CN111932548A - Subway tunnel fine crack segmentation method based on deep learning - Google Patents
Subway tunnel fine crack segmentation method based on deep learning Download PDFInfo
- Publication number
- CN111932548A CN111932548A CN202010319469.7A CN202010319469A CN111932548A CN 111932548 A CN111932548 A CN 111932548A CN 202010319469 A CN202010319469 A CN 202010319469A CN 111932548 A CN111932548 A CN 111932548A
- Authority
- CN
- China
- Prior art keywords
- image
- deep learning
- tunnel
- subway tunnel
- crack
- 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
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a subway tunnel micro crack segmentation method based on deep learning, which comprises the following specific steps of S1: automatically acquiring images of the surface of the subway tunnel by using a mobile trolley, manually selecting and marking the images containing cracks, and establishing a tunnel surface crack image database; s2: carrying out image preprocessing on an original tunnel image to construct a deep learning model; s3: and identifying and segmenting the fine cracks of the subway tunnel. The invention provides an efficient automatic detection and segmentation method for the fine cracks of the subway tunnel surface image, saves the labor cost, improves the detection efficiency, effectively avoids the wrong selection and the missed selection caused by artificial subjective reasons, and has important practical application significance.
Description
Technical Field
The invention relates to the field of image processing and deep learning, in particular to a subway tunnel micro-crack segmentation method based on deep learning.
Background
In recent years, with the increase of the construction strength of urban subway tunnels in China, urban rail transit is rapidly developed. However, under the influence of the environment and load, as the use time of the tunnel increases, the tunnel surface inevitably has structural health problems. Cracks are the most common tunnel defects, are important reasons for damaging the tunnel structure and seriously affect the traffic safety operation of the tunnel. The traditional tunnel surface crack detection mainly adopts manual visual detection, is low in efficiency and is easily influenced by artificial subjective factors. Therefore, the development of an efficient automatic subway tunnel crack segmentation method is of great significance.
At present, with the greatly improved processing capability of the GPU, the deep learning technology is rapidly developed and is successfully applied to tasks such as image recognition, classification and image segmentation. However, the conventional deep learning method for crack detection has a problem of low segmentation accuracy, and it is difficult to effectively segment a fine crack.
Disclosure of Invention
The invention aims to solve the technical problem that a novel subway tunnel fine crack segmentation method based on deep learning is provided to solve the problems that in the prior art, a deep learning method for crack detection has low segmentation precision and effective segmentation of fine cracks is difficult to realize.
In order to solve the technical problems, the technical scheme of the invention is as follows: the method for segmenting the fine cracks of the subway tunnel based on deep learning has the innovation points that: the method comprises the following steps:
s1, acquiring an image and establishing a database: automatically acquiring images of the surface of the subway tunnel by using a mobile trolley, manually selecting and marking the images containing cracks, and establishing a tunnel surface crack image database;
s2, establishing a learning model: carrying out image preprocessing on an original tunnel image to construct a deep learning model;
and S3, identifying and segmenting the fine cracks of the subway tunnel.
Further, the step S1 specifically includes the following steps:
(1) moving a moving trolley carrying a plurality of cameras in the tunnel to automatically acquire the surface image of the subway tunnel;
(2) carrying out image splicing operation on an original image acquired by a camera to obtain a real tunnel section panoramic image, and cutting the panoramic image to obtain a picture with a fixed size;
(3) and expanding the subway tunnel surface image containing the cracks, carrying out manual annotation, and establishing a tunnel surface crack image database.
Further, the step (2) of performing image stitching operation on the original image to obtain a real tunnel section panoramic image specifically comprises the following steps:
A. feature extraction: extracting local features of the image by using a Scale Invariant Feature Transform (SIFT) algorithm;
B. and (3) feature matching: matching the characteristic points of the image to be spliced by adopting a mode of comparing the nearest neighbor distance with the next nearest neighbor distance, and eliminating mismatching characteristic point pairs by utilizing a random sampling consistency RANSAC algorithm;
C. image fusion: and calculating a mathematical conversion model according to the matched characteristic point pairs, converting the images to be spliced into a coordinate system of a reference image, completing the same coordinate transformation, and fusing the overlapped areas of the images to be spliced by adopting a weighted average image fusion method to obtain a reconstructed smooth seamless panoramic image.
