CN108615057A - A kind of abnormality recognition method of the cable tunnel lighting apparatus based on CNN - Google Patents
A kind of abnormality recognition method of the cable tunnel lighting apparatus based on CNN Download PDFInfo
- Publication number
- CN108615057A CN108615057A CN201810404956.6A CN201810404956A CN108615057A CN 108615057 A CN108615057 A CN 108615057A CN 201810404956 A CN201810404956 A CN 201810404956A CN 108615057 A CN108615057 A CN 108615057A
- Authority
- CN
- China
- Prior art keywords
- lighting apparatus
- cnn
- training
- layer
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- 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
Abstract
The present invention relates to the technical fields of Computer Image Processing, more specifically, it is related to a kind of abnormality recognition method of the cable tunnel lighting apparatus based on CNN, the ability for having selected the mode of trained CNN convolutional neural networks model to take full advantage of convolutional neural networks extraction two-dimension picture feature compensates for conventional method feature description deficiency or is difficult to select the defect of suitable characteristics.The present invention lighting apparatus situation that can be in detection image is simultaneously with good stability, other noises, light are not influenced in by testing image, and open and close two kinds of objects that accurately can be detected and orient during crusing robot to be detected shoots lighting apparatus picture in lighting apparatus, with stronger anti-interference ability and preferable robustness, the accuracy rate of cable tunnel interior lighting equipment detection can be improved.There is versatility in tunnel internal equipment detection dim, under complex background, there is the wider scope of application.
Description
Technical field
The present invention relates to the technical fields of Computer Image Processing, more particularly, to a kind of cable tunnel based on CNN
The abnormality recognition method of lighting apparatus.
Background technology
Power cable is in closed tunnel environment, internal crowded, dim, to the inspection quality of cable tunnel, inspection
Speed usually determines by the light luminance of tunnel internal, thus cable tunnel inside is provided unique light source lighting apparatus it is different
Often detection becomes the guarantee that cable tunnel inspection is smoothed out.Currently, detecting lighting apparatus frequently with the mode of manual inspection
Exception, but since cable laying length is longer, internal environment is crowded, the inefficiency of manual inspection, is not easy to different to occurring
Normal lighting apparatus makes quick, correct processing.
Invention content
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of cable tunnel lighting apparatus based on CNN
Abnormality recognition method, image processing techniques can be utilized to complete to the on-line monitoring of lighting apparatus, convenient for operation maintenance personnel to going out
Now abnormal lighting apparatus makes quick, correct processing, realizes intelligent, the rapid and accuracy of cable tunnel inspection.
In order to solve the above technical problems, the technical solution adopted by the present invention is：
A kind of abnormality recognition method of the cable tunnel lighting apparatus based on CNN is provided, is included the following steps：
S1. by tunnel crusing robot camera shooting, collecting lighting apparatus sample image, image set is obtained；
S2. sample image all in image set described in traversal step S1 surrounds every sample image with encirclement frame
Complete lighting apparatus simultaneously marks processing for the lighting apparatus sample image with encirclement frame, and the pixel in encirclement frame is marked
For lighting apparatus classification, the pixel outside encirclement frame is labeled as background classification, obtains the first training set；
S3. it is directed to the lighting apparatus sample image with encirclement frame described in step S2 and carries out scaling processing, it will
The longer sides of the encirclement frame of every sample image transform to goal-selling size, and shorter edge transforms to goal-selling according to longer sides
The zoom scale of size carries out the scaling of same ratio, obtains the second training set；
S4. using COCO data sets input CNN models to carry out pre-training, simultaneously iteration pre-training obtains model after pre-training, will
Model carries out being directed to training after the second training set input pre-training described in step S3 and iteration obtains tunnel photograph for training
Bright equipment detection model；
S5. the testing image for acquiring tunnel illumination equipment in real time, zooms in and out according to the zoom scale described in step S3
It is input to afterwards in the tunnel illumination equipment detection model described in step S4, calculates the classification results that output confidence level is more than 90%
Recognition result as testing image.
