CN108615057B - CNN-based abnormity identification method for cable tunnel lighting equipment - Google Patents
CNN-based abnormity identification method for cable tunnel lighting equipment Download PDFInfo
- Publication number
- CN108615057B CN108615057B CN201810404956.6A CN201810404956A CN108615057B CN 108615057 B CN108615057 B CN 108615057B CN 201810404956 A CN201810404956 A CN 201810404956A CN 108615057 B CN108615057 B CN 108615057B
- Authority
- CN
- China
- Prior art keywords
- lighting equipment
- layer
- image
- cnn
- training
- 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.)
- Active
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 invention relates to the technical field of computer image processing, in particular to an anomaly identification method of cable tunnel lighting equipment based on CNN (convolutional neural network), which selects a mode of training a CNN convolutional neural network model, fully utilizes the capability of a convolutional neural network for extracting two-dimensional picture features, and makes up the defect that the traditional method is insufficient in feature description or is difficult to select proper features. The method can detect the condition of the lighting equipment in the image, has good stability, is not influenced by other noises and light rays in the image to be detected, can accurately detect and position the target object in the picture of the lighting equipment shot by the inspection robot to be detected under the two conditions of opening and closing of the lighting equipment, has strong anti-jamming capability and good robustness, and can improve the detection accuracy of the lighting equipment in the cable tunnel. The tunnel internal equipment detection under dim and complex background has universality and wider application range.
Description
Technical Field
The invention relates to the technical field of computer image processing, in particular to a CNN-based abnormity identification method for cable tunnel lighting equipment.
Background
The power cable is in the confined tunnel environment, and inside crowdedly, dim, to cable tunnel's quality of patrolling and examining, patrol and examine speed and determine by the inside light luminance of tunnel usually, consequently provide the unusual guarantee that the lighting apparatus of only light source patrols and examines smoothly to cable tunnel to the inside detection of anomaly of cable tunnel. At present, the mode of manual inspection is often adopted to detect the abnormity of the lighting equipment, but because the cable laying length is longer, the internal environment is crowded, the efficiency of manual inspection is low, and the lighting equipment with abnormity is inconvenient to rapidly and correctly process.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a CNN-based abnormity identification method for cable tunnel lighting equipment, which can complete online monitoring of the lighting equipment by using an image processing technology, is convenient for operation and maintenance personnel to quickly and correctly process the abnormal lighting equipment and realizes intellectualization, rapidness and accuracy of cable tunnel inspection.
In order to solve the technical problems, the invention adopts the technical scheme that:
the method for identifying the abnormity of the CNN-based cable tunnel lighting equipment comprises the following steps:
s1, shooting and collecting a sample image of the lighting equipment through a camera of a tunnel inspection robot to obtain an image set;
s2, traversing all sample images in the image set in the step S1, surrounding complete lighting equipment by using a surrounding frame for each sample image, marking the sample images as lighting equipment sample images with the surrounding frame, marking pixel points in the surrounding frame as a lighting equipment category, and marking pixel points outside the surrounding frame as a background category to obtain a first training set;
s3, carrying out scale scaling processing on the lighting equipment sample image with the surrounding frame in the step S2, converting the longer side of the surrounding frame of each sample image into a preset target size, and carrying out scaling with the same proportion on the shorter side according to the scaling scale converted from the longer side to the preset target size to obtain a second training set;
s4, inputting a COCO data set into the CNN model for pre-training and iterative pre-training to obtain a pre-trained model, inputting the second training set in the step S3 into the pre-trained model for targeted training and iterative targeted training to obtain a tunnel lighting equipment detection model;
s5, acquiring an image to be detected of the tunnel lighting equipment in real time, zooming the image according to the zooming scale in the step S3, inputting the zoomed image into the tunnel lighting equipment detection model in the step S4, and calculating a classification result with the output confidence coefficient larger than 90% as an identification result of the image to be detected.
According to the abnormity identification method of the cable tunnel lighting equipment based on the CNN, the mode of training the CNN convolutional neural network model is selected, the capability of the convolutional neural network for extracting the two-dimensional picture features is fully utilized, and the defect that the traditional method is insufficient in feature description or is difficult to select proper features is overcome. The method can detect the condition of the lighting equipment in the image, has good stability, is not influenced by other noises and light rays in the image to be detected, can accurately detect and position the target object in the picture of the lighting equipment shot by the inspection robot to be detected under the two conditions of opening and closing of the lighting equipment, has strong anti-jamming capability and good robustness, and can improve the detection accuracy of the lighting equipment in the cable tunnel. The tunnel internal equipment detection under dim and complex background has universality and wider application range.
