CN110197483A - Deep basal pit crack detection method based on vision signal - Google Patents
Deep basal pit crack detection method based on vision signal Download PDFInfo
- Publication number
- CN110197483A CN110197483A CN201910483858.0A CN201910483858A CN110197483A CN 110197483 A CN110197483 A CN 110197483A CN 201910483858 A CN201910483858 A CN 201910483858A CN 110197483 A CN110197483 A CN 110197483A
- Authority
- CN
- China
- Prior art keywords
- crack
- deep basal
- basal pit
- image
- vision signal
- 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
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
-
- 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
- G06T2207/20081—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]
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to field of video monitoring, disclose a kind of deep basal pit crack detection method based on vision signal, including obtaining deep basal pit crack sample image, processing sample simultaneously obtains crack training set and verifying collection, based on training set training deep basal pit Crack Detection model, and using crack verifying collection Optimized model parameter, automatic detection is carried out to the image of acquisition based on deep basal pit Crack Detection model.The present invention can be to the crack progress automatic detection of deep basal pit in vision signal, with good real-time and higher accuracy, effectively the crack in deep basal pit can be detected, and robustness with higher, can adapt to the variation of deep basal pit background and context.
Description
Technical field
The present invention relates to deep basal pit field of video monitoring, and in particular to a kind of deep basal pit Crack Detection side of vision signal
Method.
Background technique
Deep basal pit refer to cutting depth be more than 5 meters (containing 5 meters) though or depth be less than 5 meters, geological conditions and surrounding ring
Border and the especially complex engineering of underground utilities are a kind of engineering projects common in current large-scale construction project.With foundation pit depth
The increase of degree, safety need to obtain absolute guarantee.Crack is a great reason for inducing deep basal pit accident, seriously
Threaten the safety of foundation pit.The characteristics of in view of deep basal pit, it is time-consuming and laborious to check foundation pit crack using the mode of artificial detection, with
The fast development of computer vision technique, based on vision signal to deep basal pit crack carry out detection become a kind of possibility.
Summary of the invention
The purpose of the present invention is to provide a kind of deep basal pit crack detection method based on vision signal.Solve deep base at present
Detect whether to there are problems that bring the detection method in the crack of security risk to lack in hole.
In order to solve the above technical problems, the invention adopts the following technical scheme: a kind of deep basal pit based on vision signal
Crack detection method, its step are as follows:
1) using image pattern of the video camera acquisition comprising deep basal pit crack information;
2) processing is carried out to image pattern and obtains training set and verifying collection, carry out the region Faster R-CNN convolutional Neural net
Network training, comprising the following steps:
Sample size is adjusted to uniform sizes and is labeled to the crack information in sample by 2.1, i.e., crack (x, y,
W, h) coordinate information and classification information, 80% is used as training set in sample set, and 20% is used as verifying collection;
Training set is inputted the region Faster R-CNN convolutional neural networks by 2.2, by sharing convolutional layer and spy in network
There is convolutional layer to handle to obtain sample characteristics Value Data, region suggestion and area score obtained by the convolutional layer in RPN network,
The pond the ROL layer that region is suggested and characteristic value information inputs in network is obtained into the feature suggested in region, finally via full connection
Layer processing obtains the coordinate and classification in crack in image;
It is verified using verifying the set pair analysis model, optimizes network;
Acquisition deep basal pit image to be detected in real time, is input to deep basal pit Crack Detection model inspection crack information.
As the supplement of the above-mentioned deep basal pit crack detection method based on vision signal, in the step 1, deep basal pit is split
Seam sample image refer to acquired in deep basal pit environment include deep basal pit crack image;It is selected at random in these images
With in different location, with the crack image under different patterns and illumination condition, and be divided into training set and verifying
Collection, for training sample set for training deep basal pit Crack Detection model, verifying sample set is used for the parameter of Optimized model.
As the supplement of the deep basal pit crack detection method based on vision signal, in the step 2, training sample is handled,
It is trained, it is characterised in that in the step 2.1, the image size of use is adjusted to 640*480, and in image
Crack is labeled, and provides the boundary coordinate in crack.
