is greater than or equal to the first default ratio, characterizes external square
Shape is more long and narrow, and server determines crack of the crack connected domain in target construction, i.e., there are cracks in target construction.
In the present embodiment, first judged according to the quantitative value of pixel in the connected domain of crack to remove the shadow of partial noise
It rings, it is ensured that the accuracy of detection.The shape of boundary rectangle is related to crack connected domain, by comparing boundary rectangle length-width ratio and
The size of first default ratio is realized from determining that target construction whether there is crack in shape, improves the standard of Crack Detection
True property.
As shown in fig. 6, in one embodiment, step 210 further includes the steps that determining whether there is crack by area,
The step specifically comprises the following steps:
Step 602, when length-width ratio is less than the first default ratio, the ratio between rectangular area and quantitative value is calculated.
Wherein, rectangular area can be the area of boundary rectangle.
Specifically, it when the length-width ratio of boundary rectangle is less than the first default ratio, needs further to be judged.Service
Device calculates the area of boundary rectangle, and the pixel value of pixel can be used as the area of crack connected domain, server in the connected domain of crack
Calculate the ratio of rectangular area and quantitative value.
Step 604, when ratio is greater than the second default ratio, determining target construction, there are cracks.
Wherein, the second default ratio can be the ratio between preset rectangular area and quantitative value.
Specifically, length-width ratio may be from the biggish bulk noise of area less than the crack connected domain of the first boundary rectangle.
When rectangular area and the ratio of quantitative value are greater than the second default ratio, server excludes the influence of bulk noise, determines target
There are cracks in building；When rectangular area and the ratio of quantitative value are less than the second default ratio, this crack connected domain is determined
For background.
In the present embodiment, the crack connected domain that length-width ratio is less than default ratio may be from noise, according to rectangular area with
The ratio of quantitative value is judged again, and when ratio is greater than the second default ratio, determining target construction, there are cracks, is eliminated
The influence of noise, improves the accuracy of Crack Detection.
As shown in fig. 7, in one embodiment, building cracks detection method further includes the step of training machine learning model
Suddenly, which specifically comprises the following steps:
Step 702, the Architectural drawing sample and corresponding label about building are obtained.
Wherein, Architectural drawing sample can be the image of the building for training machine learning model.Label can be
To the mark of the FRACTURE CHARACTERISTICS addition in Architectural drawing sample.
Specifically, terminal acquisition may include splitting in Architectural drawing sample largely about the Architectural drawing sample of building
Feature is stitched, can not also include FRACTURE CHARACTERISTICS.Testing staff's fracture feature is labeled to obtain corresponding with Architectural drawing sample
Label.Architectural drawing sample and corresponding label are sent to server, the architectural drawing that server receiving terminal is sent by terminal
Decent and corresponding label.
Step 704, Architectural drawing sample is handled according to different image procossing modes respectively, obtains multiple groups building
Image pattern.
Specifically, server needs to carry out the Architectural drawing sample got data extending decent with abundant architectural drawing
This.Server handles Architectural drawing sample according to different image procossing modes, has handled when carrying out data extending
Multiple groups Architectural drawing sample is obtained after.
Step 706, obtained multiple groups Architectural drawing sample input machine learning model is trained respectively, is instructed
Practice FRACTURE CHARACTERISTICS image.
Wherein, training FRACTURE CHARACTERISTICS image can be the image that machine learning model exports in training, be labeled with machine
The FRACTURE CHARACTERISTICS that learning model detects.
Specifically, the multiple groups Architectural drawing sample obtained after data extending is input in machine learning model by server,
Machine learning model handles Architectural drawing sample, detects the FRACTURE CHARACTERISTICS in Architectural drawing sample and is labeled, obtains
To training FRACTURE CHARACTERISTICS image, and export obtained training FRACTURE CHARACTERISTICS image.
In one embodiment, it before being trained machine learning model, needs first to construct machine learning model.Structure
The machine learning model built can use convolutional neural networks, convolutional neural networks can by input layer, convolutional layer, pond layer and
Output layer composition, wherein convolutional layer and pond layer can be overlapping in this way with convolutional layer, pond layer, convolutional layer, pond layer
Form constructed.
For first layer convolutional layer, input as Architectural drawing sample, the expression formula of first layer convolutional layer are as follows:
Wherein Architectural drawing sample can use RGB (Red Green Blue, RGB) color mode, I1-I3It is respectively
The pixel value in tri- channels R, G and B of the Architectural drawing sample of input.
When being pond layer for upper one layer of convolutional layer, the calculation formula of l layers of convolutional layer are as follows:
Wherein,For l layers of j-th of Feature Mapping, ulFor after convolution bias operation but also not by activation primitive f1
When variable, MjFor the Feature Mapping set of input,For l-1 layers of ith features, (l-1 layers of output is
L layers of input),The convolution kernel of feature is exported to correspond to i-th of input feature vector in l layers with j-th,For corresponding to
The bias term of j-th of output feature.f1For activation primitive, can be ReLU activation primitive (Rectified Linear Unit,
Line rectification function), expression formula are as follows:
Unilateral inhibition is carried out using ReLU, and training speed is faster.
