CN110020652A - The dividing method of Tunnel Lining Cracks image - Google Patents
The dividing method of Tunnel Lining Cracks image Download PDFInfo
- Publication number
- CN110020652A CN110020652A CN201910012867.1A CN201910012867A CN110020652A CN 110020652 A CN110020652 A CN 110020652A CN 201910012867 A CN201910012867 A CN 201910012867A CN 110020652 A CN110020652 A CN 110020652A
- Authority
- CN
- China
- Prior art keywords
- image
- crack
- training sample
- sample set
- tunnel lining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 80
- 230000011218 segmentation Effects 0.000 claims abstract description 33
- 238000010606 normalization Methods 0.000 claims abstract description 19
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 18
- 230000006870 function Effects 0.000 claims description 37
- 238000012545 processing Methods 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 description 12
- 238000013528 artificial neural network Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008439 repair process Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of dividing method of Tunnel Lining Cracks image, belongs to image identification technical field.This method is first: direct collected original fracture image being divided into multiple images block, selects there are the image blocks in crack and corresponding bianry image as training sample set；Mean normalization is carried out there are the image block in crack to training sample concentration to handle；Using convolutional neural networks, in conjunction with loss function, training sample set is trained to removing mean normalization treated, obtains crack parted pattern；It after going mean normalization to handle crack image to be split, is input in the parted pattern of crack, exports segmentation result image.The present invention is based on the methods of depth convolutional neural networks to carry out automatic identification to image, reduces human intervention, has evaded the influence of tester's subjective factor, improved accuracy of identification and recognition efficiency, reduced rate of false alarm.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a segmentation method of a tunnel lining crack image.
Background
The construction process of infrastructures such as roads, railway tunnels and the like is faster and faster, so that the requirement on safety is higher, and the requirement on perfecting protective measures is increased day by day. Due to factors such as weather and construction process, cracks and other diseases can appear on the inner wall of the tunnel in different degrees. When the crack is seriously damaged, the collapse accident of the tunnel can be caused. Therefore, the regular inspection of the diseases in the tunnel environment has a crucial influence on the road traffic safety operation.
In the past, the tunnel repair and protection work is mainly carried out manually by using professional instruments and equipment, so that the accuracy is high. But the time and the labor are wasted, the working efficiency is low, and the defects are difficult to be checked and maintained in time. In recent years, it has been desired to take pictures of tunnels with the aid of industrial cameras, technicians not having to arrive at the site but automatically identifying cracks from the acquired images of the tunnel lining by means of digital image processing. The komass Engineering Crop acquires the tunnel lining image through the laser scanner, and the system is very effective in detecting disease areas with relatively large areas such as stripping and fragments, but cannot achieve the automatic effect on crack detection. Seung-NamYu et al, university of Hanyang, Korea, improved on the basis of this, and proposed an automated crack detection method, which has a relatively slow vehicle speed, although the detection accuracy is high, and needs further optimization.
The Lenan et al of Changan university classifies crack images based on the pavement crack recognition of a deep learning framework Caffe. Cha, y.j. et al of civil engineering department of manitoba university, canada, detects cracks with respect to concrete and steel bar surface images using a deep neural network, intercepts small-sized images from large-sized images using a sliding window for classification, and identifies image blocks where cracks are located. Therefore, the specific position of the crack can be more accurately positioned. However, such recognition accuracy cannot accurately obtain the attribute characteristics of the crack, such as the direction, length, width, etc. of the crack. On concrete surfaces such as highway tunnels, railway tunnels, bridges and the like, cracks in the tunnels are detected and subjected to attribute evaluation according to relevant standard specification requirements, and the method has very important significance for timely maintenance and repair of the tunnels.
Disclosure of Invention
The invention aims to provide a segmentation method and a segmentation system for a tunnel lining crack image, which improve the identification precision and the identification efficiency and solve the technical problems in the background technology.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a segmentation method of a tunnel lining crack image, which comprises the following flow steps:
step S110: dividing an original crack image directly acquired into a plurality of image blocks, and selecting the image blocks with cracks and corresponding binary images as training sample sets;
step S120: carrying out mean value removing normalization processing on the image blocks with cracks in the training sample set;
step S130: training the training sample set subjected to mean value removal normalization processing by using a convolutional neural network in combination with a loss function to obtain a crack segmentation model;
step S140: and after mean value removing and normalization processing are carried out on the crack images to be segmented, inputting the crack images into the crack segmentation model, and outputting segmentation result images.