As the supplement of the deep basal pit crack detection method based on vision signal, area score is used in the step 2.2
Non-maxima suppression, threshold value 0.75.
As the supplement of the deep basal pit crack detection method based on vision signal, in the step 2.2, characteristic model damage
It is as follows to lose function:
Wherein, piThe probability in crack is predicted to be for current region;pi *It is 1, when negative sample label in positive sample label
It is 0; ti={ tx, ty, tw, th};Ncls=128;Lcls(pi, pi *)=- log [pi *pi+(1-pi *)(1-pi)];Nreg=2400;
Lreg(ti, ti *)=R (ti-ti *)。
As the supplement of the above-mentioned deep basal pit crack detection method based on vision signal, in the step 3, sample is used
Verifying collection, by the continuous repetitive exercise Optimized model relevant parameter of method in step 2, until model training error tends towards stability, most
Deep basal pit Crack Detection model is obtained eventually.
As the supplement of the deep basal pit crack detection method based on vision signal, in the step 4, the deep base that acquires in real time
Live image is cheated, and by method in step 2 by the input of image to be detected deep basal pit Crack Detection model, output is split after detection
Tape edge frame coordinate and classification.
The utility model has the advantages that
Effect 1: the method for the present invention proposes one for the insufficient status of fracture automatic detection in current deep basal pit project
Deep basal pit crack detection method of the kind based on vision signal.It can be used for carrying out automatic detection to the crack of deep basal pit, it is ensured that deep
Deep pit monitor.
Effect 2: the deep basal pit crack detection method based on vision signal that the present invention uses is compared to traditional artificial investigation
Mode, accuracy high-efficient and with higher.
Detailed description of the invention
Fig. 1 is the deep basal pit crack detection method process of the invention based on vision signal.
Fig. 2 is Faster R-CNN network structure.
Specific embodiment
Below in conjunction with the attached drawing in specification, technical solution in the embodiment of the present invention is clearly and completely retouched
It states, it is clear that described example is only a part of example of the present invention, instead of all the embodiments.Present example is base
Deep basal pit crack is detected in vision signal, is solved the problems, such as at present to deep basal pit Crack Detection inefficiency.
Its step are as follows:
Step 1: including the image pattern of deep basal pit crack information using video camera acquisition;
Step 2: processing being carried out to image pattern and obtains training set and verifying collection, carries out the region Faster R-CNN convolution mind
Through network training, comprising the following steps:
Sample size is adjusted to uniform sizes and is labeled to the crack information in sample by 2.1, i.e., crack (x, y,
W, h) coordinate information and classification information, 80% is used as training set in sample set, and 20% is used as verifying collection.
Training set is inputted the region Faster R-CNN convolutional neural networks by 2.2, by sharing convolutional layer and spy in network
There is convolutional layer to handle to obtain sample characteristics Value Data, region suggestion and area score obtained by the convolutional layer in RPN network,
The pond the ROL layer that region is suggested and characteristic value information inputs in network is obtained into the feature suggested in region, finally via full connection
Layer processing obtains the coordinate and classification in crack in image.
Step 3: being verified using verifying the set pair analysis model, optimize network.
Step 4: acquisition deep basal pit image to be detected in real time is input to deep basal pit Crack Detection model inspection crack information.
In the step 1, deep basal pit crack sample image refers to that acquiring in deep basal pit environment includes deep basal pit crack
Image.Selected at random in these images in different location, with the crack pattern under different patterns and illumination condition
Picture, and it is divided into training set and verifying collection, training sample set verifies sample for training deep basal pit Crack Detection model
Collection is used for the parameter of Optimized model.
In the step 2, the image size of use is adjusted to 640*480, and be labeled to the crack in image, given
The boundary coordinate in crack out;Area score uses non-maxima suppression, threshold value 0.75.Characteristic model loss function is as follows:
Wherein, piThe probability in crack is predicted to be for current region;pi *It is 1, when negative sample label in positive sample label
It is 0;
ti={ tx, ty, tw, th};Ncls=128;Lcls(pi, pi *)=- log [pi *pi+(1-pi *)(1-pi)];Nreg=
2400; Lreg(ti, ti *)=R (ti-ti *)。
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in the use of the new type
Spirit and principle within, any modification, equivalent replacement, improvement and so on, should be included in invention protection scope it
It is interior.