Traditional convolutional neural networks when carrying out convolution algorithm, convolutional layer will in the edge filling 0 of input picture so that
Input picture and output image keep identical size.And the convolutional layer established in this programme is utilized without filling
All information derive from Architectural drawing sample itself.Because using small-sized image when this programme training, input is big when identification
Sized image, if carrying out the operation of edge filling 0 when training, edge filling will not be 0 when detection, but large-size images
In some small block edge pixel, cause training with detection used in information it is inconsistent, influence testing result.In addition, if
Using the convolution kernel having a size of 3 × 3, then the size of the characteristic image exported will reduce 2 than input size.
The characteristic image progress that pond layer is used to export convolutional layer is down-sampled, its calculation formula is:
Wherein,For l layers of j-th of characteristic image,Indicate the characteristic pattern exported to l-1 layers
It is down-sampled as carrying out, it is down-sampled to use the method being averaging, then in the characteristic image that pond layer is exported according to l-1 layers
The average value of each 2 × 2 fritter obtains a point in the characteristic image of l layers of output, i.e., the spy exported l-1 layers
2 × 2 times of size reduction for levying image.WithRespectively multiplication biasing and addition biasing.
Output layer in this programme is that the full articulamentum of traditional convolutional neural networks is replaced with convolutional layer, so that convolution
Neural network can handle the image of arbitrary dimension.
The loss function of convolutional neural networks can be cross entropy loss function, and cross entropy loss function can be two classification
Cross entropy loss function can detecte in the image of input with corresponding to convolutional neural networks comprising FRACTURE CHARACTERISTICS or not comprising splitting
Stitch feature.The calculation formula of two classification cross entropy loss functions is as follows:
Wherein y is the training FRACTURE CHARACTERISTICS image of output, and y^ is label corresponding with Architectural drawing sample, and Loss is to calculate
Obtained error.
Step 708, the error between training FRACTURE CHARACTERISTICS image and label is calculated.
Specifically, after server obtains the training FRACTURE CHARACTERISTICS image of machine learning model output, by training FRACTURE CHARACTERISTICS
Image and label corresponding with Architectural drawing sample substitute into loss function and are calculated, and obtain training FRACTURE CHARACTERISTICS image and mark
Error between label.Server can calculate error according to formula (5).
Step 710, according to the model parameter in error transfer factor machine learning model.
Wherein, model parameter can be the parameter in machine learning model.
Specifically, after server obtains error, according to the model in the trend adjustment machine learning model for minimizing error
Parameter, and continue training parameter machine learning model adjusted, until error is less than preset error threshold, deconditioning,
Obtain the machine learning model that training finishes.
In one embodiment, when server adjustment model parameter, the frequency of training counter in machine learning model is certainly
Increase by 1, when the frequency of training of frequency of training counters count is greater than preset frequency of training threshold value, then deconditioning, obtains
The machine learning model that training finishes.
In one embodiment, server adjusts the model parameter in convolutional layer and pond layer by gradient descent method.Clothes
Device be engaged in when carrying out parameter adjustment to convolutional layer, calculates separately error Loss to convolution kernelAnd biasingPartial derivative, will count
The product of obtained partial derivative and learning rate η (can be constant), as convolution kernelAnd biasingIn each adjustment
Variable quantity, calculation formula are as follows:
Wherein,It is the convolution kernel before parameter adjustment,It is parameter convolution kernel adjusted；
It is the biasing before parameter adjustment,It is parameter biasing adjusted.
Model parameter in the layer of pond is multiplication biasingIt is biased with additionCalculation formula when adjustment is as follows:
Wherein,It is the multiplication biasing before parameter adjustment,It is parameter multiplication biasing adjusted；It is the addition biasing before parameter adjustment,It is parameter addition biasing adjusted.Server can be first according to public affairs
Formula (5) calculates error, and the model parameter calculated in convolutional layer and pond layer is adjusted separately further according to formula (6) and formula (7).
In one embodiment, server is randomly provided the convolution kernel in convolutional layerAnd biasingMultiplying in the layer of pond
Method biasingIt is biased with additionBy learning rate η, error threshold e and frequency of training threshold value n is manually arranged, wherein learning rate η,
Error threshold e and frequency of training threshold value n is constant.The mistake between trained FRACTURE CHARACTERISTICS image and label is calculated in server
After difference, error is carried out by backpropagation by chain rule, model parameter is adjusted until training terminates.
In the present embodiment, Architectural drawing sample is handled according to different image procossing modes respectively, so as to machine
The Architectural drawing sample that device learning model is trained is more abundant.The multiple groups Architectural drawing sample that will be obtained respectively in training
Machine learning model is inputted, obtains training FRACTURE CHARACTERISTICS image, calculates the error between training FRACTURE CHARACTERISTICS image and label, and
According to the model parameter in error transfer factor machine learning model, it ensure that the machine learning model after the completion of training can be accurately
Carry out Crack Detection.