The abnormality recognition method of the cable tunnel lighting apparatus based on CNN of the present invention, has selected trained CNN convolutional Neurals
The mode of network model takes full advantage of the ability of convolutional neural networks extraction two-dimension picture feature, compensates for conventional method feature
Description is insufficient or is difficult to select the defect of suitable characteristics.The present invention can be in detection image lighting apparatus situation and with good
Stability, other noises, light are not influenced in by testing image, and in the case that lighting apparatus open and close two kinds all
The object in crusing robot shooting lighting apparatus picture to be detected accurately can be detected and be oriented, is had stronger anti-
Interference performance and preferable robustness can improve the accuracy rate of cable tunnel interior lighting equipment detection.In dim, complicated
Tunnel internal equipment detection under background has versatility, has the wider scope of application.
Preferably, lighting apparatus described in step S1 is the image of the emergent pilot light equipment of tunnel internal, described to answer
Anxious pilot light equipment includes square shell and two groups of lighting pilot lamps being connected on square shell.Lighting pilot lamp point
For " bright " and " going out " two class state, " bright " state indicates that illumination exception does not occur in the lighting pilot lamp, and " going out " state then indicates
It is abnormal that illumination occurs in the lighting pilot lamp.
Preferably, for camera using lighting apparatus as object, coverage is that horizontal left avertence difference is 15 °~30 ° visual angles, water
Flat right avertence difference is range between 15 °~30 ° visual angles and overlooks that deviation is 50 °~70 ° visual angles, to look up deviation be 50 °~70 °
Range between visual angle.Lighting apparatus sample image can be acquired with different view, obtained various sample image, be can be improved
The accuracy of anomalous identification.
Preferably, the encirclement frame described in step S2 is rectangle frame, and complete lighting apparatus is non-targeted in encirclement frame
The region area of object accounts for lighting apparatus of the ratio less than 15% of encirclement frame area.
Preferably, the CNN models described in step S4 are by input layer, convolutional layer, pond layer, full articulamentum and output layer structure
It builds.
Preferably, the establishment step of model is as follows after pre-training in step S4:
S41. using lighting apparatus sample image as input layer, sample image convolution extraction characteristic value is checked in convolutional layer；
S42. the characteristic value output in convolutional layer is input in the layer of pond as input and carries out maximum pond operation, contracting
The information of small convolutional layer；
S43. through the maximum pond operation in the multiple step S42 of convolution sum described in multilayer step S41, by pond layer
Output carries out each characteristic value using different weights the operation of full articulamentum as input, and the two-dimensional signal of image is turned
It is changed to one-dimension information；
S44. classified to sample image according to the value of sample image one-dimension information, classification results are exported by output layer.
Input picture is described by the feature extraction frame of CNN and detects the relationship between target, reduces the mistake of target
Error detection, the step of avoiding non-maxima suppression, solve detection and positioning of the lighting apparatus under normally-open and abnormal closing
Problem.
Preferably, the neural network structure of the convolutional layer, pond layer stacked using four layers.
Preferably, weight described in step S43 is calculated as follows：
In formula, loss function is mean square deviation MSE functions, WiIndicate i-th of weight of convolutional layer, biIndicate convolutional layer i-th partially
It setting, Y indicates that entire sample set, Y (i) indicate the corresponding mark value of i-th of sample,Indicate i-th of sample input training network
The output token value of output layer afterwards, η indicate the learning efficiency of back-propagation algorithm.
Preferably, the model training error of model is less than 10% after pre-training in step S4, tunnel illumination monitoring of equipment mould
The average value of the model training error of type is less than 5%.
Compared with prior art, the beneficial effects of the invention are as follows：
The present invention lighting apparatus situation that can be in detection image is simultaneously with good stability, not its in by testing image
The influence of his noise, light, and open and close two kinds in lighting apparatus and accurately can detect and orient and wait for
The object in crusing robot shooting lighting apparatus picture is detected, there is stronger anti-interference ability and preferable robustness,
The accuracy rate of cable tunnel interior lighting equipment detection can be improved.In tunnel internal equipment inspection dim, under complex background
Measuring tool has versatility, has the wider scope of application.
Specific implementation mode
The present invention is further illustrated With reference to embodiment.
