CN116228641A - Micro fatigue crack length calculation method based on U-net network - Google Patents
Micro fatigue crack length calculation method based on U-net network Download PDFInfo
- Publication number
- CN116228641A CN116228641A CN202211605743.2A CN202211605743A CN116228641A CN 116228641 A CN116228641 A CN 116228641A CN 202211605743 A CN202211605743 A CN 202211605743A CN 116228641 A CN116228641 A CN 116228641A
- Authority
- CN
- China
- Prior art keywords
- layer
- net network
- convolution
- fatigue crack
- module
- 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
Images
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/30164—Workpiece; Machine component
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
Abstract
The invention discloses a micro fatigue crack length calculation method based on a U-net network, which relates to the technical field of crack detection and comprises the following steps: training the optimized U-net network by using a certain type of initial images of the fretting fatigue cracks to obtain a trained U-net network model; acquiring the micro fatigue crack image similar to the step S1 as an experimental image; identifying micro fatigue cracks in the experimental image by adopting the trained U-net network model to obtain a segmentation detection result graph; and measuring the length of the fretting fatigue crack in the segmentation detection result graph. The first convolution pooling layer group and the fusion layer which are sequentially connected are arranged between the up-sampling module and the down-sampling module in the U-net network structure, so that the calculated amount can be reduced, the network speed can be improved, and meanwhile, the characteristic diagram is expanded to capture local information and detail information, so that the characteristic information of the micro fatigue crack can be accurately extracted.
Description
Technical Field
The invention relates to the technical field of crack detection, in particular to a micro fatigue crack length calculation method based on a U-net network.
Background
Under long-term high-frequency micro-motion working conditions, micro-motion fatigue cracks are easy to generate locally, along with the increase of load and cycle times, the crack growth rate can be continuously increased, the early micro-fatigue cracks can continue to develop into obvious cracks, the local rigidity of the part is weakened, even local damage is induced, and the safety of the whole mechanical structure is threatened. Therefore, the length of the fretting fatigue crack is monitored, the specific influence of the fretting crack on the equipment performance is quantified, and the method has important significance for predicting the safety life of the whole mechanical structure.
In the prior art, a patent document with publication number of CN111445446B discloses an improved U-net-based concrete surface crack detection method, wherein the calculated amount of a neural network template used by the method is relatively large, and only long cracks can be obtained; patent document with publication number of CN113284107A discloses a concrete crack real-time detection method of an attention-drawing mechanism improved U-net, but the detection model only can acquire micro cracks and cannot acquire length information of the micro cracks.
In general, the crack is identified by using a deep learning mode at present, but the micro fatigue crack size is usually quite fine, generally between one hundred micrometers and several hundred micrometers, and the micro fatigue crack cannot be identified effectively by using the existing crack detection technology based on the deep learning because the micro fatigue crack size is extremely small and cannot be obviously distinguished from the surface of a sample, so that an identification method suitable for the micro fatigue crack of an engine blade tenon sample and the like needs to be developed for identifying the micro crack, thereby realizing early identification of the micro fatigue crack.
Disclosure of Invention
In view of the above problems, an object of the present invention is to provide a method for calculating the length of fretting fatigue cracks based on a U-net network, which is suitable for monitoring the lengths of fretting fatigue cracks generated in fretting of parts such as a tenon sample of an engine blade with different cycle numbers.
The technical scheme of the invention is as follows:
a micro fatigue crack length calculation method based on a U-net network comprises the following steps:
s1, training an optimized U-net network by using a certain type of initial images of fretting fatigue cracks to obtain a trained U-net network model, and specifically comprising the following steps of:
s11, acquiring original images of some kinds of fretting fatigue cracks, wherein the fretting fatigue cracks generated by different equipment and different movement modes are different, so that the acquired original images are the fretting fatigue cracks obtained by the same movement mode of the same equipment, such as the fretting fatigue cracks generated by the cyclic stress of the tenon sample of the engine blade, such as the fretting fatigue cracks generated by the cyclic stress of the steel wire, such as the fatigue cracks generated by the cyclic stress of the axle of the high-speed train.