Further, the step S110 specifically includes:
dividing each directly acquired original crack image into a plurality of image blocks, wherein the number of pixels of each image block is lower than that of the original crack image, and selecting the image blocks with cracks and corresponding binary images as training sample sets.
Further, the size of the original crack image is 1000 × 4000 pixels, the number of the image blocks is 100, and the size of the image block is 200 × 200 pixels.
Further, the step S120 specifically includes:
average gray value of image blocks with cracks in training sample setThe following constraints are satisfied:
wherein h represents the height of the image blocks in the training sample set, w represents the width of the image blocks in the training sample set, N is the total number of the image blocks in the training sample set,expressing the gray value of the pixel point of the ith row and the jth column in each n image blocks in the training sample set;
then the gray value X of the pixel point of the ith row and the jth column in the image block after the mean value normalization processing is removedi′,jComprises the following steps:
wherein, Xi,jAnd expressing the gray value of the pixel point of the ith row and the jth column of the image block in the training sample set.
Further, in the step S130, the convolutional neural network is a VGG-16 network.
Further, in step S130, the loss function is a class-balanced cross entropy loss function:
wherein the training sample set is S { (X)n，Yn)，n＝1，2，...，N},XnRepresenting the nth image block in the training sample set,whether the jth pixel in the binary image corresponding to the nth image block is a crack pixel or not is shown, 0 is the crack pixel, and 1 is a non-crack background pixel;
q represents the set of weight parameters of the convolutional neural network, the update is trained by back propagation, Y+Representing foreground crack pixels, Y-Representing a background non-slit pixel, β ═ Y-I/Y is used to balance the number of crack pixels with the number of non-crack pixels, and the probability P (-) indicates that the Sigmoid σ (-) function is used as the activation function at the last convolution layer.
Further, the step S130 specifically includes:
training by adopting a random gradient descent method, inputting a feature map output by an image block in a training sample set through a convolutional neural network and a binary image corresponding to the image block into a loss function together, and calculating a loss value; and (4) iteratively updating the weight information of the network, and minimizing the loss function until the network converges to obtain the fracture segmentation model.
Further, the training sample set is divided into a plurality of batches, each batch has 16 training samples, the momentum is 0.9, the weight attenuation is 0.0001, and the update rule of the weight q is as follows:
qm+1:＝qm+vm+1
wherein epsilon represents the learning rate, m represents the number of iterations,representing an objective functionDerivative Q with respect to QmAverage value of (a).
The invention has the beneficial effects that: the method based on the deep convolutional neural network automatically identifies the image, reduces human intervention, avoids the influence of subjective factors of a detector, improves identification precision and identification efficiency, and reduces false alarm rate.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a segmentation method for a tunnel lining crack image according to a first embodiment of the present invention.
Fig. 2 is a diagram of an image block with a crack according to a first embodiment of the present invention.
Fig. 3 is a binary image of the image block shown in fig. 2.
Fig. 4 is an image after the image block de-averaging normalization process shown in fig. 2.
Fig. 5 is a flowchart of a training test of a deep neural network model for segmenting a crack region in a tunnel lining image according to the second embodiment of the present invention.
Fig. 6 is a schematic diagram of a VGG-16 network structure according to a second embodiment of the present invention.
Fig. 7 is an image of a crack to be segmented according to the second embodiment of the present invention.
Fig. 8 is a segmentation result image according to the second embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or modules having the same or similar functionality throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
For the convenience of understanding of the embodiments of the present invention, the following description will be further explained by taking specific embodiments as examples with reference to the drawings, and the embodiments are not to be construed as limiting the embodiments of the present invention.
It will be understood by those of ordinary skill in the art that the figures are merely schematic representations of one embodiment and that the elements or devices in the figures are not necessarily required to practice the present invention.