Further, the process of preprocessing the image in step S2 is as follows: performing contrast enhancement processing on the original image by adopting a histogram equalization method, and improving the image quality of the data set; the process of constructing the deep learning model comprises the following steps: and constructing a depth convolution encoder-decoder model for image segmentation based on deep learning, and performing iterative training and weight updating on the encoder-decoder model by using a tunnel crack image database to generate a deep learning model.
Further, the encoder-decoder model includes an encoder for feature extraction and a decoding network for feature fusion and down-sampling of images,
further, a specific method for performing iterative training and weight updating on the encoder-decoder model by using the tunnel crack image database comprises the following steps: the encoder part uses conv1-conv5 convolutional network (including maximum pooling layer therein) of ResNet101, and feature maps generated by the last convolutional layer of conv3, conv4 and conv5 are respectively marked as feature1, feature2 and feature 3; the decoder first downsamples feature1 to the same size as the image resolution of feature2, then performs feature fusion on the downsampled result graph and feature2, then downsamples the Fused result graph to match the resolution of feature3, performs feature fusion on the downsampled result graph and feature3 to obtain a Fused feature map, performs dimensionality reduction processing with a 1 × 1 convolution kernel, and uses data-dependent upsampling to obtain a final result graph.
Further, the encoder-decoder deep learning model comprises a loss function training model. The calculation formula of the loss function training model is as follows:
L(F,Y)=||F-Y′||2
wherein, F represents the generated feature map, Y represents the corresponding ground-route, Y 'represents the image obtained by reducing the resolution of Y, and Y' has the same resolution as F.
Further, the process of Y deresolved generating Y' includes the steps of:
(1) let Y be an element of RH×W×CDividing Y into H/r multiplied by W/r sub-windows, wherein the size of each sub-window S is r multiplied by C;
(2) for each window S, the value range of each pixel point is [0,1 ]]It is converted into a vector v ∈ {0, 1}N,N=r*r*C;
(3) The vector v is compressed into a low-dimensional vector x by using linear projection, and the compressed x corresponding to all H/r multiplied by W/r data blocks is combined into Y'.
Further, the linear projective transformation formula in step 3 is:
x=Pv;v′=Wx
wherein P ∈ RC′×NFor compressing v to x, W ∈ RN×C′For the reconstruction matrix, x is reconstructed as v, and v' represents the reconstructed vector v.
Further, the process of crack identification and segmentation in step S3 is as follows: and inputting the image to be detected into the deep learning model trained in the step S2, and outputting the segmentation result of the fine cracks of the tunnel surface image.
Compared with the prior art, the invention has the following beneficial effects:
the novel subway tunnel fine crack segmentation method based on deep learning can realize high-efficiency and high-precision segmentation of fine cracks in tunnel surface images, effectively saves labor cost, improves detection efficiency, and has important practical application significance.
Drawings
FIG. 1 is a flow chart of a subway tunnel micro crack segmentation method based on deep learning according to the present invention;
FIG. 2 is a schematic diagram of the deep learning model of the present invention;
FIG. 3 is a schematic diagram of data dependent upsampling;
FIG. 4 is a schematic representation of the fracture image segmentation results of the present invention;
FIG. 5 is a statistical table of fracture segmentation results of the present invention;
Detailed Description
To further illustrate the working procedures of the present invention in detail, the present invention is further described in detail with reference to the accompanying drawings and specific embodiments.