And S12, marking the crack areas in the original image pixel by pixel so that the crack areas form a closed graph, wherein specific standard software is various, such as LabelMe, labelimg and rolabelmg.
S13, the marked original images are arranged into a complete data set format in batches by using the image segmentation kit, the complete data set format is used as a training set, a testing set and a verification set of the optimized U-net model, and the trained U-net model is trained to obtain a trained U-net network model. The image segmentation suite of this step is not limited to the PaddleSeg and vox datasets.
S2, acquiring an fretting fatigue crack image similar to the step S1 as an experimental image; and carrying out segmentation detection on the experimental image by adopting the trained U-net network model to identify the fretting fatigue crack in the experimental image so as to obtain a segmentation detection result graph.
S3, measuring the length of the fretting fatigue crack in the segmentation detection result graph, wherein a plurality of specific measuring modes exist, and the invention provides a specific measuring method which comprises the following steps:
s31, generating a central axis of the fretting fatigue crack in the segmentation detection result diagram through a central axis algorithm;
s32, calculating the length of the central axis to obtain the length of the fretting fatigue crack: and counting the number of the pixels on the central axis of the fretting fatigue crack, and obtaining the total length of the central axis of the fretting fatigue crack by unit conversion of the total number of the pixels according to the resolution size of the segmentation detection result graph.
In the U-Net network structure, the probability that pixels are correctly marked is reduced by downsampling, the capability of the upsampling of the U-Net network to recover characteristic information is limited, so that the width information of cracks is not obvious, and the phenomenon of tiny cracks cannot be detected. On the premise of maximum receptive field, the maximum resolution is maintained, so that the detection of the fine cracks is realized, and specifically, the optimized U-net network comprises the following components.
The optimized U-net network comprises an up-sampling module, a connection operation, a down-sampling module and a layer fusion module positioned at the bottom of the up-sampling module and between the bottoms of the down-sampling modules;
the layer fusion module comprises a first convolution pooling layer group and a fusion layer which are connected in sequence; the first convolution pooling layer group comprises 1 depth separable convolution module and 4 first convolution pooling layers which are connected in parallel, and each first convolution pooling layer comprises a maximum pooling layer with a step length of 2, a 1 multiplied by 1 depth separable convolution module and a 2 multiplied by 2 upsampling layer; the feature map output by the downsampling module is input into a first convolution pooling layer group, the fusion layer performs addition fusion on the input image features, and the feature map output by the fusion layer is input into the upsampling module; the U-net network adopts the nonlinear activation function ReLU to process the output result of each layer, so that the nonlinear expression of the network can be improved, and the gradient disappearance phenomenon in the convolution process can be reduced.
The depth of the eigenvector of the U-net network is 2048, the up-sampling module and the down-sampling module comprise five second convolution layer groups which are sequentially connected, and each second convolution layer group comprises a residual block, an attention mechanism module and 2 transposition convolution layers which are sequentially connected;
the residual block comprises a depth separable convolution layer, a transposed convolution layer and a BN layer, and the input of the residual block is fused with the output of the residual block after convolution in the addition layer through skip connection;
the attention mechanism module comprises an encoding block and a decoding block, wherein the encoding block comprises a maximum pooling layer and a convolution layer, and the decoding block comprises an upsampling layer and a convolution layer; the characteristic information output by the residual block is input into an attention mechanism, the soft attention mechanism is excessively parameterized during training, the input of the attention mechanism module is fused with the output of the attention mechanism module after the Sigmoid function is classified by a skip connection, and the fusion is used as an output, so that the problem of redundant parameter calculation in the soft attention mechanism is solved, and the convolutional neural network is optimized.