Preferably, the lighting device in step S1 is an image of an emergency indicator light lighting device inside a tunnel, and the emergency indicator light lighting device includes a square housing and two sets of lighting indicator lights connected to the square housing. The illumination indicator lamp is divided into two states of 'on' and 'off', wherein the 'on' state indicates that the illumination indicator lamp has no abnormal illumination, and the 'off' state indicates that the illumination indicator lamp has abnormal illumination.
Preferably, the camera takes the lighting device as a target object, and the shooting range is a range with a horizontal left deviation of 15-30 degrees of visual angle and a horizontal right deviation of 15-30 degrees of visual angle, and a range with a top-view deviation of 50-70 degrees of visual angle and a bottom-view deviation of 50-70 degrees of visual angle. The sample images of the lighting equipment can be collected from different visual angles, various sample images are obtained, and the accuracy of abnormal recognition can be improved.
Preferably, the enclosure frame in step S2 is a rectangular frame, and the complete lighting device is a lighting device in which the proportion of the area of the non-target object in the enclosure frame to the area of the enclosure frame is less than 15%.
Preferably, the CNN model described in step S4 is constructed by an input layer, a convolutional layer, a pooling layer, a fully-connected layer, and an output layer.
Preferably, the pre-trained model in step S4 is established as follows:
s41, taking the sample image of the lighting equipment as an input layer, checking the convolution of the sample image in a convolution layer and extracting a characteristic value;
s42, outputting the characteristic value in the convolutional layer as input, inputting the characteristic value into a pooling layer to perform maximum pooling operation, and reducing the information of the convolutional layer;
s43, performing convolution in the multilayer step S41 and maximum pooling operation in the multiple steps S42, taking the output of a pooling layer as input, performing full-connection layer operation on each characteristic value by adopting different weights, and converting two-dimensional information of the image into one-dimensional information;
and S44, classifying the sample images according to the values of the one-dimensional information of the sample images, and outputting the classification results by an output layer.
The relation between the input picture and the detection target is described through a characteristic extraction frame of the CNN, the error detection of the target is reduced, the step of non-maximum value inhibition is avoided, and the problems of detection and positioning of the lighting equipment under normal opening and abnormal closing are solved.
Preferably, the stack of the convolutional layer and the pooling layer adopts a neural network structure with four layers.
Preferably, the weight in step S43 is calculated as follows:
wherein the loss function is a Mean Square Error (MSE) function, WiDenotes the ith weight of the convolutional layer, biDenotes the ith offset of the convolutional layer, Y denotes the entire sample set, Y (i) denotes the label value corresponding to the ith sample,the output label value of the output layer after the ith sample is input into the training network is shown, and η shows the learning efficiency of the back propagation algorithm.
Preferably, the model training error of the pre-trained model in step S4 is less than 10%, and the average value of the model training errors of the tunnel lighting device monitoring model is less than 5%.
Compared with the prior art, the invention has the beneficial effects that:
the method can detect the condition of the lighting equipment in the image, has good stability, is not influenced by other noises and light rays in the image to be detected, can accurately detect and position the target object in the picture of the lighting equipment shot by the inspection robot to be detected under the two conditions of opening and closing of the lighting equipment, has strong anti-jamming capability and good robustness, and can improve the detection accuracy of the lighting equipment in the cable tunnel. The tunnel internal equipment detection under dim and complex background has universality and wider application range.
Detailed Description
The present invention will be further described with reference to the following embodiments.
Example one
The embodiment is a first embodiment of a CNN-based abnormality identification method for a cable tunnel lighting device, and includes the following steps:
s1, shooting and collecting a sample image of the lighting equipment through a camera of a tunnel inspection robot to obtain an image set;
s2, traversing all sample images in the image set in the step S1, surrounding complete lighting equipment by using a surrounding frame for each sample image, marking the sample images into lighting equipment sample images with the surrounding frame, marking pixel points in the surrounding frame as a lighting equipment category, and marking pixel points outside the surrounding frame as a background category to obtain a first training set;
s3, carrying out scale scaling processing on the lighting equipment sample image with the surrounding frame in the step S2, converting the longer side of the surrounding frame of each sample image into a preset target size, and carrying out scaling with the same proportion on the shorter side according to the scaling scale converted from the longer side to the preset target size to obtain a second training set;
s4, inputting a COCO data set into the CNN model for pre-training and iterative pre-training to obtain a pre-trained model, inputting the second training set in the step S3 into the pre-trained model for targeted training and iterative targeted training to obtain a tunnel lighting equipment detection model;
s5, acquiring an image to be detected of the tunnel lighting equipment in real time, zooming the image according to the zooming scale in the step S3, inputting the zoomed image into the tunnel lighting equipment detection model in the step S4, and calculating a classification result with the output confidence coefficient larger than 90% as an identification result of the image to be detected.