As shown in figure 8, in one embodiment, step 704 specifically further includes the steps that constructing multiple groups Architectural drawing sample,
The step specifically comprises the following steps:
It step 802, is multiple Architectural drawing blocks by Architectural drawing sample decomposition.
Wherein, Architectural drawing block can be Architectural drawing sample is split after obtained image.
Specifically, Architectural drawing sample decomposition can be the image of default size by server, obtain multiple Architectural drawings
Block.For example, Architectural drawing sample decomposition can be the Architectural drawing block that multiple spatial resolutions are 38*38 by server.
Step 804, image transformation is carried out to resulting Architectural drawing block respectively, obtains the first image block；Image transformation packet
Include at least one of rotation processing and scaling processing.
Specifically, after server obtains Architectural drawing block, image transformation can be carried out to Architectural drawing block, image transformation can
To include at least one of rotation processing and scaling processing, after the completion of image transformation, server obtains the first image block.
When server carries out rotation processing to Architectural drawing block, horizon glass image rotation turn, vertical mirror rotation can be carried out, also
Multiple rotary can be carried out by origin of the center of Architectural drawing block.Architectural drawing after the determining rotation of calculating that server passes through matrix
As the distribution of pixel in block, for example, when Architectural drawing block is rotated by 90 ° around center, calculation formula are as follows:
Wherein, x0And y0It is the abscissa and ordinate of pixel in Architectural drawing block before rotating respectively；X and y is rotation respectively
The abscissa and ordinate of the pixel afterwards.
Step 806, image enhancement processing is carried out to the first authentic copy of the first image block, obtains the second image block.
Wherein, image enhancement processing is that some information or transformation data are added to image, to improve the visual effect of image.
Specifically, after server obtains the first image block, multiple copies is carried out to the first image block and obtain multiple copies.When
When server replicates the first image block twice, two copies, the respectively first authentic copy and triplicate are obtained.
Server can carry out different image enhancement processings to the first authentic copy, obtain the second image block.At image enhancement
Reason, which can be, is adjusted the brightness of the first authentic copy, to obtain the first authentic copy of multiple and different brightness.Server is bright in progress
When degree adjustment, the histogram of the first authentic copy can be compressed or be stretched along horizontal axis, to simulate different illumination.
Step 808, image filtering processing is carried out to the triplicate of the first image block, obtains third image block.
Specifically, server can carry out image filtering processing to triplicate, obtain third image block.Carrying out image
When filtering processing, triplicate is handled by a variety of filtering methods, random noise can also be added in triplicate.
Step 810, the first image block, the second image block and third image block are configured to multiple groups Architectural drawing sample.
Specifically, after server is handled Architectural drawing block according to different image procossing modes, will obtain
One image block, the second image block and third image block obtain multiple groups Architectural drawing sample as Architectural drawing sample.
Fig. 9 is the schematic diagram of the multiple groups Architectural drawing sample of server construction in one embodiment, specifically, reference Fig. 9,
The first row, which can be, carries out the first image block that rotation processing obtains to Architectural drawing block；Second row can be to the first image block
The first authentic copy carry out the second image block that enhancing is handled；The third line can be the triplicate progress to the first image block
Obtained third image block is filtered.In Fig. 9, according to 1 Architectural drawing block that segmentation obtains, 12 Architectural drawings are obtained
Sample.
In the present embodiment, Architectural drawing sample is split, then carries out different processing, with simulate it is a variety of under the conditions of
Architectural drawing sample ensure that the available sufficient training of machine learning model.
It should be understood that although each step in the flow chart of Fig. 2-8 is successively shown according to the instruction of arrow,
These steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly stating otherwise herein, these steps
Execution there is no stringent sequences to limit, these steps can execute in other order.Moreover, at least one in Fig. 2-8
Part steps may include that perhaps these sub-steps of multiple stages or stage are not necessarily in synchronization to multiple sub-steps
Completion is executed, but can be executed at different times, the execution sequence in these sub-steps or stage is also not necessarily successively
It carries out, but can be at least part of the sub-step or stage of other steps or other steps in turn or alternately
It executes.
Figure 10 is the schematic diagram of building cracks detection in one embodiment, specifically, referring to Fig.1 0, building cracks detection can
To be divided into training stage and detection-phase.Wherein, the training stage obtains Architectural drawing sample first, then to Architectural drawing sample
It is split, data extending then is carried out to Architectural drawing block and obtains multiple groups Architectural drawing sample, then constructs machine learning model,
According to multiple groups Architectural drawing sample training machine learning model.In detection-phase, the Architectural drawing of target construction is obtained first,
Architectural drawing is inputted into trained machine learning model and obtains FRACTURE CHARACTERISTICS image, then is secondary really by the progress of crack connected domain
It is fixed, that is, determine that target construction whether there is crack.