The Concate operation is to splice the feature map generated by the up-sampling module and the feature map sampled by the down-sampling module by an np-con cate function, a BN layer is added before a nonlinear activation function ReLU in the Concate operation, the input distribution of the upper layer is relieved to be slowly close to two ends of the nonlinear function, the BN layer performs normalization processing of normal distribution of N (0, 1) on input data, and finally the normalization processing is input to the activation function ReLU, a more obvious gradient can be generated in the back propagation, and the network is effectively helped to converge so as to improve the gradient dispersion phenomenon.
The beneficial effects are that:
(1) The method can accurately identify the tiny fatigue cracks and measure the lengths of the cracks.
(2) The first convolution pooling layer group and the fusion layer which are sequentially connected are arranged between the up-sampling module and the down-sampling module in the U-net network structure, so that the calculated amount can be reduced, the network speed can be improved, and meanwhile, the characteristic diagram is expanded to capture local information and detail information, so that tiny fretting fatigue cracks can be accurately identified. In addition, the channel dimension is spliced through layer fusion, so that training time of a network model can be reduced, a feature map can be enlarged, and resolution can be increased. The depth separable convolution module is arranged in the first convolution pooling layer group, so that the parameter calculation amount can be effectively reduced, and meanwhile, the convolution kernel size of the separable convolution module is set to be 1 multiplied by 1, so that the number of channels can be reduced.
Drawings
FIG. 1 is a flow chart of an optimized U-net network according to embodiment 1 of the present invention;
FIG. 2 is a flow chart of a layer 1 fusion module according to an embodiment of the present invention;
FIG. 3 is a block flow diagram of a residual block of embodiment 1 of the present invention;
FIG. 4 is a flow chart of the attention mechanism module of embodiment 1 of the present invention;
fig. 5 is a flowchart of crack length calculation of the experimental image of example 1 of the present invention.
Detailed Description
For a clearer understanding of the technical features, objects and advantages of the present invention, an embodiment of the present invention will be further described with reference to the accompanying drawings. The examples are intended to be illustrative only and are not to be construed as limiting the scope of the invention, as many insubstantial modifications and variations that may be made by a person skilled in the art in light of the teachings of this invention are intended to fall within the scope of this invention.
Example 1
The embodiment takes micro fatigue cracks generated by engine blade tenons through cyclic vibration as an example, and the method for calculating the length of the micro fatigue cracks specifically comprises the following steps:
s1, training an optimized U-net network by using an initial image of micro fatigue cracks generated by engine blade tenons through cyclic vibration to obtain a trained U-net network model, and specifically comprising the following steps of:
s11, acquiring an original image of micro fatigue cracks generated by engine blade tenons through cyclic vibration: in order to obtain clear initial images of micro fatigue cracks, the embodiment adopts a high-frequency fatigue testing machine of a central machine GPS100, a digital microscope is placed on the side surface of a tenon sample on the basis of the device to monitor crack growth conditions of two sides of the tenon sample, and image data are collected.
S12, marking the crack region in the original image pixel by using LabelMe software, so that the crack region forms a closed pattern.
S13, using an image segmentation kit PaddleG to arrange the marked original images into a complete data set format in batches and using the complete data set format as a training set, a test set and a verification set of the optimized U-net model, and training the optimized U-net model to obtain a trained U-net network model.
In the U-Net network structure, the probability that pixels are correctly marked is reduced by downsampling, the capability of the upsampling of the U-Net network to recover characteristic information is limited, so that the width information of cracks is not obvious, and the phenomenon of tiny cracks cannot be detected. On the premise of maximum receptive field, the maximum resolution is maintained, so that the detection of the fine cracks is realized, and specifically, the optimized U-net network comprises the following components.
Fig. 1 is a flow chart of an optimized U-net network according to this embodiment, where the optimized U-net network includes an upsampling module, a connect operation, a downsampling module, and a layer fusion module located at the bottom of the upsampling module and between the bottoms of the downsampling modules.