In step S1, the lighting device is an image of an emergency indicator light lighting device inside the tunnel, where the emergency indicator light lighting device includes a square housing and two sets of illumination indicator lights connected to the square housing; in step S1, the camera takes the lighting device as a target, and the shooting range is a range with a horizontal left deviation of 15 ° to 30 ° viewing angle and a horizontal right deviation of 15 ° to 30 ° viewing angle, and a range with a top-view deviation of 50 ° to 70 ° viewing angle and a bottom-view deviation of 50 ° to 70 ° viewing angle. The lighting equipment sample image of this embodiment can gather from different perspectives, obtains diversified sample image, improves the degree of accuracy of abnormal recognition. In the first training set, the lighting device sample image marks each pixel point therein to form an image mark set, and the image mark set stores mark data in the following form:
{image_name,label,x1,y1}
the image _ name represents the name of an image of lighting equipment shot by the inspection robot, the label represents the type of a pixel point, the x1 represents the abscissa of the pixel point, and the y1 represents the ordinate of the pixel point.
The enclosure frame in the step S2 is a rectangular frame, and the complete lighting device is a lighting device in which the proportion of the area of the non-target object in the enclosure frame to the area of the enclosure frame is less than 15%.
The CNN model described in step S4 is constructed by an input layer, a convolutional layer, a pooling layer, a full-link layer, and an output layer, the model training error of the pre-trained model is less than 10%, the average value of the model training errors of the tunnel lighting device monitoring model is less than 5%, and the building steps of the pre-trained model are as follows:
s41, taking the sample image of the lighting equipment as an input layer, checking the convolution of the sample image in a convolution layer and extracting a characteristic value;
s42, outputting the characteristic value in the convolutional layer as input, inputting the characteristic value into a pooling layer to perform maximum pooling operation, and reducing the information of the convolutional layer;
s43, performing convolution in the multilayer step S41 and maximum pooling operation in the multiple steps S42, taking the output of a pooling layer as input, performing full-connection layer operation on each characteristic value by adopting different weights, and converting two-dimensional information of the image into one-dimensional information;
and S44, classifying the sample images according to the values of the one-dimensional information of the sample images, and outputting the classification results by an output layer.
The relation between the input picture and the detection target is described through a characteristic extraction frame of the CNN, the error detection of the target is reduced, the step of non-maximum value inhibition is avoided, and the problems of detection and positioning of the lighting equipment under normal opening and abnormal closing are solved.
In step S43, the weight calculation of each layer is performed by inverse gradient calculation:
wherein the loss function is a Mean Square Error (MSE) function, WiDenotes the ith weight of the convolutional layer, biDenotes the ith offset of the convolutional layer, Y denotes the entire sample set, Y (i) denotes the label value corresponding to the ith sample,the output label value of the output layer after the ith sample is input into the training network is shown, and η shows the learning efficiency of the back propagation algorithm.
The convolutional layer and the pooling layer of the present embodiment are stacked in a four-layer neural network structure:
obtaining the output of the convolutional layer, then performing normalization processing by using a BN layer (batch normalization), then using a Re L U function (Rectified L initial Units) as a nonlinear activation function for activation, and finally performing pooling by using a maximum pooling layer (Maxpooling) with a window size of 3 × 3, wherein the sampling stride of the maximum pooling layer (Maxpooling) is 2;
after the output of the convolutional layer is obtained, a BN layer (batch normalization) is used for normalization, then a Re L U function (Rectified L initial Units) is used as a nonlinear activation function for activation, finally a maximum pooling layer (Maxpooling) with the window size of 3 × 3 is used for pooling, and the sampling stride of the maximum pooling layer (Maxpooling) is 2;
the third layer, firstly using convolution layer, using 96 convolution filters with size of 3 × 3, convolution step is 1, setting convolution offset distance pad to be 1 to make the dimension of input graph equal to output graph, and outputting 96 feature graphs with arbitrary size;
the fourth layer, use convolution layer first, convolution layer uses 48 convolution filters with size 3 × 3, convolution step is 1, and set convolution offset pad to be 1, and then use Re L U function (Rectified L initial Units) as activation function to activate after convolution;
the structure of the full-connection layer is that 256-dimensional features are processed and output by two full-connection layers, then a frame regression layer (smooth L1L oss L eye) is used for processing, the frame regression layer outputs a frame, four elements of the frame are obtained, the four elements are respectively the horizontal and vertical coordinates x and y of the upper left corner of the frame output by the frame regression layer and the width w and the height h of the frame output by the frame regression layer, and the frame is used as a real area of the lighting equipment possibility area as a target position.