Example one
As shown in fig. 1, a first embodiment of the present invention provides a method for segmenting a tunnel lining crack image, including the following steps:
step S110: dividing an original crack image directly acquired into a plurality of image blocks, and selecting the image blocks with cracks and corresponding binary images as training sample sets;
step S120: carrying out mean value removing normalization processing on the image blocks with cracks in the training sample set;
step S130: training the training sample set subjected to mean value removal normalization processing by using a convolutional neural network in combination with a loss function to obtain a crack segmentation model;
step S140: and after mean value removing and normalization processing are carried out on the crack images to be segmented, inputting the crack images into the crack segmentation model, and outputting segmentation result images.
In the first embodiment of the present invention, the step S110 specifically includes:
dividing each directly acquired original crack image into a plurality of image blocks, wherein the number of pixels of each image block is lower than that of the original crack image, and selecting the image block with cracks (shown in figure 2) and the corresponding binary image (shown in figure 3) as a training sample set.
Since the acquired crack image usually has a large amount of noise similar to the background gray scale, which results in low contrast between the crack and the background, it can be preprocessed by the mean value removing method. Therefore, the texture features of the background part can be removed, the difference between the crack and the background is highlighted, and the neural network is favorable for extracting the crack features.
In the first embodiment of the present invention, the step S120 specifically includes:
average gray value of image blocks with cracks in training sample setThe following constraints are satisfied:
wherein h represents the height of the image blocks in the training sample set, w represents the width of the image blocks in the training sample set, N is the total number of the image blocks in the training sample set,expressing the gray value of the pixel point of the ith row and the jth column in each n image blocks in the training sample set;
then the gray value X of the pixel point of the ith row and the jth column in the image block after the mean value normalization processing is removedi′,jComprises the following steps:
wherein, Xi,jAnd expressing the gray value of the pixel point of the ith row and the jth column of the image block in the training sample set. As shown in fig. 4, the image is the image after the mean value removing normalization process of the crack image block shown in fig. 2.
In the first embodiment of the invention, aiming at high-precision segmentation of a crack region, a VGG-16 network is selected as a basic network frame for deep learning, and the network can automatically extract features without artificial design features.
The neural network is trained in a mode of iteratively updating the network weight through a minimum loss function, so that the neural network can learn abundant crack characteristics, and a high-precision segmentation effect is finally obtained.
In a crack image of a tunnel lining, a crack area is elongated, and compared with a large number of background pixels, pixel points where cracks are located are sparse. The crack type pixel points and the background non-crack type pixel points have serious unbalance in number, and a loss function for solving the unbalance problem is necessary. Therefore, in the design of the cross entropy loss function, the problem of the class imbalance is solved by formalizing the problem of the imbalance by using coefficients.
In a first embodiment of the present invention, in the step S130, the loss function is a class-balanced cross entropy loss function:
wherein the training sample set is S { (X)n，Yn)，n＝1，2，...，N},XnRepresenting the nth image block in the training sample set,whether the jth pixel in the binary image corresponding to the nth image block is a crack pixel or not is shown, 0 is the crack pixel, and 1 is a non-crack background pixel;
q represents the set of weight parameters of the convolutional neural network, the update is trained by back propagation, Y+Representing foreground crack pixels, Y-Representing a background non-slit pixel, β ═ Y-I/Y is used to balance crack pixel points with non-crack pixel points, and the probability P (.) represents that a Sigmoid sigma (.) function is used as an activation function at the last convolution layer.
Specifically, the step S130 specifically includes:
training by adopting a random gradient descent method, inputting a feature map output by an image block in a training sample set through a convolutional neural network and a binary image corresponding to the image block into a loss function together, and calculating a loss value; and (4) iteratively updating the weight information of the network, and minimizing the loss function until the network converges to obtain the fracture segmentation model.
In the training process, the training sample set is divided into a plurality of batches, each batch has 16 training samples, the momentum is 0.9, the weight attenuation is 0.0001, and the updating rule of the weight q is as follows:
qm+1:＝qm+vm+1
wherein epsilon represents the learning rate, m represents the number of iterations,representing the derivative Q of the objective function with respect to QmAverage value of (a).