FIG. 2 is a flow diagram of a layer fusion module including a first convolution pooling layer group and a fusion layer connected in sequence; the first convolution pooling layer group comprises 1 depth separable convolution module and 4 first convolution pooling layers which are connected in parallel, and each first convolution pooling layer comprises a maximum pooling layer with a step length of 2, a 1 multiplied by 1 depth separable convolution module and a 2 multiplied by 2 upsampling layer; the feature map output by the downsampling module is input into a first convolution pooling layer group, the fusion layer performs addition fusion on the input image features, and the feature map output by the fusion layer is input into the upsampling module; the U-net network adopts the nonlinear activation function ReLU to process the output result of each layer, so that the nonlinear expression of the network can be improved, and the gradient disappearance phenomenon in the convolution process can be reduced.
With continued reference to fig. 2, the up-sampling module and the down-sampling module each include five second convolution layer groups connected in sequence, where each second convolution layer group includes a residual block, an attention mechanism module and 2 transposed convolution layers connected in sequence.
Referring to fig. 3, fig. 3 is a block flow diagram of a residual block, where the residual block includes a depth separable convolution, a transposed convolution layer, and a BN layer, and an input of the residual block is fused with an output of the residual block after convolution at an addition layer through skip connection.
Referring to fig. 4, fig. 4 is a flow diagram of an attention mechanism module, the attention mechanism module including an encoding block including a max-pooling layer and a convolutional layer, and a decoding block including an upsampling layer and a convolutional layer; the characteristic information output by the residual block is input into an attention mechanism, the attention mechanism is excessively parameterized during training, the input of the attention mechanism module is fused with the output of the attention mechanism module after the classification of the Sigmoid function through a skip connection, and the output is formed after the fusion, so that the problem of redundant parameter calculation in a soft attention mechanism is solved, and the convolutional neural network is optimized.
The Concate operation is to splice the feature map generated by the up-sampling module and the feature map sampled by the down-sampling module by an np-con cate function, a BN layer is added before a nonlinear activation function ReLU in the Concate operation, the input distribution of the upper layer is relieved to be slowly close to two ends of the nonlinear function, the BN layer performs normalization processing of normal distribution of N (0, 1) on input data, and finally the normalization processing is performed on the input data, the normalization processing is finally input to the value of the activation function ReLU, a more obvious gradient can be generated in the back propagation, and the network is effectively helped to converge so as to improve the gradient dispersion phenomenon.
According to the embodiment, the first convolution pooling layer group and the fusion layer which are sequentially connected are arranged between the up-sampling module and the down-sampling module, so that the calculated amount can be reduced, the network speed is improved, and meanwhile, the feature map is enlarged to capture local information and detail information. In addition, the channel dimension is spliced through layer fusion, so that training time of a network model can be reduced, a feature map can be enlarged, and resolution can be increased. The depth separable convolution module is arranged in the first convolution pooling layer group, so that the parameter calculation amount can be effectively reduced, and meanwhile, the convolution kernel size of the separable convolution module is set to be 1 multiplied by 1, so that the number of channels can be reduced.
S2, acquiring an fretting fatigue crack image similar to the step S1 as an experimental image; dividing, detecting and identifying fretting fatigue cracks in the experimental image by adopting the trained U-net network model to obtain a division detection result graph; acquiring a central axis in a crack segmentation result diagram by adopting a central axis algorithm; and counting the number of the pixel points contained in the central axis by adopting a non-zero element counting function, wherein a specific detection and identification flow is shown in figure 5.
S3, measuring the length of the fretting fatigue crack in the segmentation detection result graph, and specifically comprising the following steps of:
s31, generating a central axis of the fretting fatigue crack in the segmentation detection result diagram through a central axis algorithm.