In one embodiment, as shown in figure 11, a kind of building cracks detection device 1100 is provided, comprising: image obtains
Modulus block 1102, image detection module 1104, image processing module 1106, connected domain extraction module 1108 and crack determining module
1110, in which:
Image collection module 1102, for obtaining the Architectural drawing of target construction.
Image detection module 1104 obtains FRACTURE CHARACTERISTICS for detecting Architectural drawing input machine learning model
Image.
Image processing module 1106 carries out binary conversion treatment for fracture characteristic image, obtains binaryzation FRACTURE CHARACTERISTICS
Image.
Connected domain extraction module 1108, for extracting crack connected domain from binaryzation FRACTURE CHARACTERISTICS image.
Crack determining module 1110, for determining target construction with the presence or absence of crack by crack connected domain.
In the present embodiment, after the Architectural drawing for obtaining target construction, Architectural drawing is inputted into machine learning model, is passed through
Machine learning model carries out first time detection to the FRACTURE CHARACTERISTICS in Architectural drawing, obtains FRACTURE CHARACTERISTICS image；Fracture is special again
It levies image and carries out binary conversion treatment, obtain binaryzation FRACTURE CHARACTERISTICS image, crack is extracted from binaryzation FRACTURE CHARACTERISTICS image and is connected
Logical domain；Second is carried out by crack connected domain to detect, and determines that target construction with the presence or absence of crack, combines in the detection process
Machine learning model and connected domain are detected twice to reduce error detection, and accuracy in detection is improved.
In one embodiment, connected domain extraction module 1108 is also used to carry out morphology to binaryzation FRACTURE CHARACTERISTICS image
Processing；Each pixel in binaryzation FRACTURE CHARACTERISTICS image after traversing through Morphological scale-space；When the pixel traversed meet it is pre-
If when fracture conditions, determining crack connected domain according to the pixel for meeting default fracture conditions.
In the present embodiment, Morphological scale-space first is carried out with the influence of less noise to binaryzation FRACTURE CHARACTERISTICS image, is improved
The accuracy of image procossing.Each pixel in binaryzation FRACTURE CHARACTERISTICS image after traversing Morphological scale-space, it is pre- according to meeting
If the pixel of fracture conditions determines crack connected domain, avoids omission and extract crack connected domain.
In one embodiment, connected domain extraction module 1108 is also used to meet the pixel of default fracture conditions as kind
Sub-pixel, and the coordinate of sub-pixel is subjected to stacking；The pixel for meeting default fracture conditions in the neighborhood of sub-pixel is made
For pixel can be filled, and the coordinate of fillable pixel is subjected to stacking；Pop-up is located at the coordinate of stack top；The coordinate institute of pop-up is right
The pixel answered continues to search fillable pixel as sub-pixel, until being sky in stack；Picture corresponding to each coordinate by pop-up
Plain region is determined as crack connected domain.
In the present embodiment, the coordinate that will meet the pixel of default fracture conditions carries out stacking, pops up stack top coordinate simultaneously one by one
The pixel for meeting default fracture conditions is continued to search, by the linear list that this operation of stack is limited, can not be found with omitting
All meet the pixel of default fracture conditions, and crack connected domain is determined according to the pixel found, improves and extract crack company
The accuracy in logical domain.
In one embodiment, crack determining module 1110 is also used to count the quantitative value of pixel in the connected domain of crack；When
When quantitative value is greater than present count magnitude, the boundary rectangle of crack connected domain is determined；Calculate the rectangular area and length and width of boundary rectangle
Than；When length-width ratio is greater than or equal to the first default ratio, determining target construction, there are cracks.
In the present embodiment, first judged according to the quantitative value of pixel in the connected domain of crack to remove the shadow of partial noise
It rings, it is ensured that the accuracy of detection.The shape of boundary rectangle is related to crack connected domain, by comparing boundary rectangle length-width ratio and
The size of first default ratio is realized from determining that target construction whether there is crack in shape, improves the standard of Crack Detection
True property.
In one embodiment, crack determining module 1110 is also used to when length-width ratio is less than the first default ratio, is calculated
Ratio between rectangular area and quantitative value；When ratio is greater than the second default ratio, determining target construction, there are cracks.
In the present embodiment, the crack connected domain that length-width ratio is less than default ratio may be from noise, according to rectangular area with
The ratio of quantitative value is judged again, and when ratio is greater than the second default ratio, determining target construction, there are cracks, is eliminated
The influence of noise, improves the accuracy of Crack Detection.
In one embodiment, building cracks detection device 1100 further include: obtain module, sample process module, model
Training module, error calculating module and parameter adjustment module, in which:
Module is obtained, for obtaining Architectural drawing sample and corresponding label about building.
Sample process module is obtained for being handled according to different image procossing modes respectively Architectural drawing sample
To multiple groups Architectural drawing sample.