Example two
As shown in fig. 5, a training process and a testing process of a deep neural network model for segmenting a crack region in a tunnel lining image according to a second embodiment of the present invention are provided.
In the second embodiment of the present invention, a large amount of detailed information needs to be extracted for fine segmentation of a long and thin crack region, so that the neural network of the present invention integrates low-dimensional features (edges, gradients, etc.) and high-dimensional features (semantic features of cracks, etc.) extracted in a training process, and a high-precision segmentation effect is finally obtained through training.
For the directly acquired crack image, the size is 1000 × 4000 pixels. During the training of the neural network, the crack image size is large for the input of the network. Dividing each image into 100 sub-blocks of 200 × 200 pixels, selecting a large number of image blocks with cracks, and using the artificially labeled binary image corresponding to each crack image block as the input of the network. During the testing of the neural network, there is no requirement on the size of the test image.
The method of the second embodiment of the present invention specifically includes the following process steps:
s1 mean value removing normalization preprocessing of crack image
The method mainly comprises the steps of carrying out mean value removing processing on an original crack image, and subtracting the average gray value of all training sample sets from the gray value of each pixel point in the image block of the training data set. Let X denote an image block of the training data set,expressing the gray value of the pixel point of the ith row and the jth column in each n image blocks, wherein X' ij expresses the gray value of the pixel point of the ith row and the jth column in the image blocks after the mean value removing normalization processing, and the gray value is expressed as follows:
wherein,the average gray value of the image blocks with cracks in the training sample set,the following constraints are satisfied:
wherein h represents the height of the image blocks in the training sample set, w represents the width of the image blocks in the training sample set, N is the total number of the image blocks in the training sample set,and expressing the gray value of the pixel point of the ith row and the jth column in each n image blocks in the training sample set.
S2 design of fracture splitting network
In order to realize fracture segmentation in the tunnel image, a network model of fracture segmentation is designed, as shown in fig. 6. The common VGG-16 network is improved, and a full connection layer in the network is removed. In the VGG-16 network, 5 convolution stages are used for providing the characteristics of a crack image, the first two convolution stages respectively have 2 convolution layers, the last 3 convolution stages respectively have 3 convolution layers, the sizes of the convolution cores are 3x3, and a ReLU activation function is connected behind each convolution layer. Between each convolution stage is a Max Pooling layer with kernel size 2x2, step size 2. Because a great deal of detail information needs to be extracted for segmenting the crack region, the last layer of each convolution stage is output laterally, channels are spliced, and low-dimensional features and high-dimensional features are fused together. The number of output channels is changed by convolution with 1 × 1, so that the feature map of the multi-channel becomes a feature map of a single channel. Can play the role of dimension reduction without changing the size of the feature map. And finally, connecting a sigmoid cross entropy loss function.
S3: design of loss function
In order to solve the problem of unbalanced quantity of two types of pixels of a crack area and a non-crack area in a crack image, a cross entropy loss function of class balance is adopted. Training data set S { (X)n，Yn)，n＝1，2，...，N},XnRepresenting the nth image block in the training data set,and whether the jth pixel in the label image corresponding to the nth picture block is a crack pixel or not is shown, 0 is the crack pixel, and 1 is a non-crack background pixel.
The cross entropy loss function of class balancing is as follows:
wherein the training sample set is S { (X)n，Yn)，n＝1，2，...，N},XnRepresenting the nth image block in the training sample set,whether the jth pixel in the binary image corresponding to the nth image block is a crack pixel or not is shown, 0 is the crack pixel, and 1 is a non-crack background pixel;
q represents a set of weight parameters of the convolutional neural network, generalTraining updates, Y, over back propagation+Representing foreground crack pixels, Y-Representing a background non-slit pixel, β ═ Y-I/Y is used to balance the number of crack pixels with the number of non-crack pixels, and the probability P (-) indicates that the Sigmoid σ (-) function is used as the activation function at the last convolution layer.