The invention provides a subway tunnel micro crack segmentation method based on deep learning, which has the specific flow as shown in figure 1, and comprises the following steps:
s1, acquiring an image and establishing a database:
the image acquisition system of the invention mainly comprises: the method comprises the following steps that a mobile trolley, a CCD camera, illumination equipment, an image acquisition card and a computer are used, images of the surface of the subway tunnel are automatically acquired by the mobile trolley, the images containing cracks are manually selected and marked, and a tunnel surface crack image database is established; the method specifically comprises the following steps:
(1) moving a moving trolley carrying a plurality of cameras in the tunnel to automatically acquire the surface image of the subway tunnel;
(2) carrying out image splicing operation on an original image acquired by a camera to obtain a real tunnel section panoramic image, and cutting the panoramic image to obtain a picture with a fixed size; the method comprises the following specific steps of carrying out image splicing operation on an original image to obtain a real tunnel section panoramic image:
A. feature extraction: extracting local features of the image by using a Scale Invariant Feature Transform (SIFT) algorithm;
B. and (3) feature matching: matching the characteristic points of the image to be spliced by adopting a mode of comparing the nearest neighbor distance with the next nearest neighbor distance, and eliminating mismatching characteristic point pairs by utilizing a random sampling consistency RANSAC algorithm;
C. image fusion: and calculating a mathematical conversion model according to the matched characteristic point pairs, converting the images to be spliced into a coordinate system of a reference image, completing the same coordinate transformation, and fusing the overlapped areas of the images to be spliced by adopting a weighted average image fusion method to obtain a reconstructed smooth seamless panoramic image.
(3) The method comprises the steps of expanding subway tunnel surface images containing cracks, carrying out manual labeling, and establishing a tunnel surface crack image database, wherein 2045 original crack images are obtained in the embodiment of the invention, and carrying out contrast enhancement treatment on the original images, the contrast enhancement method comprises horizontal/vertical turning and random rotation operation, 2045 original images and 8180 expansion images are finally obtained and used for establishing the tunnel surface crack image database, 80% of sample images in the database are randomly selected as a training set, and 20% of sample images are selected as a testing set.
S2, constructing a deep learning model:
carrying out image preprocessing on an original tunnel image, and constructing a deep learning model, wherein the image preprocessing process comprises the following steps: a histogram equalization method is adopted to improve the image quality of the data set; the process of constructing the deep learning model comprises the following steps: and constructing a depth convolution encoder-decoder model for image segmentation based on deep learning, and performing iterative training and weight updating on the encoder-decoder model by using a tunnel crack image database to generate a deep learning model.
The encoder-decoder model of the invention comprises an encoder for feature extraction and a decoding network for feature fusion and image down-sampling, as shown in fig. 2, the specific method for performing iterative training and weight updating on the encoder-decoder model by using a tunnel crack image database comprises the following steps: the encoder part uses conv1-conv5 convolutional network (including maximum pooling layer therein) of ResNet101, and the feature maps generated by the last convolutional layer of conv3, conv4 and conv5 are respectively marked as feature1, feature2 and feature 3; the decoder first downsamples feature1 to the same size as the image resolution of feature2, then performs feature fusion on the downsampled result graph and feature2, then downsamples the Fused result graph to match the resolution of feature3, performs feature fusion on the downsampled result graph and feature3 to obtain a Fused feature map, performs dimensionality reduction processing with a 1 × 1 convolution kernel, and uses data-dependent upsampling to obtain a final result graph.
The deep learning model of the encoder-decoder comprises a loss function training model, and the calculation formula of the loss function training model is as follows:
L(F,Y)=||F-Y′||2
wherein, F represents the generated feature map, Y represents the corresponding ground-route, Y 'represents the image obtained by reducing the resolution of Y, and Y' has the same resolution as F. Wherein the process of generating Y' by reducing the resolution of Y comprises the following steps:
(1) let Y be an element of RH×W×CDividing Y into H/r multiplied by W/r sub-windows, wherein the size of each sub-window S is r multiplied by C;
(2) for each window S, the value range of each pixel point is [0,1 ]]It is converted into a vector v ∈ {0, 1}N,N=r*r*C;
(3) Compressing the vector v into a low-dimensional vector x by using linear projection, and combining the compressed x corresponding to all H/r multiplied by W/r data blocks into Y', wherein the linear projection transformation formula is as follows:
x=Pv;v′=Wx
wherein P ∈ RC′×NFor compressing v to x, W ∈ RN×C′For the reconstruction matrix, x is reconstructed as v, and v' represents the reconstructed vector v.
