CN114387261A - Automatic detection method suitable for railway steel bridge bolt diseases - Google Patents
Automatic detection method suitable for railway steel bridge bolt diseases Download PDFInfo
- Publication number
- CN114387261A CN114387261A CN202210046066.9A CN202210046066A CN114387261A CN 114387261 A CN114387261 A CN 114387261A CN 202210046066 A CN202210046066 A CN 202210046066A CN 114387261 A CN114387261 A CN 114387261A
- Authority
- CN
- China
- Prior art keywords
- data set
- bolt
- diseases
- image
- steel bridge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an automatic detection method suitable for railway steel bridge bolt diseases, which comprises the following steps: acquiring an image of the steel bridge bolt; preprocessing the acquired image; classifying and labeling unit images containing damaged bolts; making a data set; building a target detection network SSD; training a target detection network (SSD); and inputting the image for identification into the SSD network model, and automatically detecting the bolt diseases. The invention provides an automatic detection method for railway steel bridge bolt diseases, which can be used for identifying a bridge inspection image, wherein the identification result comprises the types of the diseases, the positions of the diseases and the confidence coefficient of the occurrence of the diseases.
Description
Technical Field
The invention belongs to the technical field of railway bridge detection and evaluation, and particularly relates to an automatic detection method suitable for railway steel bridge bolt diseases.
Background
With the rapid development of railway construction in China, in recent years, the construction scale and the number of railway bridges are greatly increased. The steel bridge is used as an important component of a railway bridge, and the later detection and operation and maintenance work of the steel bridge plays a vital role in the safe operation of the railway.
In the diseases of the railway steel bridge bolt, two diseases of bolt corrosion and bolt shedding loss are common. At present, the detection of the steel bridge bolt diseases is carried out by adopting a manual detection mode. The mode of patrolling and examining through the manual work of bridge inspection personnel adopts closely visual observation, strikes traditional detection means such as waiting and detects the bolt disease of steel bridge. Because railway steel bridge bolt uses a large amount, and detection workload is big, relies on traditional artifical mode that detects to can not satisfy the detection needs gradually to traditional artifical detection precision is on the low side, very relies on detection personnel's experience and responsibility, and to the position that detection personnel are difficult for arriving, it is big to develop the degree of difficulty to the detection of bolt, produces the condition of false retrieval and even missed retrieval easily.
With the rapid development of artificial intelligence technology represented by deep learning in recent years, a bridge detection method based on the combination of computer vision and deep learning is gradually applied to bridge inspection work. By introducing the automatic detection method, the inspection efficiency of the bridge is greatly improved, the overall detection precision is improved, the probability of false detection and missed detection is reduced, and the labor intensity of inspection personnel is reduced. The application and research of the existing automatic detection technology for the steel bridge bolt diseases have defects, and an urgent need exists for researching an automatic detection method suitable for the railway steel bridge bolt diseases.
Disclosure of Invention
The invention is provided for solving the problems in the prior art, and aims to provide an automatic detection method suitable for railway steel bridge bolt diseases.
The technical scheme of the invention is as follows: an automatic detection method suitable for railway steel bridge bolt diseases comprises the following steps:
acquiring images of steel bridge bolts
Ii, preprocessing the collected image
Iii, classifying and labeling unit images containing damaged bolts
Iv. making a data set
V. building a target detection network SSD
Vi.training target detection networks SSD
And vii, inputting the image for identification into the SSD network model, and automatically detecting the bolt diseases.
Furthermore, step i is to collect the image of the steel bridge bolt, and the specific process is as follows:
the high-definition camera or the industrial camera is adopted to collect the image of the steel bridge bolt, and the collected image data is stored in a collection database.
Further, step ii is to pre-process the acquired image, specifically as follows:
firstly, carrying out size transformation on an image in an acquisition database to obtain a target image;
then, the transformed image is divided into a plurality of unit images;
finally, the obtained unit image is stored.
Further, step iii classifies and labels the unit images containing the damaged bolts, and the specific process is as follows:
firstly, selecting the positions of the damaged bolts in the unit images,
then, the disease type of the selected position is confirmed,
and finally, generating a data file, wherein the data file comprises the marked position information and the marked damage information of the damaged bolt.