Embodiment one
The present embodiment is a kind of first embodiment of the abnormality recognition method of the cable tunnel lighting apparatus based on CNN, packet
Include following steps：
S1. by tunnel crusing robot camera shooting, collecting lighting apparatus sample image, image set is obtained；
S2. sample image all in image set described in traversal step S1 surrounds every sample image with encirclement frame
Simultaneously processing is marked for the lighting apparatus sample image with encirclement frame, by the pixel in encirclement frame in complete lighting apparatus
Labeled as lighting apparatus classification, the pixel outside encirclement frame is labeled as background classification, obtains the first training set；
S3. it is directed to the lighting apparatus sample image with encirclement frame described in step S2 and carries out scaling processing, it will
The longer sides of the encirclement frame of every sample image transform to goal-selling size, and shorter edge transforms to goal-selling according to longer sides
The zoom scale of size carries out the scaling of same ratio, obtains the second training set；
S4. using COCO data sets input CNN models to carry out pre-training, simultaneously iteration pre-training obtains model after pre-training, will
Model carries out being directed to training after the second training set input pre-training described in step S3 and iteration obtains tunnel photograph for training
Bright equipment detection model；
S5. the testing image for acquiring tunnel illumination equipment in real time, zooms in and out according to the zoom scale described in step S3
It is input to afterwards in the tunnel illumination equipment detection model described in step S4, calculates the classification results that output confidence level is more than 90%
Recognition result as testing image.
Wherein, lighting apparatus described in step S1 is the image of the emergent pilot light equipment of tunnel internal, described emergent
Pilot light equipment includes square shell and two groups of lighting pilot lamps being connected on square shell；In step S1, camera shooting
Head using lighting apparatus as object, coverage be horizontal left avertence difference be 15 °~30 ° visual angles, horizontal right avertence difference is 15 °~30 °
Range between visual angle and overlook that deviation is 50 °~70 ° visual angles, to look up deviation be the range between 50 °~70 ° visual angles.This
The lighting apparatus sample image of embodiment can be acquired with different view, obtain various sample image, improve anomalous identification
Accuracy.In first training set, lighting apparatus sample image is marked to form image tagged to each pixel therein
Collection, image tagged collection store flag data using following form：
{ image_name, label, x1, y1 }
Wherein, image_name indicates that crusing robot shoots lighting apparatus Image Name, and label indicates the class of pixel
Not, x1 indicates that the abscissa of pixel, y1 indicate the ordinate of pixel.
Encirclement frame described in step S2 is rectangle frame, and complete lighting apparatus is the region of non-targeted object in encirclement frame
Area accounts for lighting apparatus of the ratio less than 15% of encirclement frame area.
CNN models described in step S4 are pre- to instruct by input layer, convolutional layer, pond layer, full articulamentum and output layer building
The model training error of model is less than 10% after white silk, and the average value of the model training error of tunnel illumination equipment monitoring model is less than
5%, the establishment step of model is as follows after pre-training:
S41. using lighting apparatus sample image as input layer, sample image convolution extraction characteristic value is checked in convolutional layer；
S42. the characteristic value output in convolutional layer is input in the layer of pond as input and carries out maximum pond operation, contracting
The information of small convolutional layer；
S43. through the maximum pond operation in the multiple step S42 of convolution sum described in multilayer step S41, by pond layer
Output carries out each characteristic value using different weights the operation of full articulamentum as input, and the two-dimensional signal of image is turned
It is changed to one-dimension information；
S44. classified to sample image according to the value of sample image one-dimension information, classification results are exported by output layer.
Input picture is described by the feature extraction frame of CNN and detects the relationship between target, reduces the mistake of target
Error detection, the step of avoiding non-maxima suppression, solve detection and positioning of the lighting apparatus under normally-open and abnormal closing
Problem.
Each layer weight calculation is calculated using reversed gradient in step S43：
In formula, loss function is mean square deviation MSE functions, WiIndicate i-th of weight of convolutional layer, biIndicate convolutional layer i-th partially
It setting, Y indicates that entire sample set, Y (i) indicate the corresponding mark value of i-th of sample,Indicate i-th of sample input training network
The output token value of output layer afterwards, η indicate the learning efficiency of back-propagation algorithm.