The output layer is specifically of a structure that a convolution kernel is used for processing one-dimensional information output by the fully-connected layer, an output feature map is fixed to be 56 × 56, then the output feature map is input into a three-layer convolution layer with the convolution kernel size of 1 × 1, wherein the first layer of convolution layer has 1024-dimensional outputs, the second layer of convolution layer has 256-dimensional outputs, the third layer of convolution layer has 4-dimensional outputs, and the outputs of the third layer of convolution layer are input into a binary classifier.
Example two
Collecting 400 experimental pictures, wherein the inspection robots with different shooting angles shoot images of the tunnel internal lighting equipment, 200 pictures are used for training, 50 pictures of the states of each switch and each indicator light are respectively used, and the rest 200 pictures are used as test set pictures. Adopting a CNN model to detect an image of the inspection robot shooting the tunnel internal lighting equipment, and obtaining a conclusion: the correct coincidence rate of the abnormal equipment and the abnormal detection result is 100%, and the correct coincidence rate of the normal equipment and the normal detection result is 100%.
In summary, under the two conditions of turning on and turning off the lighting equipment, according to the different angles of the shot images, the method can also accurately detect and position the lighting equipment in the images and finish accurate result detection, so that the placement position of the camera and the fixed-point inspection position of the inspection robot can be more free, and some complex environmental conditions can be effectively dealt with.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (9)
1. An abnormality identification method for a CNN-based cable tunnel lighting device is characterized by comprising the following steps:
s1, shooting and collecting a sample image of the lighting equipment through a camera of a tunnel inspection robot to obtain an image set;
s2, traversing all sample images in the image set in the step S1, surrounding complete lighting equipment by using a surrounding frame for each sample image, marking the sample images into lighting equipment sample images with the surrounding frame, marking pixel points in the surrounding frame as a lighting equipment category, and marking pixel points outside the surrounding frame as a background category to obtain a first training set;
s3, carrying out scale scaling processing on the lighting equipment sample image with the surrounding frame in the step S2, converting the longer side of the surrounding frame of each sample image into a preset target size, and carrying out scaling with the same proportion on the shorter side according to the scaling scale converted from the longer side to the preset target size to obtain a second training set;
s4, inputting a COCO data set into the CNN model for pre-training and iterative pre-training to obtain a pre-trained model, inputting the second training set in the step S3 into the pre-trained model for targeted training and iterative targeted training to obtain a tunnel lighting equipment detection model;
s5, acquiring an image to be detected of the tunnel lighting equipment in real time, zooming the image according to the zooming scale in the step S3, inputting the zoomed image into the tunnel lighting equipment detection model in the step S4, and calculating a classification result with the output confidence coefficient larger than 90% as an identification result of the image to be detected.
2. The abnormality recognition method for CNN-based cable tunnel illumination equipment according to claim 1, wherein the illumination equipment in step S1 is an image of an emergency light illumination equipment inside a tunnel, the emergency light illumination equipment including a square housing and two sets of illumination lights connected to the square housing.
3. The method for identifying abnormality in CNN-based cable tunnel illumination equipment according to claim 1, wherein in step S1, the camera takes the illumination equipment as a target, and the photographing ranges are a range of viewing angles with a horizontal left deviation of 15 ° to 30 ° and a horizontal right deviation of 15 ° to 30 °, and a range of viewing angles with a top deviation of 50 ° to 70 ° and a range of viewing angles with a bottom deviation of 50 ° to 70 °.
4. The method for identifying abnormality in CNN-based cable tunnel illumination apparatus according to claim 1, wherein the surrounding frame in step S2 is a rectangular frame, and the complete illumination apparatus is an illumination apparatus in which the ratio of the area of the non-target object in the surrounding frame to the area of the surrounding frame is less than 15%.
5. The abnormality recognition method for CNN-based cable tunnel illumination apparatus according to claim 1, wherein the CNN model in step S4 is constructed of an input layer, a convolutional layer, a pooling layer, a full connection layer, and an output layer.