Claims (7)
1. the deep basal pit crack detection method based on vision signal, which is characterized in that method includes the following steps:
1) using image pattern of the video camera acquisition comprising deep basal pit crack information;
2) processing is carried out to image pattern and obtains training set and verifying collection, carry out Faster R-CNN region convolutional neural networks instruction
Practice, comprising the following steps:
Sample size is adjusted to uniform sizes and is labeled to the crack information in sample by 2.1, i.e. (x, y, w, the h) in crack
Coordinate information and classification information, 80% is used as training set in sample set, and 20% is used as verifying collection;
Training set is inputted the region Faster R-CNN convolutional neural networks by 2.2, by sharing convolutional layer and peculiar volume in network
Lamination handles to obtain sample characteristics Value Data, region suggestion and area score is obtained by the convolutional layer in RPN network, by region
It is recommended that obtaining the feature suggested in region with the pond the ROL layer in characteristic value information input network, finally handled via full articulamentum
Obtain the coordinate and classification in crack in image;
3) it is verified using verifying the set pair analysis model, optimizes network;
Acquisition deep basal pit image to be detected in real time, is input to deep basal pit Crack Detection model inspection crack information.
2. the deep basal pit crack detection method according to claim 1 based on vision signal, which is characterized in that the step
In 1, deep basal pit crack sample image refer to acquired in deep basal pit environment include deep basal pit crack image;In these images
In select at random in different location, with the crack image under different patterns and illumination condition, and be divided into training
Collection and verifying collection, for training sample set for training deep basal pit Crack Detection model, verifying sample set is used for the parameter of Optimized model.
3. the deep basal pit crack detection method according to claim 1 based on vision signal, which is characterized in that the step
In 2.1, the image size of use is adjusted to 640*480, and be labeled to the crack in image, the boundary for providing crack is sat
Mark.
4. the deep basal pit crack detection method shown according to claim 1 based on vision signal, which is characterized in that the step
Area score uses non-maxima suppression, threshold value 0.75 in 2.2.
5. the deep basal pit crack detection method according to claim 1 based on vision signal, which is characterized in that the step
In 2.2, characteristic model loss function is as follows:
Wherein, piThe probability in crack is predicted to be for current region;pi *It is 1 in positive sample label, is 0 when negative sample label;
ti={ tx, ty, tw, th};Ncls=128;Lcls(pi, pi *)=- log [pi *pi+(1-pi *)(1-pi)];Nreg=2400;
Lreg(ti, ti *)=R (ti-ti *)。
6. the deep basal pit crack detection method according to claim 1 based on vision signal, which is characterized in that the step
It in 3, is verified and is collected using sample, by the continuous repetitive exercise Optimized model relevant parameter of method in step 2, until model training error
It tends towards stability, it is final to obtain deep basal pit Crack Detection model.