The convolutional layer of the present embodiment, the stacking of pond layer use four layers of neural network structure：
First layer, it is 2 first to use convolutional layer, 48 convolution filters that convolutional layer is 7 × 7 using size, convolution stride,
Export the characteristic pattern of 48 arbitrary sizes；After the output for obtaining convolutional layer, carried out using BN layers (batch normalization)
Then normalized uses ReLU functions (Rectified Linear Units) to be swashed as nonlinear activation function
Living, the maximum pond layer (Maxpooling) for being again finally 3 × 3 with a window size carries out pond, maximum pond layer
(Maxpooling) sampling stride is 2；
The second layer, it is 2 first to use convolutional layer, 96 convolution filters that convolutional layer is 5 × 5 using size, convolution stride,
Export the characteristic pattern of 96 arbitrary sizes；After the output for obtaining convolutional layer, carried out using BN layers (batch normalization)
Then normalized uses ReLU functions (Rectified Linear Units) to be swashed as nonlinear activation function
Living, the maximum pond layer (Maxpooling) for being again finally 3 × 3 with a window size carries out pond, maximum pond layer
(Maxpooling) sampling stride is 2；
Third layer, it is 1 first to use convolutional layer, 96 convolution filters that convolutional layer is 3 × 3 using size, convolution stride,
And convolution offset distance pad is arranged makes the dimension of input figure be equal to output figure for 1, exports the characteristic pattern of 96 arbitrary sizes；
4th layer, it is 1 first to use convolutional layer, 48 convolution filters that convolutional layer is 3 × 3 using size, convolution stride,
And it is 1 that convolution offset distance pad, which is arranged, and ReLU functions (Rectified Linear Units) are reused after convolution and are used as activation primitive
Into line activating；
The structure of full articulamentum is as follows：Using the feature of two complete 256 dimensions of articulamentum processing output, then returned using frame
Layer (smoothL1Loss Layer) is returned to handle, frame returns layer and exports frame, obtains four elements of frame, four elements point
It is not wide w, the high h of transverse and longitudinal coordinate x, the y and frame recurrence the exported frame of layer in the upper left corner that frame returns the exported frame of layer；
Using frame as lighting apparatus possibility region as the real estate of target location.
Output layer is specifically using such as lower structure：It is handled using convolution kernel for the one-dimension information of full articulamentum output,
The characteristic pattern of output is fixed as 56 × 56, and then convolution kernel size that the characteristic pattern of output is inputted to one three layers is 1 × 1 convolution
There are 1024 dimensions to export for layer, wherein first layer convolutional layer, and there are the second layer 256 dimensions to export, and third layer has 4 dimensions
Degree output, the output of the convolutional layer of third layer are input to binary classifier.
Embodiment two
400 experiment pictures are collected, including the crusing robot of different shooting angles shoots tunnel internal lighting apparatus
Image, wherein the picture for training has 200, each 50 of the picture of each switch and LED status, remaining 200 conduct
Test set picture.The image of tunnel internal lighting apparatus is shot using CNN model inspection crusing robots, it is concluded that：Equipment
Exception and testing result are that abnormal correct coincidence factor is 100%, and equipment is normal and testing result is normal correct coincidence factor
It is 100%.
In summary, two kinds for opening, closing in lighting apparatus, the angle according to shooting image is different, this hair
Bright method also accurately can detect and orient the lighting apparatus in image and complete accurate result detection, this can allow and take the photograph
The placement position of camera and the fixed point inspection position of crusing robot are freer, can effectively cope with some complicated rings
Border situation.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
All any modification, equivalent and improvement etc., should be included in the claims in the present invention made by within the spirit and principle of invention
Protection domain within.
Claims (9)
1. a kind of abnormality recognition method of the cable tunnel lighting apparatus based on CNN, which is characterized in that include the following steps：
S1. by tunnel crusing robot camera shooting, collecting lighting apparatus sample image, image set is obtained；
S2. sample image all in image set described in traversal step S1 surrounds completely every sample image encirclement frame
Lighting apparatus and be marked processing be the lighting apparatus sample image with encirclement frame, by encirclement frame pixel mark
For lighting apparatus classification, the pixel outside encirclement frame is labeled as background classification, obtains the first training set；
S3. it is directed to the lighting apparatus sample image with encirclement frame described in step S2 and carries out scaling processing, by every
The longer sides of the encirclement frame of sample image transform to goal-selling size, and shorter edge transforms to goal-selling size according to longer sides
Zoom scale carry out same ratio scaling, obtain the second training set；
S4. using COCO data sets input CNN models to carry out pre-training, simultaneously iteration pre-training obtains model after pre-training, by step
Model carries out being directed to training after the second training set input pre-training described in S3 and iteration obtains tunnel illumination for training and sets
Standby detection model；
S5. the testing image for acquiring tunnel illumination equipment in real time, zooms in and out rear defeated according to the zoom scale described in step S3
Enter into the tunnel illumination equipment detection model described in step S4, calculates the classification results conduct that output confidence level is more than 90%
The recognition result of testing image.
2. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 1 based on CNN, which is characterized in that
Lighting apparatus described in step S1 is the image of the emergent pilot light equipment of tunnel internal, the emergent pilot light equipment
Including square shell and two groups of lighting pilot lamps being connected on square shell.
3. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 1 based on CNN, which is characterized in that
In step S1, for camera using lighting apparatus as object, coverage is that horizontal left avertence difference is 15 °~30 ° visual angles, horizontal right avertence
Difference between 15 °~30 ° visual angles range and overlook deviation be 50 °~70 ° visual angles, look up deviation and be 50 °~70 ° visual angles it
Between range.
4. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 1 based on CNN, which is characterized in that
Encirclement frame described in step S2 is rectangle frame, and complete lighting apparatus is that the region area of non-targeted object in encirclement frame accounts for packet
The ratio of peripheral frame area is less than 15% lighting apparatus.
5. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 1 based on CNN, which is characterized in that
CNN models described in step S4 are by input layer, convolutional layer, pond layer, full articulamentum and output layer building.
6. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 5 based on CNN, which is characterized in that
The establishment step of model is as follows after pre-training in step S4:
S41. using lighting apparatus sample image as input layer, sample image convolution extraction characteristic value is checked in convolutional layer；
S42. the characteristic value output in convolutional layer is input in the layer of pond as input and carries out maximum pond operation, reduce volume
The information of lamination；
S43. through the maximum pond operation in the multiple step S42 of convolution sum described in multilayer step S41, by the output of pond layer
As input, the operation of full articulamentum is carried out using different weights to each characteristic value, the two-dimensional signal of image is converted to
One-dimension information；
S44. classified to sample image according to the value of sample image one-dimension information, classification results are exported by output layer.
7. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 6 based on CNN, which is characterized in that
The neural network structure of the convolutional layer, pond layer stacked using four layers.
8. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 6 based on CNN, which is characterized in that
Weight is calculated as follows described in step S43：
In formula, loss function is mean square deviation MSE functions, WiIndicate i-th of weight of convolutional layer, biIndicate i-th of biasing of convolutional layer, Y
Indicate that entire sample set, Y (i) indicate the corresponding mark value of i-th of sample,Indicate defeated after network is trained in i-th of sample input
Go out the output token value of layer, η indicates the learning efficiency of back-propagation algorithm.
9. the abnormality recognition method of the cable tunnel lighting apparatus according to claim 1 based on CNN, which is characterized in that
The model training error of model is less than 10% after pre-training in step S4, the model training error of tunnel illumination equipment monitoring model
Average value be less than 5%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810404956.6A CN108615057B (en) | 2018-04-28 | 2018-04-28 | CNN-based abnormity identification method for cable tunnel lighting equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810404956.6A CN108615057B (en) | 2018-04-28 | 2018-04-28 | CNN-based abnormity identification method for cable tunnel lighting equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108615057A true CN108615057A (en) | 2018-10-02 |
| CN108615057B CN108615057B (en) | 2020-07-14 |
Family
ID=63661797
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810404956.6A Active CN108615057B (en) | 2018-04-28 | 2018-04-28 | CNN-based abnormity identification method for cable tunnel lighting equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108615057B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110688806A (en) * | 2019-11-29 | 2020-01-14 | 清华四川能源互联网研究院 | Hydraulic tunnel risk assessment method and device and terminal equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014191376A (en) * | 2013-03-26 | 2014-10-06 | Nohmi Bosai Ltd | Smoke detection device and smoke detection method |
| CN106504233A (en) * | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
