CN113362302A - Fault detection method of subway train electric box cover based on image recognition - Google Patents
Fault detection method of subway train electric box cover based on image recognition Download PDFInfo
- Publication number
- CN113362302A CN113362302A CN202110620998.5A CN202110620998A CN113362302A CN 113362302 A CN113362302 A CN 113362302A CN 202110620998 A CN202110620998 A CN 202110620998A CN 113362302 A CN113362302 A CN 113362302A
- Authority
- CN
- China
- Prior art keywords
- nameplate
- sample
- box cover
- fault
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- 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
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention relates to the technical field of train fault detection, in particular to a fault detection method of an electric box cover of a subway train based on image recognition, which comprises the following steps: 1. acquiring a positive and negative sample data set of the electrical box cover component by a line scanning camera; 2. preprocessing the data set and converting the data set into a pseudo-color image; 3. using the data set of the positive sample for the Nanodet training; 4. obtaining a classification positive sample set through cutting; turning over the negative samples, adding noise to enhance data, expanding the number of the negative samples, positioning and cutting to obtain a classified negative sample set; 5. training the classified negative sample set and the classified positive sample set together to finally obtain a fault detection model; 6. and inputting the sample to be detected into a fault detection model for testing. The invention can effectively detect the fault of the cover part of the electrical box and has high accuracy.
Description
Technical Field
The invention relates to the technical field of train fault detection, in particular to a fault detection method of an electric box cover of a subway train based on image recognition.
Background
In recent years, with the development of rail transit industry, the mileage and the number of trains of rail transit in China are increasing and are first in the world for a long time. Regular inspection of the train is the key for ensuring safe operation of the train, and the number of the overhaul trains brings great pressure to overhaul work. At present, the inspection work of the train still highly depends on manual work, and the fault detection of key parts at the bottom of the train needs to depend on the naked eyes of workers to detect each part. However, the maintenance mode is not only inefficient, but also has poor reliability, and wastes a large amount of manpower and material resources, so that a new technology is urgently needed for realizing efficient maintenance in the train inspection work.
With the continuous breakthrough of the related technology in the field of computer vision, the fault diagnosis technology based on computer vision is developed rapidly, which brings a new revolution for the revolution of train inspection work. At present, computer vision is mainly applied to the fields of security protection, automatic driving, face recognition and the like, and the land falling of the fields depends on a large number of related samples. However, a large number of normal samples exist in the train inspection work, and fault samples are extremely rare, so that the fault detection of train components by using computer vision is difficult.
Disclosure of Invention
The invention provides a fault detection method of a subway train electric box cover based on image recognition, which can overcome certain defects in the prior art.
The invention discloses a fault detection method of an electric box cover of a subway train based on image recognition, which comprises the following steps of:
1) acquiring a data set of the electrical box cover component by a line scanning camera, namely a data set containing a positive sample for subsequent training; a fault data set containing a negative sample is also obtained through the line scanning camera;
2) preprocessing the data set and converting the data set into a pseudo-color image;
3) using the data set of the positive sample for the Nanodet training;
4) when the target detection network training is finished, cutting the rectangular nameplate, the triangular nameplate and the lock catch from the data set of the electric box cover, and storing the rectangular nameplate, the triangular nameplate and the lock catch as a classified positive sample set; turning over the negative sample of the electric box cover, adding noise to enhance data, expanding the number of the negative sample, positioning parts on the expanded negative sample through a trained Nanodet neural network, cutting the positioned parts, and storing the parts as a classified negative sample set;
5) training classified negative sample sets and classified positive sample sets of various categories by using a pretrained VGG16 model carried by a keras, and finally obtaining three fault detection models respectively used for detecting a rectangular nameplate, a triangular nameplate and a lock catch;
6) and inputting the sample to be detected into the fault detection model for testing.
Preferably, step 6) comprises the steps of:
6.1) converting the sample to be detected into a pseudo-color image;
6.2) inputting the pseudo-color image into a trained lightweight target detection Nanodet model, positioning and cutting out a rectangular nameplate, a triangular nameplate and a lock catch;
6.3) respectively inputting the cut rectangular nameplate, the triangular nameplate and the lock catch data into the corresponding VGG16 classification networks to respectively obtain the number of the rectangular nameplate, the triangular nameplate and the lock catch in a fault-free mode;
6.4) comparing the number of the rectangular nameplates, the triangular nameplates and the latches which are obtained through statistics and have no fault with the number of the rectangular nameplates, the triangular nameplates and the latches under the condition of the positive sample of the electric box cover data set, and if the number of the rectangular nameplates, the triangular nameplates and the latches is equal, judging that the sample to be detected is a positive sample, otherwise, judging that the sample to be detected is a negative sample.
Preferably, in step 1), the number of positive samples is 1470 and the number of negative samples is 53.
Preferably, in step 2), the pseudo-color image is 3 channels.
Preferably, in step 3), the learning rate of the Nanodet training is set to 0.0001, and the number of iterations is set to 100 epochs.
The invention has the following beneficial effects:
1. in the actual operation process of the train, the model can effectively detect the faults of the cover part of the electric box, has high accuracy, the false alarm rate is only 1.89 percent, the missing report rate is 0, and can effectively avoid the interference caused by non-relevant factors such as rainwater, oil stain, dust, illumination and the like.
2. The model is suitable for a fault detection scene with simple and crude equipment and high requirement on detection speed, and although more parts to be detected exist on one electric box cover sample, the detection time of each picture in the CPU environment only needs 0.39s, and real-time detection can be realized.