S32, calculating the length of the central axis to obtain the length of the fretting fatigue crack: and counting the number of the pixels on the central axis of the fretting fatigue crack, and obtaining the total length of the central axis of the fretting fatigue crack by unit conversion of the total number of the pixels according to the resolution size of the segmentation detection result graph. For example, the number of pixels of a central axis of a certain crack in a certain image is counted to be 80, and the number of pixels of a diagonal line of the image is 1000, and the length of the diagonal line is 43cm, that is, in this case, 1000 px=43 cm, that is, 1 px=0.43 mm, and the length of the central axis is: 80 x 0.43mm = 34.4mm.
The foregoing description of the invention has been presented to enable one of ordinary skill in the art to practice the invention based on such description. Based on the foregoing, all other embodiments that may be obtained by one of ordinary skill in the art without undue burden are within the scope of the present invention.
Claims (4)
1. The micro fatigue crack length calculating method based on the U-net network is characterized by comprising the following steps of:
s1, training an optimized U-net network by using an original image of a certain type of fretting fatigue crack to obtain a trained U-net network model;
s2, acquiring an fretting fatigue crack image similar to the step S1 as an experimental image; dividing, detecting and identifying fretting fatigue cracks in the experimental image by adopting the trained U-net network model to obtain a division detection result graph;
s3, measuring the length of the fretting fatigue crack in the segmentation detection result graph;
the optimized U-net network comprises an up-sampling module, a connection operation, a down-sampling module and a layer fusion module positioned at the bottom of the up-sampling module and between the bottoms of the down-sampling modules;
the layer fusion module comprises a first convolution pooling layer group and a fusion layer which are connected in sequence; the first convolution pooling layer group comprises 1 depth separable convolution module and 4 first convolution pooling layers which are connected in parallel, and each first convolution pooling layer comprises a maximum pooling layer with a step length of 2, a 1 multiplied by 1 depth separable convolution module and a 2 multiplied by 2 upsampling layer; the feature map output by the downsampling module is input into a first convolution pooling layer group, the fusion layer performs addition fusion on the input image features, and the feature map output by the fusion layer is input into the upsampling module; the nonlinear activation function adopted by the U-net network is ReLU;
the depth of the eigenvector of the U-net network is 2048, the up-sampling module and the down-sampling module comprise five second convolution layer groups which are sequentially connected, and each second convolution layer group comprises a residual block, an attention mechanism module and 2 transposition convolution layers which are sequentially connected;
the residual block comprises a depth separable convolution layer, a transposed convolution layer and a BN layer, and the input of the residual block is fused with the output of the residual block after convolution in the addition layer through skip connection;
the attention mechanism module comprises an encoding block and a decoding block, wherein the encoding block comprises a maximum pooling layer and a convolution layer, and the decoding block comprises an upsampling layer and a convolution layer; inputting the characteristic information output by the residual block into an attention mechanism module, and fusing the input of the attention mechanism module with the output of the attention mechanism module after the classification of the Sigmoid function by a skip connection;
splicing the feature map generated by the up-sampling module and the feature map subjected to down-sampling by an np-concatate function, adding a BN layer before a nonlinear activation function ReLU in the concatate operation, performing normalization processing of normal distribution of N (0, 1) on input data, and inputting the normalized data to the activation function ReLU.
2. The micro fatigue crack length calculating method based on the U-net network according to claim 1, wherein the step S1 includes the steps of:
s11, obtaining an original image of a certain type of fretting fatigue crack;
s12, marking the crack areas in the original image pixel by pixel so that the crack areas form a closed graph;
s13, the marked original images are arranged into a complete data set format in batches by using the image segmentation kit, the complete data set format is used as a training set, a testing set and a verification set of the optimized U-net model, and the trained U-net model is trained to obtain a trained U-net network model.
3. The micro fatigue crack length calculating method based on the U-net network according to claim 1, wherein the step S3 includes the steps of:
s31, generating a central axis of the fretting fatigue crack in the segmentation detection result diagram through a central axis algorithm;
s32, calculating the length of the central axis to obtain the length of the fretting fatigue crack.