Sigmoid function is as follows:
s4 training and testing of fracture splitting network
A random gradient descent method is adopted to train a fracture image segmentation network model, a small batch processing mode is adopted in the training process, all training data sets are divided into a plurality of batches, each batch is provided with 16 training samples, the momentum is 0.9, and the weight attenuation is 0.0001. The update rule of the weight q is:
qm+1:＝qm+vm+1
wherein epsilon represents the learning rate, m represents the number of iterations,representing the derivative Q of the objective function with respect to QmAverage value of (a).
The pre-training model trained on the ImageNet data set is used for initializing the weight of the network, so that the network training process can be accelerated. In the process of model training, a feature map output by an image block in a training sample set through a convolutional neural network and a binary image corresponding to the image block are input into a loss function together, and a loss value is calculated. And (4) iteratively updating the weight information of the network, and minimizing the loss function until the network converges to obtain a network model for fracture segmentation.
In the process of network testing, the prepared crack image to be segmented (without size requirement) is input into the trained network model after the preprocessing of mean value removal as shown in fig. 7, and the segmentation result image is output as shown in fig. 8.
In conclusion, the invention provides a model of a deep convolutional neural network for identifying the cracks of the tunnel lining image aiming at the requirement of high-precision identification in the tunnel lining image. Because the difference between the crack area and the background in the aspect of gray scale, gradient and the like is small in the crack image, the model based on the deep convolutional neural network is used for identifying and segmenting the cracks of the tunnel lining image, the construction joints, the later repair marks and the like which are easy to interfere with crack identification in the background can be accurately distinguished, the crack identification precision is improved, meanwhile, the false alarm rate is reduced, the human intervention is reduced, the influence of subjective factors of a detector is avoided, and the identification efficiency is improved.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (8)
1. A segmentation method of a tunnel lining crack image is characterized by comprising the following flow steps:
step S110: dividing an original crack image directly acquired into a plurality of image blocks, and selecting the image blocks with cracks and corresponding binary images as training sample sets;
step S120: carrying out mean value removing normalization processing on the image blocks with cracks in the training sample set;
step S130: training the training sample set subjected to mean value removal normalization processing by using a convolutional neural network in combination with a loss function to obtain a crack segmentation model;
step S140: and after mean value removing and normalization processing are carried out on the crack images to be segmented, inputting the crack images into the crack segmentation model, and outputting segmentation result images.
2. The segmentation method for the image of the tunnel lining crack as claimed in claim 1, wherein the step S110 specifically comprises:
dividing each directly acquired original crack image into a plurality of image blocks, wherein the number of pixels of each image block is lower than that of the original crack image, and selecting the image blocks with cracks and corresponding binary images as training sample sets.
3. The method for segmenting the tunnel lining fractured image according to claim 2, wherein the size of the original fractured image is 1000 x 4000 pixels, the number of the image blocks is 100, and the size of the image blocks is 200 x 200 pixels.
4. The segmentation method for the image of the tunnel lining crack according to claim 1, wherein the step S120 specifically comprises:
average gray value of image blocks with cracks in training sample setThe following constraints are satisfied:
wherein h represents the height of the image blocks in the training sample set, w represents the width of the image blocks in the training sample set, N is the total number of the image blocks in the training sample set,representing every n images in a training sample setThe gray value of the pixel point of the ith row and the jth column in the block;
removing the gray value X 'of the pixel point of the jth line and the jth line in the image block after the mean value normalization processing'i,jComprises the following steps:
wherein, Xi,jAnd expressing the gray value of the pixel point of the ith row and the jth column of the image block in the training sample set.
5. The segmentation method for the image of the tunnel lining crack of claim 4, wherein in the step S130, the convolutional neural network is a VGG-16 network.
6. The method for segmenting the image of the tunnel lining crack as claimed in claim 5, wherein in the step S130, the loss function is a class-balanced cross entropy loss function:
wherein the training sample set is S { (X)n，Yn)，n＝1，2，...，N},XnRepresenting the nth image block in the training sample set,whether the jth pixel in the binary image corresponding to the nth image block is a crack pixel or not is shown, 0 is the crack pixel, and 1 is a non-crack background pixel;
q represents the set of weight parameters of the convolutional neural network, the update is trained by back propagation, Y+Representing foreground crack pixels, Y-Representing a background non-slit pixel, β ═ Y-I/Y is used to balance the number of crack pixels with the number of non-crack pixels, and the probability P (-) indicates that the Sigmoid σ (-) function is used as the activation function at the last convolution layer.