Model training module, for respectively instructing obtained multiple groups Architectural drawing sample input machine learning model
Practice, obtains training FRACTURE CHARACTERISTICS image.
Error calculating module, for calculating the error between trained FRACTURE CHARACTERISTICS image and label.
Parameter adjustment module, for according to the model parameter in error transfer factor machine learning model.
In the present embodiment, Architectural drawing sample is handled according to different image procossing modes respectively, so as to machine
The Architectural drawing sample that device learning model is trained is more abundant.The multiple groups Architectural drawing sample that will be obtained respectively in training
Machine learning model is inputted, obtains training FRACTURE CHARACTERISTICS image, calculates the error between training FRACTURE CHARACTERISTICS image and label, and
According to the model parameter in error transfer factor machine learning model, it ensure that the machine learning model after the completion of training can be accurately
Carry out Crack Detection.
In one embodiment, sample process module is also used to Architectural drawing sample decomposition be multiple Architectural drawing blocks；
Image transformation is carried out to resulting Architectural drawing block respectively, obtains the first image block；Image transformation includes rotation processing and scaling
At least one of processing；Image enhancement processing is carried out to the first authentic copy of the first image block, obtains the second image block；To first
The triplicate of image block carries out image filtering processing, obtains third image block；By the first image block, the second image block and third
Image block is configured to multiple groups Architectural drawing sample.
In the present embodiment, Architectural drawing sample is split, then carries out different processing, with simulate it is a variety of under the conditions of
Architectural drawing sample ensure that the available sufficient training of machine learning model.
Specific about building cracks detection device limits the limit that may refer to above for building cracks detection method
Fixed, details are not described herein.Modules in above-mentioned building cracks detection device can fully or partially through software, hardware and its
Combination is to realize.Above-mentioned each module can be embedded in the form of hardware or independently of in the processor in computer equipment, can also be with
It is stored in the memory in computer equipment in a software form, in order to which processor calls the above modules of execution corresponding
Operation.
In one embodiment, a kind of computer equipment is provided, which can be server, internal junction
Composition is shown in Fig.12.The computer equipment include by system bus connect processor, memory, network interface and
Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment
Include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data
Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The database of machine equipment is used to store the data for building cracks detection.The network interface of the computer equipment is used for and outside
Terminal by network connection communication.To realize a kind of building cracks detection method when the computer program is executed by processor.
It will be understood by those skilled in the art that structure shown in Figure 12, only part relevant to application scheme
The block diagram of structure, does not constitute the restriction for the computer equipment being applied thereon to application scheme, and specific computer is set
Standby may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, memory is stored with meter
Calculation machine program, when computer program is executed by processor, so that the step of processor executes above-mentioned building cracks detection method.This
The step of locating building cracks detection method can be the step in the building cracks detection method of above-mentioned each embodiment.
In one embodiment, a kind of computer readable storage medium is provided, computer program, computer journey are stored with
When sequence is executed by processor, so that the step of processor executes above-mentioned building cracks detection method.Building cracks detection side herein
The step of method, can be the step in the building cracks detection method of above-mentioned each embodiment.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
1. a kind of building cracks detection method, which comprises
Obtain the Architectural drawing of target construction；
Architectural drawing input machine learning model is detected, FRACTURE CHARACTERISTICS image is obtained；
Binary conversion treatment is carried out to the FRACTURE CHARACTERISTICS image, obtains binaryzation FRACTURE CHARACTERISTICS image；
Crack connected domain is extracted from the binaryzation FRACTURE CHARACTERISTICS image；
Determine the target construction with the presence or absence of crack by the crack connected domain.
2. the method according to claim 1, wherein described extract from the binaryzation FRACTURE CHARACTERISTICS image is split
Stitching connected domain includes:
Morphological scale-space is carried out to the binaryzation FRACTURE CHARACTERISTICS image；
Each pixel in binaryzation FRACTURE CHARACTERISTICS image after traversing through Morphological scale-space；
When the pixel traversed meets default fracture conditions, determine that crack connects according to the pixel for meeting the default fracture conditions
Logical domain.
3. according to the method described in claim 2, it is characterized in that, the basis pixel that meets the default fracture conditions is true
Determining crack connected domain includes:
The pixel for meeting the default fracture conditions is subjected to stacking as sub-pixel, and by the coordinate of the sub-pixel；
The pixels of the default fracture conditions will be met in the neighborhood of the sub-pixel as fillable pixel, and will described in can
The coordinate of filler pixels carries out stacking；
Pop-up is located at the coordinate of stack top；
Fillable pixel is continued to search using pixel corresponding to the coordinate of pop-up as sub-pixel, until being sky in stack；
Pixel region corresponding to each coordinate by pop-up is determined as crack connected domain.
4. the method according to claim 1, wherein described determine that the target is built by the crack connected domain
Object, which is built, with the presence or absence of crack includes:
Count the quantitative value of pixel in the crack connected domain；
When the quantitative value is greater than present count magnitude, the boundary rectangle of the crack connected domain is determined；
Calculate the rectangular area and length-width ratio of the boundary rectangle；
When the length-width ratio is greater than or equal to the first default ratio, determine that there are cracks for the target construction.