The embodiment of the invention can minimize reconstruction errors on a training set, and learn to obtain a transformation matrix P and a reconstruction matrix W, wherein the minimization formula is as follows:
wherein, P*Representing the transformation matrix, W, obtained after minimizing the reconstruction errors*Representing the reconstruction matrix obtained after minimizing the reconstruction error.
To calculate the above-mentioned loss function L (F, Y) | | | F-Y' | non-calculation using a more intuitive method2Using the above formula to learn to obtain a reconstruction matrix W to upsample F, that is, using data-dependent upsampling operation to replace a coarser bilinear interpolation upsampling algorithm, the specific process of upsampling is shown in fig. 3, and a loss function training model L (F, Y) | | F-Y' | survival rate2The conversion is to more intuitive L (F, Y) ═ Loss (dupsmax (F)), Y.
S3, identifying and segmenting the fine cracks of the subway tunnel:
the process of identifying and segmenting the fine cracks comprises the following steps: and inputting the image to be detected into the deep learning model trained in the step S2, and outputting the segmentation result of the fine cracks of the tunnel surface image.
The results of the embodiment of the present invention are shown in fig. 4, where graph (a) is the input image; FIG. (b) is the manually annotated group route corresponding to FIG. (a); the graph (c) is a division result output by the deep learning model input to the graph (a). The statistical table in fig. 5 is an average intersection ratio (mlou) of the division results of the fine cracks of the subway tunnel in the present embodiment. As can be seen from the table data in fig. 5 and the segmentation result in fig. 4, the deep learning model based on the encoder-decoder architecture constructed in the present embodiment has a higher segmentation accuracy for the micro cracks on the surface of the subway tunnel.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above-described embodiments. It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to make many modifications without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A subway tunnel micro crack segmentation method based on deep learning is characterized in that: the method comprises the following steps:
s1, acquiring an image and establishing a database: automatically acquiring images of the surface of the subway tunnel by using a mobile trolley, manually selecting and marking the images containing cracks, and establishing a tunnel surface crack image database;
s2, establishing a learning model: carrying out image preprocessing on an original tunnel image to construct a deep learning model;
and S3, identifying and segmenting the fine cracks of the subway tunnel.
2. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 1, wherein: the step S1 specifically includes the following steps:
(1) moving a moving trolley carrying a plurality of cameras in the tunnel to automatically acquire the surface image of the subway tunnel;
(2) carrying out image splicing operation on an original image acquired by a camera to obtain a real tunnel section panoramic image, and cutting the panoramic image to obtain a picture with a fixed size;
(3) and expanding the subway tunnel surface image containing the cracks, carrying out manual annotation, and establishing a tunnel surface crack image database.
3. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 2, wherein: the image stitching operation is performed on the original image in the step (2), and the specific steps of obtaining a real tunnel section panoramic image are as follows:
A. feature extraction: extracting local features of the image by using a Scale Invariant Feature Transform (SIFT) algorithm;
B. and (3) feature matching: matching the characteristic points of the image to be spliced by adopting a mode of comparing the nearest neighbor distance with the next nearest neighbor distance, and eliminating mismatching characteristic point pairs by utilizing a random sampling consistency RANSAC algorithm;
C. image fusion: and calculating a mathematical conversion model according to the matched characteristic point pairs, converting the images to be spliced into a coordinate system of a reference image, completing the same coordinate transformation, and fusing the overlapped areas of the images to be spliced by adopting a weighted average image fusion method to obtain a reconstructed smooth seamless panoramic image.
4. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 1, wherein: the process of preprocessing the image in step S2 is as follows: performing contrast enhancement processing on the original image by adopting a histogram equalization method, and improving the image quality of the data set; the process of constructing the deep learning model comprises the following steps: and constructing a depth convolution encoder-decoder model for image segmentation based on deep learning, and performing iterative training and weight updating on the encoder-decoder model by using a tunnel crack image database to generate a deep learning model.
5. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 4, wherein: the encoder-decoder model includes an encoder for feature extraction and a decoding network for feature fusion and image down-sampling.
6. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 4, wherein: the specific method for performing iterative training and weight updating on the encoder-decoder model by using the tunnel crack image database comprises the following steps: the encoder part uses conv1-conv5 convolutional network (including maximum pooling layer therein) of ResNet101, and feature maps generated by the last convolutional layer of conv3, conv4 and conv5 are respectively marked as feature1, feature2 and feature 3; the decoder first downsamples feature1 to the same size as the image resolution of feature2, then performs feature fusion on the downsampled result graph and feature2, then downsamples the Fused result graph to match the resolution of feature3, performs feature fusion on the downsampled result graph and feature3 to obtain a Fused feature map, performs dimensionality reduction processing with a 1 × 1 convolution kernel, and uses data-dependent upsampling to obtain a final result graph.
7. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 6, wherein: the deep learning model of the encoder-decoder comprises a loss function training model, and the calculation formula of the loss function training model is as follows:
L(F,Y)=||F-Y′||2
wherein, F represents the generated feature map, Y represents the corresponding ground-route, Y 'represents the image obtained by reducing the resolution of Y, and Y' has the same resolution as F.
8. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 7, wherein: the process of Y deresolved generation Y' includes the steps of:
(1) let Y be an element of RH×W×CDividing Y into H/r multiplied by W/r sub-windows, wherein the size of each sub-window S is r multiplied by C;
(2) for each window S, the value range of each pixel point is [0,1 ]]It is converted into a vector v ∈ {0, 1}N,N=r*r*C;
(3) The vector v is compressed into a low-dimensional vector x by using linear projection, and the compressed x corresponding to all H/r multiplied by W/r data blocks is combined into Y'.
9. The subway tunnel micro-crack segmentation method based on deep learning of claim 8, wherein: the linear projection transformation formula in the step 3 is as follows:
x=Pv:v′=Wx
wherein P ∈ RC′×NFor compressing v to x, W ∈ RN×C′For the reconstruction matrix, x is reconstructed as v, and v' represents the reconstructed vector v.
10. The subway tunnel micro-crack segmentation method based on deep learning as claimed in claim 1, wherein: the process of crack identification and segmentation in step S3 is as follows: and inputting the image to be detected into the deep learning model trained in the step S2, and outputting the segmentation result of the fine cracks of the tunnel surface image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010319469.7A CN111932548B (en) | 2020-04-21 | 2020-04-21 | Subway tunnel fine crack segmentation method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010319469.7A CN111932548B (en) | 2020-04-21 | 2020-04-21 | Subway tunnel fine crack segmentation method based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111932548A true CN111932548A (en) | 2020-11-13 |
CN111932548B CN111932548B (en) | 2022-04-22 |
Family
ID=73316311
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010319469.7A Active CN111932548B (en) | 2020-04-21 | 2020-04-21 | Subway tunnel fine crack segmentation method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111932548B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112632680A (en) * | 2020-12-30 | 2021-04-09 | 同济大学 | Large civil engineering structure water leakage condition reconstruction method based on deep learning |
CN113205133A (en) * | 2021-04-30 | 2021-08-03 | 成都国铁电气设备有限公司 | Tunnel water stain intelligent identification method based on multitask learning |