7. The segmentation method for the image of the tunnel lining crack as claimed in claim 6, wherein the step S130 specifically comprises:
training by adopting a random gradient descent method, inputting a feature map output by an image block in a training sample set through a convolutional neural network and a binary image corresponding to the image block into a loss function together, and calculating a loss value; and (4) iteratively updating the weight information of the network, and minimizing the loss function until the network converges to obtain the fracture segmentation model.
8. The method for segmenting the tunnel lining crack image according to claim 7, wherein the training sample set is divided into a plurality of batches, each batch has 16 training samples, the momentum is 0.9, the weight attenuation is 0.0001, and the update rule of the weight q is as follows:
qm+1:＝qm+vm+1
wherein epsilon represents the learning rate, m represents the number of iterations,representing the derivative Q of the objective function with respect to QmAverage value of (a).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910012867.1A CN110020652A (en) | 2019-01-07 | 2019-01-07 | The dividing method of Tunnel Lining Cracks image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910012867.1A CN110020652A (en) | 2019-01-07 | 2019-01-07 | The dividing method of Tunnel Lining Cracks image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110020652A true CN110020652A (en) | 2019-07-16 |
Family
ID=67188736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910012867.1A Pending CN110020652A (en) | 2019-01-07 | 2019-01-07 | The dividing method of Tunnel Lining Cracks image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110020652A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415233A (en) * | 2019-07-26 | 2019-11-05 | 东南大学 | Pavement crack rapid extracting method based on two step convolutional neural networks |
| CN110895814A (en) * | 2019-11-30 | 2020-03-20 | 南京工业大学 | Intelligent segmentation method for aero-engine hole detection image damage based on context coding network |
| CN110910343A (en) * | 2019-09-29 | 2020-03-24 | 北京建筑大学 | Method and device for detecting pavement cracks and computer equipment |
| CN111080602A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting foreign matters in water leakage hole of railway wagon |
| CN111160301A (en) * | 2019-12-31 | 2020-05-15 | 同济大学 | Tunnel disease target intelligent identification and extraction method based on machine vision |
| CN111539286A (en) * | 2020-04-15 | 2020-08-14 | 创新奇智(合肥)科技有限公司 | Lining line identification method and device and readable storage medium |
| CN111932548A (en) * | 2020-04-21 | 2020-11-13 | 南京航空航天大学 | Subway tunnel fine crack segmentation method based on deep learning |
| CN112036425A (en) * | 2020-05-09 | 2020-12-04 | 中铁四局集团有限公司 | Tunnel cavity state radar spectrum image recognition model construction method and tunnel cavity state radar spectrum image recognition method |
| CN112698392A (en) * | 2020-11-20 | 2021-04-23 | 中国石油天然气股份有限公司 | Multilayer system reservoir three-dimensional drawing and oil-gas spatial distribution and constant volume method |
| CN115761613A (en) * | 2022-08-10 | 2023-03-07 | 中铁隧道局集团有限公司 | Automatic tunnel crack detection method based on convolutional network |
| WO2023207104A1 (en) * | 2022-04-26 | 2023-11-02 | 云南航天工程物探检测股份有限公司 | Ground penetrating radar tunnel lining quality inspection method based on self-supervised learning |
| CN117292266A (en) * | 2023-11-24 | 2023-12-26 | 河海大学 | Method and device for detecting concrete cracks of main canal of irrigation area and storage medium |
| CN118505690A (en) * | 2024-07-17 | 2024-08-16 | 华东交通大学 | Tunnel lining crack detection method and system based on deep learning and electronic equipment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107092960A (en) * | 2017-04-17 | 2017-08-25 | 中国民航大学 | A kind of improved parallel channel convolutional neural networks training method |