5. according to the method described in claim 4, it is characterized in that, the method also includes:
When the length-width ratio is less than the first default ratio, the ratio between the rectangular area and the quantitative value is calculated
Value；
When the ratio is greater than the second default ratio, determine that there are cracks for the target construction.
6. the method according to claim 1, wherein being gone back before the Architectural drawing for obtaining target construction
Include:
Obtain the Architectural drawing sample and corresponding label about building；
The Architectural drawing sample is handled according to different image procossing modes respectively, it is decent to obtain multiple groups architectural drawing
This；
Obtained multiple groups Architectural drawing sample input machine learning model is trained respectively, obtains training FRACTURE CHARACTERISTICS figure
Picture；
Calculate the error between the trained FRACTURE CHARACTERISTICS image and the label；
According to the model parameter in machine learning model described in the error transfer factor.
7. according to the method described in claim 6, it is characterized in that, it is described to the Architectural drawing sample respectively according to different
Image procossing mode is handled, and is obtained multiple groups Architectural drawing sample and is included:
It is multiple Architectural drawing blocks by the Architectural drawing sample decomposition；
Image transformation is carried out to resulting Architectural drawing block respectively, obtains the first image block；Described image transformation includes at rotation
At least one of reason and scaling processing；
Image enhancement processing is carried out to the first authentic copy of the first image block, obtains the second image block；
Image filtering processing is carried out to the triplicate of the first image block, obtains third image block；
The first image block, second image block and the third image block are configured to multiple groups Architectural drawing sample.
8. a kind of building cracks detection device, which is characterized in that described device includes:
Image collection module, for obtaining the Architectural drawing of target construction；
Image detection module obtains FRACTURE CHARACTERISTICS image for detecting Architectural drawing input machine learning model；
Image processing module obtains binaryzation FRACTURE CHARACTERISTICS image for carrying out binary conversion treatment to the FRACTURE CHARACTERISTICS image；
Connected domain extraction module, for extracting crack connected domain from the binaryzation FRACTURE CHARACTERISTICS image；
Crack determining module, for determining the target construction with the presence or absence of crack by the crack connected domain.
9. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, which is characterized in that the processor realizes any one of claims 1 to 7 institute when executing the computer program
The step of stating method.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claims 1 to 7 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910298780.5A CN110163842B (en) | 2019-04-15 | 2019-04-15 | Building crack detection method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910298780.5A CN110163842B (en) | 2019-04-15 | 2019-04-15 | Building crack detection method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110163842A true CN110163842A (en) | 2019-08-23 |
| CN110163842B CN110163842B (en) | 2021-06-25 |
Family
ID=67639583
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910298780.5A Active CN110163842B (en) | 2019-04-15 | 2019-04-15 | Building crack detection method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110163842B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110929565A (en) * | 2019-10-15 | 2020-03-27 | 平安科技（深圳）有限公司 | Risk monitoring method and device based on machine learning, storage medium and electronic equipment |
| CN112001359A (en) * | 2020-09-09 | 2020-11-27 | 郑州航空工业管理学院 | Heuristic multi-branch line repairing method based on pattern recognition and optimal distribution |
| CN112016848A (en) * | 2020-09-11 | 2020-12-01 | 范玲珍 | Intelligent detection management system for quality supervision, acceptance and acceptance of constructional engineering based on data scheduling |
| CN112837290A (en) * | 2021-02-03 | 2021-05-25 | 中南大学 | Crack image automatic identification method based on seed filling algorithm |
| CN113160168A (en) * | 2021-04-16 | 2021-07-23 | 浙江高速信息工程技术有限公司 | Noise processing method and system based on fracture image binarization connected region |
| CN113643233A (en) * | 2021-07-01 | 2021-11-12 | 深圳市格灵精睿视觉有限公司 | Oily coating detection method, system and equipment and computer readable storage medium |
| CN113658111A (en) * | 2021-07-27 | 2021-11-16 | 武汉理工大学 | Crack detection method, system, device and medium for building or structure |