CN114419421A (en) * | 2022-01-21 | 2022-04-29 | 中国地质大学(北京) | Subway tunnel crack identification system and method based on images |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180174275A1 (en) * | 2016-12-15 | 2018-06-21 | WaveOne Inc. | Autoencoding image residuals for improving upsampled images |
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 |
CN110020652A (en) * | 2019-01-07 | 2019-07-16 | 新而锐电子科技(上海)有限公司 | The dividing method of Tunnel Lining Cracks image |
CN111028217A (en) * | 2019-12-10 | 2020-04-17 | 南京航空航天大学 | Image crack segmentation method based on full convolution neural network |
-
2020
- 2020-04-21 CN CN202010319469.7A patent/CN111932548B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180174275A1 (en) * | 2016-12-15 | 2018-06-21 | WaveOne Inc. | Autoencoding image residuals for improving upsampled images |
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 |
CN110020652A (en) * | 2019-01-07 | 2019-07-16 | 新而锐电子科技(上海)有限公司 | The dividing method of Tunnel Lining Cracks image |
CN111028217A (en) * | 2019-12-10 | 2020-04-17 | 南京航空航天大学 | Image crack segmentation method based on full convolution neural network |
Non-Patent Citations (1)
Title |
---|
QIAN XIE 等: "Automatic Detection and Classification of Sewer", 《IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112632680A (en) * | 2020-12-30 | 2021-04-09 | 同济大学 | Large civil engineering structure water leakage condition reconstruction method based on deep learning |
CN113205133A (en) * | 2021-04-30 | 2021-08-03 | 成都国铁电气设备有限公司 | Tunnel water stain intelligent identification method based on multitask learning |
CN113205133B (en) * | 2021-04-30 | 2024-01-26 | 成都国铁电气设备有限公司 | Tunnel water stain intelligent identification method based on multitask learning |
CN114419421A (en) * | 2022-01-21 | 2022-04-29 | 中国地质大学(北京) | Subway tunnel crack identification system and method based on images |
Also Published As
Publication number | Publication date |
---|---|
CN111932548B (en) | 2022-04-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111932548B (en) | Subway tunnel fine crack segmentation method based on deep learning | |
CN111401384B (en) | Transformer equipment defect image matching method | |
CN112734646B (en) | Image super-resolution reconstruction method based on feature channel division | |
WO2022160170A1 (en) | Method and apparatus for detecting metal surface defects | |
CN110070091B (en) | Semantic segmentation method and system based on dynamic interpolation reconstruction and used for street view understanding | |
CN111369565A (en) | Digital pathological image segmentation and classification method based on graph convolution network | |
CN112101138B (en) | Bridge inhaul cable surface defect real-time identification system and method based on deep learning | |
US20220366682A1 (en) | Computer-implemented arrangements for processing image having article of interest | |
CN114092439A (en) | Multi-organ instance segmentation method and system | |
CN111881743B (en) | Facial feature point positioning method based on semantic segmentation | |
CN112101262B (en) | Multi-feature fusion sign language recognition method and network model | |
CN112580515A (en) | Lightweight face key point detection method based on Gaussian heat map regression | |
CN112906675B (en) | Method and system for detecting non-supervision human body key points in fixed scene | |
CN110895815A (en) | Chest X-ray pneumothorax segmentation method based on deep learning | |
CN115937626A (en) | Automatic generation method of semi-virtual data set based on instance segmentation | |
CN116797609A (en) | Global-local feature association fusion lung CT image segmentation method | |
Zeng et al. | Self-attention learning network for face super-resolution | |
CN116468732A (en) | Lung CT image segmentation method and imaging method based on deep learning | |
CN110807416A (en) | Digital instrument intelligent recognition device and method suitable for mobile detection device | |
CN117541574A (en) | Tongue diagnosis detection method based on AI semantic segmentation and image recognition | |
CN112200766A (en) | Industrial product surface defect detection method based on area-associated neural network | |
CN117291902A (en) | Detection method for pixel-level concrete cracks based on deep learning | |
CN116362995A (en) | Tooth image restoration method and system based on standard prior | |
CN113947567B (en) | Defect detection method based on multitask learning | |
CN116188352A (en) | Pulmonary nodule segmentation method based on enhanced edge features |
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 |