Further, step iv is to create a data set, specifically including the following steps:
the established data sets include an A data set, an I data set and a J data set
The data set A stores position information and disease information of the disease bolt;
the I data set stores image names corresponding to the training data set, the test data set and the verification data set;
the J dataset stores the original segmented unit images.
And dividing the data set A, the data set I and the data set J into a training data set, a testing data set and a verification data set according to a fixed proportion mode.
Further, step v, building a target detection network SSD, specifically including the following steps:
the target detection network SSD belongs to a single-stage target detection network and comprises a backbone feature extraction network, an auxiliary convolution network and a prediction convolution network, wherein the target detection network SSD is used for classifying and position regression of a template frame to obtain the type of the bolt diseases and the positions of the bolts where the diseases occur.
Further, step vi, training the target detection network SSD, specifically including the following processes:
the method comprises the steps of classifying the disease types of template frames and performing position regression, obtaining the confidence coefficient of each template frame belonging to a certain type and the position prediction information of the template frame, and evaluating the boundary error of the SSD network prediction template frame and a real frame through the weighted loss function of the confidence coefficient and the position prediction of each disease type.
Furthermore, step vii inputs the image for recognition into the SSD network model, performs automatic detection on the bolt damage, and the trained training target detection network SSD can output the detection result of the bolt damage.
The invention provides an automatic detection method for railway steel bridge bolt diseases, which can be used for identifying a bridge inspection image, wherein the identification result comprises the types of the diseases, the positions of the diseases and the confidence coefficient of the occurrence of the diseases.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a graph of the predicted results of the present invention.
Detailed Description
The present invention is described in detail below with reference to the accompanying drawings and examples:
as shown in figures 1-2, an automatic detection method suitable for railway steel bridge bolt diseases comprises the following steps:
acquiring images of steel bridge bolts
Ii, preprocessing the collected image
Iii, classifying and labeling unit images containing damaged bolts
Iv. making a data set
V. building a target detection network SSD
Vi.training target detection networks SSD
And vii, inputting the image for identification into the SSD network model, and automatically detecting the bolt diseases.
Step i, collecting an image of the steel bridge bolt, wherein the specific process is as follows:
the high-definition camera or the industrial camera is adopted to collect the image of the steel bridge bolt, and the collected image data is stored in a collection database.
Step ii, preprocessing the acquired image, specifically comprising the following steps:
firstly, carrying out size transformation on an image in an acquisition database to obtain a target image;
then, the transformed image is divided into a plurality of unit images;
finally, the obtained unit image is stored.
Step iii, classifying and labeling the unit images containing the damaged bolts, wherein the specific process is as follows:
firstly, selecting the positions of the damaged bolts in the unit images,
then, the disease type of the selected position is confirmed,
and finally, generating a data file, wherein the data file comprises the marked position information and the marked damage information of the damaged bolt.
Step iv, data set manufacturing, which comprises the following specific processes:
the established data sets include an A data set, an I data set and a J data set
The data set A stores position information and disease information of the disease bolt;
the I data set stores image names corresponding to the training data set, the test data set and the verification data set;
the J dataset stores the original segmented unit images.
And dividing the data set A, the data set I and the data set J into a training data set, a testing data set and a verification data set according to a fixed proportion mode.
The training data set is a data sample used for model training fitting and is used for training and fitting SSD model parameters;
and the verification data set is used for adjusting the hyper-parameters of the model in the model training, preliminarily evaluating the prediction capability of the model in the iterative training process, and verifying the generalization capability of the current model to determine whether to stop the training.
The test data set is used to evaluate the generalization ability of the model that is ultimately trained.
And step v, building a target detection network SSD, wherein the specific process is as follows:
the target detection network SSD belongs to a single-stage target detection network and comprises a backbone feature extraction network, an auxiliary convolution network and a prediction convolution network, wherein the target detection network SSD is used for classifying and position regression of a template frame to obtain the type of the bolt diseases and the positions of the bolts where the diseases occur.
Step vi, training the target detection network SSD, and the specific process is as follows:
the method comprises the steps of classifying the disease types of template frames and performing position regression, obtaining the confidence coefficient of each template frame belonging to a certain type and the position prediction information of the template frame, and evaluating the boundary error of the SSD network prediction template frame and a real frame through the weighted loss function of the confidence coefficient and the position prediction of each disease type.