7. the deep basal pit crack detection method according to claim 1 based on vision signal, which is characterized in that the step
In 4, the deep basal pit live image acquired in real time, and by method in step 2 by image to be detected deep basal pit Crack Detection model
Input exports crack frame coordinate and classification after detection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910483858.0A CN110197483A (en) | 2019-05-31 | 2019-05-31 | Deep basal pit crack detection method based on vision signal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910483858.0A CN110197483A (en) | 2019-05-31 | 2019-05-31 | Deep basal pit crack detection method based on vision signal |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110197483A true CN110197483A (en) | 2019-09-03 |
Family
ID=67753985
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910483858.0A Pending CN110197483A (en) | 2019-05-31 | 2019-05-31 | Deep basal pit crack detection method based on vision signal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110197483A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111047585A (en) * | 2019-12-25 | 2020-04-21 | 苏州奥易克斯汽车电子有限公司 | Pavement crack detection method |
CN111428592A (en) * | 2020-03-12 | 2020-07-17 | 厦门路桥信息股份有限公司 | Bored concrete pile identification method and system based on deep learning |
CN114034260A (en) * | 2021-09-18 | 2022-02-11 | 南京市江北新区中央商务区建设管理办公室 | Deep foundation pit support structure deformation diagnosis system based on streaming media and BIM |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN108416307A (en) * | 2018-03-13 | 2018-08-17 | 北京理工大学 | A kind of Aerial Images road surface crack detection method, device and equipment |
CN109285139A (en) * | 2018-07-23 | 2019-01-29 | 同济大学 | A kind of x-ray imaging weld inspection method based on deep learning |
CN109767423A (en) * | 2018-12-11 | 2019-05-17 | 西南交通大学 | A kind of crack detection method of bituminous pavement image |
-
2019
- 2019-05-31 CN CN201910483858.0A patent/CN110197483A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN108416307A (en) * | 2018-03-13 | 2018-08-17 | 北京理工大学 | A kind of Aerial Images road surface crack detection method, device and equipment |
CN109285139A (en) * | 2018-07-23 | 2019-01-29 | 同济大学 | A kind of x-ray imaging weld inspection method based on deep learning |
CN109767423A (en) * | 2018-12-11 | 2019-05-17 | 西南交通大学 | A kind of crack detection method of bituminous pavement image |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111047585A (en) * | 2019-12-25 | 2020-04-21 | 苏州奥易克斯汽车电子有限公司 | Pavement crack detection method |
CN111428592A (en) * | 2020-03-12 | 2020-07-17 | 厦门路桥信息股份有限公司 | Bored concrete pile identification method and system based on deep learning |
CN114034260A (en) * | 2021-09-18 | 2022-02-11 | 南京市江北新区中央商务区建设管理办公室 | Deep foundation pit support structure deformation diagnosis system based on streaming media and BIM |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107767376B (en) | X-ray bone age prediction method and system based on deep learning | |
CN108960135B (en) | Dense ship target accurate detection method based on high-resolution remote sensing image | |
CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
CN104483326B (en) | High-voltage line defects of insulator detection method and system based on depth belief network | |
CN104361314B (en) | Based on infrared and transformer localization method and device of visual image fusion | |
CN110197483A (en) | Deep basal pit crack detection method based on vision signal | |
CN109859171A (en) | A kind of flooring defect automatic testing method based on computer vision and deep learning | |
CN107451999A (en) | foreign matter detecting method and device based on image recognition | |
CN111091544B (en) | Method for detecting breakage fault of side integrated framework of railway wagon bogie | |
CN104867144B (en) | IC element welding point defect detection methods based on mixed Gauss model | |
CN104657706B (en) | The fracture of high ferro circuit line bar and draw bail body method for detecting abnormality based on image | |
Wang et al. | Investigation into recognition algorithm of helmet violation based on YOLOv5-CBAM-DCN | |
CN107782733A (en) | Image recognition the cannot-harm-detection device and method of cracks of metal surface | |
CN111462058B (en) | Method for rapidly detecting effective rice ears | |
CN111852792B (en) | Fan blade defect self-diagnosis positioning method based on machine vision | |
CN104850832B (en) | A kind of large-scale image sample mask method and system based on classification iteration | |
CN108416774A (en) | A kind of fabric types recognition methods based on fine granularity neural network | |
CN108009592A (en) | A kind of diabetic retinal classification of images method | |
CN105447859A (en) | Field wheat aphid counting method | |
CN104867145B (en) | IC element welding point defect detection methods based on VIBE models | |
CN110020691B (en) | Liquid crystal screen defect detection method based on convolutional neural network impedance type training | |
CN109087305A (en) | A kind of crack image partition method based on depth convolutional neural networks | |
CN106651893A (en) | Edge detection-based wall body crack identification method | |
CN106548131A (en) | A kind of workmen's safety helmet real-time detection method based on pedestrian detection | |
CN108109133A (en) | A kind of silkworm seed automatic counting method based on digital image processing techniques |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190903 |