| CN114782845A (en) * | 2022-06-20 | 2022-07-22 | 青岛理工大学 | Method, device, equipment and medium for identifying new cracks of building after blasting |
| CN115393270A (en) * | 2022-07-14 | 2022-11-25 | 北京建筑大学 | Automatic identification and prediction method, device and equipment for architecture heritage diseases |
| CN116721354A (en) * | 2023-08-08 | 2023-09-08 | 中铁七局集团电务工程有限公司武汉分公司 | Building crack defect identification method, system and readable storage medium |
| WO2024087870A1 (en) * | 2022-10-26 | 2024-05-02 | 上海船舶工艺研究所(中国船舶集团有限公司第十一研究所) | Defect identification method for x-ray weld seam image, device, and storage medium |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103528515A (en) * | 2013-10-15 | 2014-01-22 | 北京交通大学长三角研究院 | Dynamic detection method for crack of bridge bottom surface |
| CN104021574A (en) * | 2014-07-04 | 2014-09-03 | 武汉武大卓越科技有限责任公司 | Method for automatically identifying pavement diseases |
| CN104792792A (en) * | 2015-04-27 | 2015-07-22 | 武汉武大卓越科技有限责任公司 | Stepwise-refinement pavement crack detection method |
| GB2542118A (en) * | 2015-09-04 | 2017-03-15 | Toshiba Res Europe Ltd | A method, apparatus, system, and computer readable medium for detecting change to a structure |
| CN106651872A (en) * | 2016-11-23 | 2017-05-10 | 北京理工大学 | Prewitt operator-based pavement crack recognition method and system |
| KR101772916B1 (en) * | 2016-12-30 | 2017-08-31 | 한양대학교 에리카산학협력단 | Device for measuring crack width of concretestructure |
| CN107230202A (en) * | 2017-05-16 | 2017-10-03 | 淮阴工学院 | The automatic identifying method and system of pavement disease image |
| CN107527354A (en) * | 2017-07-06 | 2017-12-29 | 长安大学 | A kind of region growing method based on composite diagram |
| CN107610092A (en) * | 2017-08-01 | 2018-01-19 | 长安大学 | Pavement crack dynamic testing method based on video flowing |
| CN108229461A (en) * | 2018-01-16 | 2018-06-29 | 上海同岩土木工程科技股份有限公司 | A kind of tunnel slot method for quickly identifying based on deep learning |
| CN108364280A (en) * | 2018-01-03 | 2018-08-03 | 东南大学 | Structural cracks automation describes and width accurately measures method and apparatus |
| WO2018165753A1 (en) * | 2017-03-14 | 2018-09-20 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN108765386A (en) * | 2018-05-16 | 2018-11-06 | 中铁科学技术开发公司 | A kind of tunnel slot detection method, device, electronic equipment and storage medium |
| CN109325468A (en) * | 2018-10-18 | 2019-02-12 | 广州智颜科技有限公司 | A kind of image processing method, device, computer equipment and storage medium |
-
2019
- 2019-04-15 CN CN201910298780.5A patent/CN110163842B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103528515A (en) * | 2013-10-15 | 2014-01-22 | 北京交通大学长三角研究院 | Dynamic detection method for crack of bridge bottom surface |
| CN104021574A (en) * | 2014-07-04 | 2014-09-03 | 武汉武大卓越科技有限责任公司 | Method for automatically identifying pavement diseases |
| CN104792792A (en) * | 2015-04-27 | 2015-07-22 | 武汉武大卓越科技有限责任公司 | Stepwise-refinement pavement crack detection method |
| GB2542118A (en) * | 2015-09-04 | 2017-03-15 | Toshiba Res Europe Ltd | A method, apparatus, system, and computer readable medium for detecting change to a structure |
| CN106651872A (en) * | 2016-11-23 | 2017-05-10 | 北京理工大学 | Prewitt operator-based pavement crack recognition method and system |
| KR101772916B1 (en) * | 2016-12-30 | 2017-08-31 | 한양대학교 에리카산학협력단 | Device for measuring crack width of concretestructure |
| WO2018165753A1 (en) * | 2017-03-14 | 2018-09-20 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN107230202A (en) * | 2017-05-16 | 2017-10-03 | 淮阴工学院 | The automatic identifying method and system of pavement disease image |
| CN107527354A (en) * | 2017-07-06 | 2017-12-29 | 长安大学 | A kind of region growing method based on composite diagram |
| CN107610092A (en) * | 2017-08-01 | 2018-01-19 | 长安大学 | Pavement crack dynamic testing method based on video flowing |
| CN108364280A (en) * | 2018-01-03 | 2018-08-03 | 东南大学 | Structural cracks automation describes and width accurately measures method and apparatus |
| CN108229461A (en) * | 2018-01-16 | 2018-06-29 | 上海同岩土木工程科技股份有限公司 | A kind of tunnel slot method for quickly identifying based on deep learning |
| CN108765386A (en) * | 2018-05-16 | 2018-11-06 | 中铁科学技术开发公司 | A kind of tunnel slot detection method, device, electronic equipment and storage medium |
| CN109325468A (en) * | 2018-10-18 | 2019-02-12 | 广州智颜科技有限公司 | A kind of image processing method, device, computer equipment and storage medium |
Non-Patent Citations (5)
| Title |
|---|
| HENRIQUE OLIVEIRA ET AL: "Road surface crack detection: Improved segmentation with pixel-based refinement", 《2017 25TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO)》 * |
| 狄亚平: "路面裂缝识别算法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 王琳等: "基于双目立体视觉技术的桥梁裂缝测量系统", 《计算机应用》 * |
| 马建等: "路面检测技术综述", 《交通运输工程学报》 * |