4. The micro fatigue crack length calculating method based on the U-net network according to claim 3, wherein the step S32 includes the steps of:
and counting the number of the pixels on the central axis of the fretting fatigue crack, and obtaining the total length of the central axis of the fretting fatigue crack by unit conversion of the total number of the pixels according to the resolution size of the segmentation detection result graph.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211605743.2A CN116228641A (en) | 2022-12-14 | 2022-12-14 | Micro fatigue crack length calculation method based on U-net network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211605743.2A CN116228641A (en) | 2022-12-14 | 2022-12-14 | Micro fatigue crack length calculation method based on U-net network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116228641A true CN116228641A (en) | 2023-06-06 |
Family
ID=86584901
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211605743.2A Pending CN116228641A (en) | 2022-12-14 | 2022-12-14 | Micro fatigue crack length calculation method based on U-net network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116228641A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117710348A (en) * | 2023-12-21 | 2024-03-15 | 广州恒沙云科技有限公司 | Pavement crack detection method and system based on position information and attention mechanism |
-
2022
- 2022-12-14 CN CN202211605743.2A patent/CN116228641A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117710348A (en) * | 2023-12-21 | 2024-03-15 | 广州恒沙云科技有限公司 | Pavement crack detection method and system based on position information and attention mechanism |
| CN117710348B (en) * | 2023-12-21 | 2024-06-11 | 广州恒沙云科技有限公司 | Pavement crack detection method and system based on position information and attention mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101896406B1 (en) | Road crack detection apparatus of pixel unit and method thereof, and computer program for executing the same | |
| JP2020126601A (en) | Fault detecting method and fault detecting device | |
| CN110660021B (en) | Pulse eddy current thermal imaging high-resolution sharpening method based on deep learning | |
| CN116552306B (en) | Monitoring system and method for direct current pile | |
| CN116228641A (en) | Micro fatigue crack length calculation method based on U-net network | |
| CN113077444A (en) | CNN-based ultrasonic nondestructive detection image defect classification method | |
| CN111899216A (en) | Abnormity detection method for insulator fastener of high-speed rail contact network | |
| CN118011994A (en) | Cross-modal industrial large model-based fault diagnosis system and method | |
| CN115018012B (en) | Internet of things time sequence anomaly detection method and system under high dimensionality characteristics | |
| CN117422695A (en) | CR-deep-based anomaly detection method | |
| CN115937651A (en) | Cylindrical roller surface detection method and system based on improved yolov5s network model | |
| CN111652212A (en) | Method for detecting loss fault of fastening bolt at end part of cross rod based on deep learning | |
| CN115330697A (en) | Tire flaw detection domain self-adaption method based on migratable Swin transducer | |
| CN114354195A (en) | Rolling bearing fault diagnosis method of depth domain self-adaptive convolution network | |
| CN113516652A (en) | Battery surface defect and adhesive detection method, device, medium and electronic equipment | |
| Zhao et al. | Image tampering detection via semantic segmentation network | |
| CN111210409B (en) | Condition-based generation confrontation network structure damage identification method | |
| CN117274197A (en) | PCB defect detection method based on YOLO v5 algorithm improvement | |
| CN115452373B (en) | Rolling bearing fault diagnosis method based on event camera | |
| CN112164040A (en) | Steel surface defect identification method based on semi-supervised deep learning algorithm | |
| CN104715257A (en) | Image median filtering detection method and device | |
| CN116380466A (en) | Rolling bearing intelligent fault diagnosis method and system based on enhanced event visual data | |
| CN112508946A (en) | Cable tunnel abnormity detection method based on antagonistic neural network | |
| CN117128862B (en) | Scanning detection analysis method for automobile parts | |
| CN113658112B (en) | Bow net anomaly detection method based on template matching and neural network algorithm |
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 |