| CN106841216A (en) * | 2017-02-28 | 2017-06-13 | 浙江工业大学 | Tunnel defect automatic identification equipment based on panoramic picture CNN |
| CN106898346A (en) * | 2017-04-19 | 2017-06-27 | 杭州派尼澳电子科技有限公司 | A kind of freeway tunnel safety monitoring system |
| KR101772916B1 (en) * | 2016-12-30 | 2017-08-31 | 한양대학교 에리카산학협력단 | Device for measuring crack width of concretestructure |
-
2018
- 2018-04-28 CN CN201810404956.6A patent/CN108615057B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014191376A (en) * | 2013-03-26 | 2014-10-06 | Nohmi Bosai Ltd | Smoke detection device and smoke detection method |
| CN106504233A (en) * | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
| KR101772916B1 (en) * | 2016-12-30 | 2017-08-31 | 한양대학교 에리카산학협력단 | Device for measuring crack width of concretestructure |
| CN106841216A (en) * | 2017-02-28 | 2017-06-13 | 浙江工业大学 | Tunnel defect automatic identification equipment based on panoramic picture CNN |
| CN106898346A (en) * | 2017-04-19 | 2017-06-27 | 杭州派尼澳电子科技有限公司 | A kind of freeway tunnel safety monitoring system |
Non-Patent Citations (4)
| Title |
|---|
| EFTYCHIOS PROTOPAPADAKIS等: "Image Based Approaches for Tunnels’ Defects Recognition via Robotic Inspectors", 《INTERNATIONAL SYMPOSIUM ON VISUAL COMPUTING》 * |
| HAO WUA等: "CNN refinement based object recognition through optimized segmentation", 《OPTIK》 * |
| HAO WU等: "Optimized CNN Based Image Recognition Through Target Region Selection，", 《OPTIK》 * |
| 汤一平等: "基于全景图像CNN的隧道病害自动识别方法", 《计算机科学》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110688806A (en) * | 2019-11-29 | 2020-01-14 | 清华四川能源互联网研究院 | Hydraulic tunnel risk assessment method and device and terminal equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108615057B (en) | 2020-07-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021138995A1 (en) | Fully automatic detection method for checkerboard corners | |
| CN110414334B (en) | Intelligent water quality identification method based on unmanned aerial vehicle inspection | |
| CN108564065B (en) | Cable tunnel open fire identification method based on SSD | |
| CN108009515B (en) | Power transmission line positioning and identifying method of unmanned aerial vehicle aerial image based on FCN | |
| CN110826520B (en) | Port grab bucket detection method based on improved YOLOv3-tiny algorithm | |
| CN105574527B (en) | A kind of quick object detecting method based on local feature learning | |
| CN108846418A (en) | A kind of positioning of cable machinery temperature anomaly and recognition methods | |
| CN107944396A (en) | A kind of disconnecting link state identification method based on improvement deep learning | |
| CN107833221A (en) | A kind of water leakage monitoring method based on multi-channel feature fusion and machine learning | |
| CN111260788B (en) | Power distribution cabinet switch state identification method based on binocular vision | |
| CN108010025B (en) | Switch and indicator lamp positioning and state identification method of screen cabinet based on RCNN | |
| CN108918536A (en) | Tire-mold face character defect inspection method, device, equipment and storage medium | |
| CN110400315A (en) | A kind of defect inspection method, apparatus and system | |
| CN111260621A (en) | Method for positioning and identifying surface defects of printed circuit board | |
| CN107240112A (en) | Individual X Angular Point Extracting Methods under a kind of complex scene | |
| CN109191421A (en) | Cylindricality lithium battery periphery pit visible detection method | |
| CN112258490A (en) | Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion | |
| CN109685774A (en) | Varistor open defect detection method based on depth convolutional neural networks | |
| CN110659637A (en) | Electric energy meter number and label automatic identification method combining deep neural network and SIFT features | |
| CN108154496B (en) | Electric equipment appearance change identification method suitable for electric power robot | |
| CN116012579A (en) | Method for detecting abnormal states of parts based on photographed images of intelligent inspection robot of train | |
| CN110321869A (en) | Personnel's detection and extracting method based on Multiscale Fusion network | |
| CN108615057A (en) | A kind of abnormality recognition method of the cable tunnel lighting apparatus based on CNN | |
| CN117314986A (en) | Unmanned aerial vehicle cross-mode power distribution equipment inspection image registration method based on semantic segmentation | |
| CN116245793A (en) | Color ring resistor resistance value detection method based on vision |
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 |