| 高尚兵等: "一种新的路面裂缝自动检测算法", 《系统仿真学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110929565B (en) * | 2019-10-15 | 2023-12-22 | 平安科技（深圳）有限公司 | Risk monitoring method and device based on machine learning, storage medium and electronic equipment |
| CN110929565A (en) * | 2019-10-15 | 2020-03-27 | 平安科技（深圳）有限公司 | Risk monitoring method and device based on machine learning, storage medium and electronic equipment |
| CN112001359B (en) * | 2020-09-09 | 2023-07-18 | 郑州航空工业管理学院 | Heuristic multi-fork line repairing method based on pattern recognition and optimal allocation |
| CN112001359A (en) * | 2020-09-09 | 2020-11-27 | 郑州航空工业管理学院 | Heuristic multi-branch line repairing method based on pattern recognition and optimal distribution |
| CN112016848A (en) * | 2020-09-11 | 2020-12-01 | 范玲珍 | Intelligent detection management system for quality supervision, acceptance and acceptance of constructional engineering based on data scheduling |
| CN112837290A (en) * | 2021-02-03 | 2021-05-25 | 中南大学 | Crack image automatic identification method based on seed filling algorithm |
| CN113160168A (en) * | 2021-04-16 | 2021-07-23 | 浙江高速信息工程技术有限公司 | Noise processing method and system based on fracture image binarization connected region |
| CN113643233A (en) * | 2021-07-01 | 2021-11-12 | 深圳市格灵精睿视觉有限公司 | Oily coating detection method, system and equipment and computer readable storage medium |
| CN113643233B (en) * | 2021-07-01 | 2024-05-24 | 深圳市格灵精睿视觉有限公司 | Oily coating detection method, system and equipment, and computer readable storage medium |
| CN113658111B (en) * | 2021-07-27 | 2023-12-12 | 武汉理工大学 | Crack detection method, system, device and medium for building or structure |
| CN113658111A (en) * | 2021-07-27 | 2021-11-16 | 武汉理工大学 | Crack detection method, system, device and medium for building or structure |
| CN114782845A (en) * | 2022-06-20 | 2022-07-22 | 青岛理工大学 | Method, device, equipment and medium for identifying new cracks of building after blasting |
| CN115393270A (en) * | 2022-07-14 | 2022-11-25 | 北京建筑大学 | Automatic identification and prediction method, device and equipment for architecture heritage diseases |
| WO2024087870A1 (en) * | 2022-10-26 | 2024-05-02 | 上海船舶工艺研究所(中国船舶集团有限公司第十一研究所) | Defect identification method for x-ray weld seam image, device, and storage medium |
| CN116721354A (en) * | 2023-08-08 | 2023-09-08 | 中铁七局集团电务工程有限公司武汉分公司 | Building crack defect identification method, system and readable storage medium |
| CN116721354B (en) * | 2023-08-08 | 2023-11-21 | 中铁七局集团电务工程有限公司武汉分公司 | Building crack defect identification method, system and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110163842B (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163842A (en) | Building cracks detection method, device, computer equipment and storage medium | |
| US11681418B2 (en) | Multi-sample whole slide image processing in digital pathology via multi-resolution registration and machine learning | |
| CN110660023B (en) | Video stitching method based on image semantic segmentation | |
| WO2021000702A1 (en) | Image detection method, device, and system | |
| CA2867365C (en) | Method, system and computer storage medium for face detection | |
| US11544916B2 (en) | Automated gauge reading and related systems, methods, and devices | |
| CN108764208A (en) | Image processing method and device, storage medium, electronic equipment | |
| Oliveira et al. | Unsupervised local color correction for coarsely registered images | |
| Christlein et al. | A study on features for the detection of copy-move forgeries | |
| JP2010045613A (en) | Image identifying method and imaging device | |
| CN109740572A (en) | A kind of human face in-vivo detection method based on partial color textural characteristics | |
| JP6047561B2 (en) | Method, computer system, and computer program for processing a plurality of images of the same scene | |
| JP2017076909A (en) | Image processing device, image processing method and program | |
| CN110349163A (en) | Image processing method and device, electronic equipment, computer readable storage medium | |
| RU2320011C1 (en) | Method for automatic correction of red-eye effect | |
| US8498453B1 (en) | Evaluating digital images using head points | |
| CN111428730B (en) | Weak supervision fine-grained object classification method | |
| CN108254380A (en) | PCB circuit board template matching method based on Digital Image Processing | |
| CN106652898A (en) | LED display point-to-point correction method | |
| CN112037128A (en) | Panoramic video splicing method | |
| WO2023225774A1 (en) | Image processing method and apparatus, and electronic device and computer-readable storage medium | |
| CN116229084A (en) | Empty target detection method | |
| CN113344011B (en) | Color constancy method based on cascade fusion feature confidence weighting | |
| CN115035281A (en) | Rapid infrared panoramic image splicing method | |
| CN115035466A (en) | Infrared panoramic radar system for safety monitoring |
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 |