| CN107292887A (en) * | 2017-06-20 | 2017-10-24 | 电子科技大学 | A kind of Segmentation Method of Retinal Blood Vessels based on deep learning adaptive weighting |
| CN107977969A (en) * | 2017-12-11 | 2018-05-01 | 北京数字精准医疗科技有限公司 | A kind of dividing method, device and the storage medium of endoscope fluorescence image |
| CN108346144A (en) * | 2018-01-30 | 2018-07-31 | 哈尔滨工业大学 | Bridge Crack based on computer vision monitoring and recognition methods automatically |
| CN108520278A (en) * | 2018-04-10 | 2018-09-11 | 陕西师范大学 | A kind of road surface crack detection method and its evaluation method based on random forest |
| CN108710919A (en) * | 2018-05-25 | 2018-10-26 | 东南大学 | A kind of crack automation delineation method based on multi-scale feature fusion deep learning |
| CN109146849A (en) * | 2018-07-26 | 2019-01-04 | 昆明理工大学 | A kind of road surface crack detection method based on convolutional neural networks and image recognition |
-
2019
- 2019-01-07 CN CN201910012867.1A patent/CN110020652A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107092960A (en) * | 2017-04-17 | 2017-08-25 | 中国民航大学 | A kind of improved parallel channel convolutional neural networks training method |
| CN107292887A (en) * | 2017-06-20 | 2017-10-24 | 电子科技大学 | A kind of Segmentation Method of Retinal Blood Vessels based on deep learning adaptive weighting |
| CN107977969A (en) * | 2017-12-11 | 2018-05-01 | 北京数字精准医疗科技有限公司 | A kind of dividing method, device and the storage medium of endoscope fluorescence image |
| CN108346144A (en) * | 2018-01-30 | 2018-07-31 | 哈尔滨工业大学 | Bridge Crack based on computer vision monitoring and recognition methods automatically |
| CN108520278A (en) * | 2018-04-10 | 2018-09-11 | 陕西师范大学 | A kind of road surface crack detection method and its evaluation method based on random forest |
| CN108710919A (en) * | 2018-05-25 | 2018-10-26 | 东南大学 | A kind of crack automation delineation method based on multi-scale feature fusion deep learning |
| CN109146849A (en) * | 2018-07-26 | 2019-01-04 | 昆明理工大学 | A kind of road surface crack detection method based on convolutional neural networks and image recognition |
Non-Patent Citations (6)
| Title |
|---|
| 孙文雅等: "基于BP神经网络管道裂缝图像分割", 《计算机测量与控制》 * |
| 宋俊芳: "基于CNN的路面裂缝图像分类", 《信息与电脑(理论版)》 * |
| 张文达等: "基于多尺度分块卷积神经网络的图像目标识别算法", 《计算机应用》 * |
| 柴雪松等: "基于深度卷积神经网络的隧道衬砌裂缝识别算法", 《铁道建筑》 * |
| 薛亚东等: "基于深度学习的盾构隧道衬砌病害识别方法", 《湖南大学学报(自然科学版)》 * |
| 赵珊珊等: "基于卷积神经网络的路面裂缝检测", 《传感器与微系统》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415233A (en) * | 2019-07-26 | 2019-11-05 | 东南大学 | Pavement crack rapid extracting method based on two step convolutional neural networks |
| CN110910343A (en) * | 2019-09-29 | 2020-03-24 | 北京建筑大学 | Method and device for detecting pavement cracks and computer equipment |
| CN110895814A (en) * | 2019-11-30 | 2020-03-20 | 南京工业大学 | Intelligent segmentation method for aero-engine hole detection image damage based on context coding network |
| CN111080602A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting foreign matters in water leakage hole of railway wagon |
| CN111080602B (en) * | 2019-12-12 | 2020-10-09 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting foreign matters in water leakage hole of railway wagon |
| CN111160301B (en) * | 2019-12-31 | 2023-04-18 | 同济大学 | Tunnel disease target intelligent identification and extraction method based on machine vision |
| CN111160301A (en) * | 2019-12-31 | 2020-05-15 | 同济大学 | Tunnel disease target intelligent identification and extraction method based on machine vision |
| CN111539286A (en) * | 2020-04-15 | 2020-08-14 | 创新奇智(合肥)科技有限公司 | Lining line identification method and device and readable storage medium |
| CN111932548A (en) * | 2020-04-21 | 2020-11-13 | 南京航空航天大学 | Subway tunnel fine crack segmentation method based on deep learning |
| CN112036425A (en) * | 2020-05-09 | 2020-12-04 | 中铁四局集团有限公司 | Tunnel cavity state radar spectrum image recognition model construction method and tunnel cavity state radar spectrum image recognition method |