6. The method for recognizing the abnormality of the CNN-based cable tunnel illumination apparatus according to claim 5, wherein the pre-trained model in step S4 is established as follows:
s41, taking the sample image of the lighting equipment as an input layer, checking the convolution of the sample image in a convolution layer and extracting a characteristic value;
s42, outputting the characteristic value in the convolutional layer as input, inputting the characteristic value into a pooling layer to perform maximum pooling operation, and reducing the information of the convolutional layer;
s43, performing convolution in the multilayer step S41 and maximum pooling operation in the multiple steps S42, taking the output of a pooling layer as input, performing full-connection layer operation on each characteristic value by adopting different weights, and converting two-dimensional information of the image into one-dimensional information;
and S44, classifying the sample images according to the values of the one-dimensional information of the sample images, and outputting the classification results by an output layer.
7. The anomaly identification method for CNN-based cable tunnel lighting equipment according to claim 6, wherein the stack of convolutional layer and pooling layer adopts a four-layer neural network structure.
8. The abnormality recognition method for CNN-based cable tunnel illumination apparatuses according to claim 6, wherein the weight in step S43 is calculated as follows:
wherein the loss function is a Mean Square Error (MSE) function, WiDenotes the ith weight of the convolutional layer, biDenotes the ith offset of the convolutional layer, Y denotes the entire sample set, Y (i) denotes the label value corresponding to the ith sample,the output label value of the output layer after the ith sample is input into the training network is shown, and η shows the learning efficiency of the back propagation algorithm.
9. The method for recognizing the abnormality of the CNN-based cable tunnel lighting apparatus according to claim 1, wherein the model training error of the pre-trained model in the step S4 is less than 10%, and the average value of the model training errors of the monitoring model of the tunnel lighting apparatus is 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 CN108615057A (en) | 2018-10-02 |
CN108615057B true 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) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110688806B (en) * | 2019-11-29 | 2020-04-28 | 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 |
---|
CNN refinement based object recognition through optimized segmentation;Hao Wua等;《Optik》;20170918;第76-82页 * |
Image Based Approaches for Tunnels’ Defects Recognition via Robotic Inspectors;Eftychios Protopapadakis等;《International Symposium on Visual Computing》;20151218;第706-716页 * |
Optimized CNN Based Image Recognition Through Target Region Selection,;Hao Wu等;《Optik》;20171122;第772–777页 * |
基于全景图像CNN的隧道病害自动识别方法;汤一平等;《计算机科学》;20171115;第207-212页 * |
Also Published As
Publication number | Publication date |
---|---|
CN108615057A (en) | 2018-10-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108009515B (en) | Power transmission line positioning and identifying method of unmanned aerial vehicle aerial image based on FCN | |
CN108564065B (en) | Cable tunnel open fire identification method based on SSD | |
CN108010025B (en) | Switch and indicator lamp positioning and state identification method of screen cabinet based on RCNN | |
CN112199993B (en) | Method for identifying transformer substation insulator infrared image detection model in any direction based on artificial intelligence | |
WO2022170844A1 (en) | Video annotation method, apparatus and device, and computer readable storage medium | |
CN111563896B (en) | Image processing method for detecting abnormality of overhead line system | |
WO2022206161A1 (en) | Feature point recognition-based block movement real-time detection method | |
CN111062938A (en) | Plate expansion plug detection system and method based on machine learning | |
CN112258490A (en) | Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion | |
CN110991256A (en) | System and method for carrying out age estimation and/or gender identification based on face features | |
CN116797977A (en) | Method and device for identifying dynamic target of inspection robot and measuring temperature and storage medium | |
CN114092478B (en) | Anomaly detection method | |
CN108615057B (en) | CNN-based abnormity identification method for cable tunnel lighting equipment | |
CN108664886A (en) | A kind of fast face recognition method adapting to substation's disengaging monitoring demand | |
CN113205507B (en) | Visual question answering method, system and server | |
CN114170686A (en) | Elbow bending behavior detection method based on human body key points | |
CN117314986A (en) | Unmanned aerial vehicle cross-mode power distribution equipment inspection image registration method based on semantic segmentation | |
CN111582332A (en) | Picture identification method for dropper component of high-speed railway contact network | |
CN114913086B (en) | Face image quality enhancement method based on generation countermeasure network | |
CN116109849A (en) | SURF feature matching-based high-voltage isolating switch positioning and state identification method | |
CN111738148B (en) | Fault identification method using infrared inspection shooting | |
TWI747686B (en) | A defect detection method and a defect detection device | |
CN115147591A (en) | Transformer equipment infrared image voltage heating type defect diagnosis method and system | |
CN110956640B (en) | Heterogeneous image edge point detection and registration method | |
McAlorum et al. | Automated concrete crack inspection with directional lighting platform |
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 |