And step vii, inputting the image for identification into the SSD network model, automatically detecting the bolt diseases, and outputting the detection result of the bolt diseases by the trained training target detection network SSD.
The data set A, the data set I and the data set J are divided into a training data set, a testing data set and a verification data set according to a fixed proportion mode, and the fixed proportion is 3:1: 1.
Example one
An automatic detection method suitable for railway steel bridge bolt diseases comprises the following steps:
acquiring images of steel bridge bolts
Ii, preprocessing the collected image
Iii, classifying and labeling unit images containing damaged bolts
Iv. making a data set
V. building a target detection network SSD
Vi.training target detection networks SSD
And vii, inputting the image for identification into the SSD network model, and automatically detecting the bolt diseases.
Step i, collecting an image of the steel bridge bolt, wherein the specific process is as follows:
the high-definition camera or the industrial camera is adopted to collect the image of the steel bridge bolt, and the collected image data is stored in a collection database.
Step ii, preprocessing the acquired image, specifically comprising the following steps:
firstly, carrying out size transformation on an image in an acquisition database to obtain a target image;
then, the transformed image is divided into a plurality of unit images;
finally, the obtained unit image is stored.
More specifically, let the size of the captured image be a × b, let x be ceil (a/300), and y be ceil (b/300), where ceil is an upward integer function. The size of the transformed image is (x 300) x (y 300). This transformation ensures that the image can be completely divided into sub-images of 300 x 300 size. And dividing the transformed picture into a plurality of unit images, wherein the size of each unit image is 300 pixels by 300 pixels. One acquired image can be segmented into x y elemental images.
Step iii, classifying and labeling the unit images containing the damaged bolts, wherein the specific process is as follows:
firstly, selecting the positions of the damaged bolts in the unit images,
then, the disease type of the selected position is confirmed,
and finally, generating a data file, wherein the data file comprises the marked position information and the marked damage information of the damaged bolt.
Step iv, data set manufacturing, which comprises the following specific processes:
the established data sets include an A data set, an I data set and a J data set
The data set A stores position information and disease information of the disease bolt;
the I data set stores image names corresponding to the training data set, the test data set and the verification data set;
the J dataset stores the original segmented unit images.
And dividing the data set A, the data set I and the data set J into a training data set, a testing data set and a verification data set according to a fixed proportion mode.
The method for training the data set includes the steps that a data set marking tool labellimg is used, and bolt corrosion diseases and bolt missing and falling diseases in unit images are marked manually.
And step v, building a target detection network SSD, wherein the specific process is as follows:
the target detection network SSD belongs to a single-stage target detection network and comprises a backbone feature extraction network, an auxiliary convolution network and a prediction convolution network, wherein the target detection network SSD is used for classifying and position regression of a template frame to obtain the type of the bolt diseases and the positions of the bolts where the diseases occur.
The detection network extracts disease features by performing convolution operation on the unit images, and the result obtained by performing convolution operation on the unit images is called a feature map. The network contains convolutional layers for performing convolutional operations.
The backbone feature extraction network is a network framework of VGG-16, and extracts 6 feature maps with different scales together with the auxiliary convolution network. One template box is generated at each grid center point position on the feature map, and 8732 template boxes are generated for 6 feature maps.
The prediction convolution network is used for classifying and position regression of the template frame to obtain the disease type of the bolt and the position of the generated disease. And classifying the disease types corresponding to the template frame by using a Softmax classifier in machine learning, wherein the classification types are three types, namely bolt corrosion, bolt loss and image background, and the disease type classification operation is completed.
The position of the template frame is described using four parameters, center point coordinates (x, y), length and width of the frame (a, b). And correcting the position of the template frame by performing 4-channel convolution operation again, wherein 4 channels respectively correspond to 4 position parameters of the template frame to obtain position prediction information of the template frame. The position regression operation is completed.
Step vi, training the target detection network SSD, and the specific process is as follows:
the method comprises the steps of classifying the disease types of template frames and performing position regression, obtaining the confidence coefficient of each template frame belonging to a certain type and the position prediction information of the template frame, and evaluating the boundary error of the SSD network prediction template frame and a real frame through the weighted loss function of the confidence coefficient and the position prediction of each disease type.