| CN112698392A (en) * | 2020-11-20 | 2021-04-23 | 中国石油天然气股份有限公司 | Multilayer system reservoir three-dimensional drawing and oil-gas spatial distribution and constant volume method |
| WO2023207104A1 (en) * | 2022-04-26 | 2023-11-02 | 云南航天工程物探检测股份有限公司 | Ground penetrating radar tunnel lining quality inspection method based on self-supervised learning |
| CN115761613A (en) * | 2022-08-10 | 2023-03-07 | 中铁隧道局集团有限公司 | Automatic tunnel crack detection method based on convolutional network |
| CN115761613B (en) * | 2022-08-10 | 2023-09-01 | 中铁隧道局集团有限公司 | Automatic tunnel crack detection method based on convolutional network |
| CN117292266A (en) * | 2023-11-24 | 2023-12-26 | 河海大学 | Method and device for detecting concrete cracks of main canal of irrigation area and storage medium |
| CN117292266B (en) * | 2023-11-24 | 2024-03-22 | 河海大学 | Method and device for detecting concrete cracks of main canal of irrigation area and storage medium |
| CN118505690A (en) * | 2024-07-17 | 2024-08-16 | 华东交通大学 | Tunnel lining crack detection method and system based on deep learning and electronic equipment |
| CN118505690B (en) * | 2024-07-17 | 2024-10-11 | 华东交通大学 | Tunnel lining crack detection method and system based on deep learning and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110020652A (en) | The dividing method of Tunnel Lining Cracks image | |
| CN107480611B (en) | Crack identification method based on deep learning convolutional neural network | |
| CN108776772B (en) | Cross-time building change detection modeling method, detection device, method and storage medium | |
| CN108960135B (en) | Dense ship target accurate detection method based on high-resolution remote sensing image | |
| CN110415233A (en) | Pavement crack rapid extracting method based on two step convolutional neural networks | |
| KR102202572B1 (en) | road crack assessment method and system using drone image and artificial intelligent algorithm | |
| CN108460764A (en) | The ultrasonoscopy intelligent scissor method enhanced based on automatic context and data | |
| CN111241994B (en) | Deep learning remote sensing image rural highway sanded road section extraction method | |
| US20220315243A1 (en) | Method for identification and recognition of aircraft take-off and landing runway based on pspnet network | |
| CN114022770A (en) | Mountain crack detection method based on improved self-attention mechanism and transfer learning | |
| CN111832607B (en) | Bridge disease real-time detection method based on model pruning | |
| CN110930320A (en) | Image defogging method based on lightweight convolutional neural network | |
| CN109064449B (en) | Method for detecting bridge surface diseases | |
| CN110909657A (en) | Method for identifying apparent tunnel disease image | |
| CN113256601A (en) | Pavement disease detection method and system | |
| CN110751644A (en) | Road surface crack detection method | |
| CN113762265B (en) | Classified segmentation method and system for pneumonia | |
| CN115719475B (en) | Three-stage trackside equipment fault automatic detection method based on deep learning | |
| CN111735815A (en) | Method and device for detecting defects of small hardware fittings of power transmission line and storage medium | |
| CN114049538A (en) | Airport crack image confrontation generation method based on UDWGAN + + network | |
| CN116664488A (en) | Bridge surface disease and structural damage monitoring method based on mobile phone app and machine vision | |
| CN115272826A (en) | Image identification method, device and system based on convolutional neural network | |
| CN115205255A (en) | Stone automatic grading method and system based on deep learning | |
| CN116246240A (en) | Pavement identification model building method, pavement damage disease automatic identification method and system | |
| CN118072193A (en) | Dam crack detection method based on unmanned aerial vehicle image and deep learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190716 |