Specifically, the weighting loss function is shown as follows:
wherein, x ═ {1,0} represents the matching condition of the template frame and the real boundary frame;
m represents the number of matched template boxes;
c represents a category confidence;
l represents the predicted result;
g represents a real bounding box;
the alpha parameter is the ratio between the adjusted confidence loss and the position loss, and the default parameter is 1.
And continuously carrying out iterative training by using a back propagation principle, and training and adjusting model parameters until the model converges to finish training. And (5) iterating until the weighting loss function value is converged, finishing training, and storing the SSD network model for the operation of the next step.
And step vii, inputting the image for identification into the SSD network model, automatically detecting the bolt diseases, and outputting the detection result of the bolt diseases by the trained training target detection network SSD.
And (3) finishing the SSD network model by training, inputting the generated 300 × 300 unit images, and repeating the disease classification and position regression operation to obtain all prediction results.
In order to ensure the identification precision, template frames with confidence degrees larger than 0.8 are reserved. And calculating the area of all the reserved template frames by adopting a non-maximum inhibition algorithm, then calculating the IOU (input output) intersection ratio of the template frame with the highest confidence coefficient of a certain type of diseases and other template frames, and deleting the template frames with the IOU larger than 0.25, thereby realizing the removal operation of the overlapped template frames and obtaining the prediction result of the sub-images. And (3) re-splicing the sub-images of which the prediction is finished by 300 × 300 to finally generate a prediction result, as shown in fig. 2.
The data set A, the data set I and the data set J are divided into a training data set, a testing data set and a verification data set according to a fixed proportion mode, and the fixed proportion is 3:1: 1.
The invention provides an automatic detection method for railway steel bridge bolt diseases, which can be used for identifying a bridge inspection image, wherein the identification result comprises the types of the diseases, the positions of the diseases and the confidence coefficient of the occurrence of the diseases.
Claims (9)
1. An automatic detection method suitable for railway steel bridge bolt diseases is characterized by comprising the following steps: the method comprises the following steps:
acquiring images of the steel bridge bolt
(ii) preprocessing the acquired image
(iii) classifying and labeling unit images containing damaged bolts
(iv) preparing a data set
(v) building a target detection network SSD
(vi) training target detection networks SSD
(vii) inputting the image for recognition into the SSD network model, and automatically detecting the bolt disease.
2. The automatic detection method for the railway steel bridge bolt diseases according to claim 1, which is characterized by comprising the following steps: acquiring an image of the steel bridge bolt, wherein the specific process is as follows:
the high-definition camera or the industrial camera is adopted to collect the image of the steel bridge bolt, and the collected image data is stored in a collection database.
3. The automatic detection method for the railway steel bridge bolt diseases according to claim 1, which is characterized by comprising the following steps: and (ii) preprocessing the acquired image, wherein the specific process is as follows:
firstly, carrying out size transformation on an image in an acquisition database to obtain a target image;
then, the transformed image is divided into a plurality of unit images;
finally, the obtained unit image is stored.
4. The automatic detection method for the railway steel bridge bolt diseases according to claim 1, which is characterized by comprising the following steps: step (iii) classifying and labeling the unit images containing the damaged bolts, wherein the specific process is as follows:
firstly, selecting the positions of the damaged bolts in the unit images,
then, the disease type of the selected position is confirmed,
and finally, generating a data file, wherein the data file comprises the marked position information and the marked damage information of the damaged bolt.
5. The automatic detection method for the railway steel bridge bolt diseases according to claim 1, which is characterized by comprising the following steps: step (iv) is to make a data set, and the specific process is as follows:
the established data sets include an A data set, an I data set and a J data set
The data set A stores position information and disease information of the disease bolt;
the I data set stores image names corresponding to the training data set, the test data set and the verification data set;
the J dataset stores the original segmented unit images.
6. The automatic detection method for the railway steel bridge bolt diseases according to claim 5, which is characterized in that: and dividing the data set A, the data set I and the data set J into a training data set, a testing data set and a verification data set according to a fixed proportion mode.
7. The automatic detection method for the railway steel bridge bolt diseases according to claim 1, which is characterized by comprising the following steps: and (v) building a target detection network (SSD), wherein the specific process is as follows:
the target detection network SSD belongs to a single-stage target detection network and comprises a backbone feature extraction network, an auxiliary convolution network and a prediction convolution network, wherein the target detection network SSD is used for classifying and position regression of a template frame to obtain the type of the bolt diseases and the positions of the bolts where the diseases occur.
8. The automatic detection method for the railway steel bridge bolt diseases according to claim 7 is characterized in that: step (vi) training the target detection network SSD, which comprises the following specific processes:
the method comprises the steps of classifying the disease types of template frames and performing position regression, obtaining the confidence coefficient of each template frame belonging to a certain type and the position prediction information of the template frame, and evaluating the boundary error of the SSD network prediction template frame and a real frame through the weighted loss function of the confidence coefficient and the position prediction of each disease type.
9. The automatic detection method for the railway steel bridge bolt diseases according to claim 1, which is characterized by comprising the following steps: and (vii) inputting the image for identification into the SSD network model, automatically detecting the bolt diseases, and outputting the detection result of the bolt diseases by the trained training target detection network SSD.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210046066.9A CN114387261A (en) | 2022-01-14 | 2022-01-14 | Automatic detection method suitable for railway steel bridge bolt diseases |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210046066.9A CN114387261A (en) | 2022-01-14 | 2022-01-14 | Automatic detection method suitable for railway steel bridge bolt diseases |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114387261A true CN114387261A (en) | 2022-04-22 |
Family
ID=81202745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210046066.9A Pending CN114387261A (en) | 2022-01-14 | 2022-01-14 | Automatic detection method suitable for railway steel bridge bolt diseases |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114387261A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188464A (en) * | 2023-04-24 | 2023-05-30 | 中铁四局集团有限公司 | Switch coupling bolt disease detection method and system |
| CN117351320A (en) * | 2023-08-25 | 2024-01-05 | 中铁大桥局集团第二工程有限公司 | Bolt step-by-step detection method based on deep learning |
-
2022
- 2022-01-14 CN CN202210046066.9A patent/CN114387261A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188464A (en) * | 2023-04-24 | 2023-05-30 | 中铁四局集团有限公司 | Switch coupling bolt disease detection method and system |
| CN117351320A (en) * | 2023-08-25 | 2024-01-05 | 中铁大桥局集团第二工程有限公司 | Bolt step-by-step detection method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
| CN111223088B (en) | Casting surface defect identification method based on deep convolutional neural network | |
| WO2022077605A1 (en) | Wind turbine blade image-based damage detection and localization method | |
| CN108711148B (en) | Tire defect intelligent detection method based on deep learning | |
| CN111696075A (en) | Intelligent fan blade defect detection method based on double-spectrum image | |
| CN111861978A (en) | Bridge crack example segmentation method based on Faster R-CNN | |
| CN111368690B (en) | Deep learning-based video image ship detection method and system under influence of sea waves | |
| CN109858367B (en) | Visual automatic detection method and system for worker through supporting unsafe behaviors | |
| CN111724355B (en) | Image measuring method for abalone body type parameters | |
| CN114387261A (en) | Automatic detection method suitable for railway steel bridge bolt diseases | |
| CN114399672A (en) | Railway wagon brake shoe fault detection method based on deep learning | |
| CN111178177A (en) | Cucumber disease identification method based on convolutional neural network | |
| CN113674216A (en) | Subway tunnel disease detection method based on deep learning | |
| CN114170511A (en) | Pavement crack disease identification method based on Cascade RCNN | |
| CN114155474A (en) | Damage identification technology based on video semantic segmentation algorithm | |
| CN112508911A (en) | Rail joint touch net suspension support component crack detection system based on inspection robot and detection method thereof | |
| CN114419421A (en) | Subway tunnel crack identification system and method based on images | |
| CN113313107A (en) | Intelligent detection and identification method for multiple types of diseases on cable surface of cable-stayed bridge | |
| CN115995056A (en) | Automatic bridge disease identification method based on deep learning | |
| CN113378831B (en) | Mouse embryo organ identification and scoring method and system | |
| CN113962929A (en) | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line | |
| CN113962951A (en) | Training method and device for detecting segmentation model, and target detection method and device | |
| CN116052110B (en) | Intelligent positioning method and system for pavement marking defects | |
| CN115830302A (en) | Multi-scale feature extraction and fusion power distribution network equipment positioning identification method | |
| Zhang et al. | Pavement crack detection based on deep learning |
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 |