3. The method has high accuracy and short detection time, and can ensure that detection personnel can be separated from the low severe working environment of the train and reduce the workload of maintenance personnel. The method has great significance for reducing the investment of manpower and material resources and ensuring the operation safety of the train.
Drawings
Fig. 1 is a flowchart of a method for detecting a fault of an electric box cover of a subway train based on image recognition in embodiment 1;
FIG. 2 is a schematic view of the rectangular nameplate, the triangular nameplate and the latch of embodiment 1.
Detailed Description
For a further understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings and examples. It is to be understood that the examples are illustrative of the invention and not limiting.
Example 1
As shown in fig. 1, the present embodiment provides a method for detecting a fault of an electric box cover of a subway train based on image recognition, which includes the following steps:
1) acquiring a data set of the electrical box cover component by a line scanning camera with the resolution of 2048, namely a data set containing 1470 positive samples for subsequent training; in addition, a fault data set containing 53 negative samples was acquired by the line scan camera. Because in this item, the border of rectangle data plate, triangle-shaped data plate and hasp is clear, simple structure, and the negative sample figure of this item simultaneously can, consequently takes the thought of first location back two categorizations to realize fault detection.
The data set of the electric appliance box cover component comprises 36 categories, and each sample to be tested has different numbers of rectangular nameplates, triangular nameplates and lock catches. The data set comprises a rectangular nameplate, a triangular nameplate and a lock catch as shown in fig. 2, wherein the technology needs to detect the fault conditions of falling, damage, displacement, shielding and the like of the three parts.
2) And because the acquired data set pictures are single-channel gray-scale pictures and the input of the neural network is 3 channels, the data set is preprocessed and converted into a 3-channel pseudo-color picture.
3) Considering that the requirement of the method on the detection time of a single picture is strict, and the length-width ratio difference of each type of electric box cover sample is extremely large, the light target detection network Nanodet is adopted in positioning, and firstly, a data set of a positive sample is used for Nanodet training; the learning rate is set to 0.0001 and the number of iterations is set to 100 epochs.
4) When the target detection network training is finished, cutting parts to be detected (a rectangular nameplate, a triangular nameplate and a lock catch) from the data set of the electric box cover, and storing the parts as a classified positive sample set; and turning over 53 negative samples of the electric box cover, adding noise to enhance data, expanding the number of the negative samples, positioning parts on the expanded negative samples through a trained Nanodet neural network, cutting the positioned parts, and storing the parts as a classified negative sample set.
5) The Keras pre-trained VGG16 model is used for training classified negative sample sets and classified positive sample sets of all categories together, and finally three fault detection models which are used for detecting the rectangular nameplate, the triangular nameplate and the lock catch are obtained.
6) Inputting a sample to be detected into a fault detection model (cascade model) for testing; the method comprises the following steps:
6.1) converting the sample picture to be detected into a pseudo-color picture;
6.2) inputting the pseudo-color image into a trained lightweight target detection Nanodet model, positioning and cutting out a rectangular nameplate, a triangular nameplate and a lock catch;
6.3) respectively inputting the cut rectangular nameplate, the triangular nameplate and the lock catch data into the corresponding VGG16 classification networks to respectively obtain the number of the rectangular nameplate, the triangular nameplate and the lock catch in a fault-free mode;
6.4) comparing the number of the rectangular nameplates, the triangular nameplates and the latches which are obtained through statistics and have no fault with the number of the rectangular nameplates, the triangular nameplates and the latches under the condition of the positive sample of the electric box cover data set, and if the number of the rectangular nameplates, the triangular nameplates and the latches is equal, judging that the sample to be detected is a positive sample, otherwise, judging that the sample to be detected is a negative sample.
In the embodiment, by using the latest lightweight target detection network Nanodet, aiming at the condition that the length-width ratio of the data set of the electric box cover has great difference, the network can effectively avoid the conditions of feature loss and the like caused by input picture compression. This network can realize effectively the accurate location to the rectangle data plate of electric box lid, triangle-shaped data plate and hasp.
In the embodiment, for the large size difference and limited sample number of various key parts, a fixed input size is set for each part, and a model is trained in a mode of fine tuning the VGG16 network by using a transfer learning technology. The final three-part classification network has high accuracy and high speed, and can effectively identify the fault form of the parts.
In the embodiment, the lightweight target detection network Nanodet and the VGG16 based on the transfer learning technology are combined to form a fault detection model of the electric box cover. The advantages of the two networks are fully exerted, and the data acquisition environments such as: the influence that illumination, rainwater, oil stain and dust etc. brought has guaranteed the rate of accuracy that detects, provides the powerful guarantee for train safe operation.
Example 2
In the embodiment, firstly, the data of a train running part is acquired by a linear array camera on the side surface of a train; then, correcting the image in a data processing center, cutting the electrical box cover part according to a preset position and transmitting the cut electrical box cover part to a server; then, detecting data through an electric box cover fault detection algorithm deployed on a server; and finally, feeding back the detection result to a maintainer, and carrying out fixed-point maintenance according to fault information.
The model training process is as follows:
firstly, preprocessing the collected pictures of the electric box cover by utilizing Labelimg to prepare a data set in a VOC2007 format.
And secondly, training a lightweight target detection network Nanodet by using the well-made data set, setting the learning rate to be 0.0001, setting the iteration cycle epoch to be 100, and obtaining an h5 weight file after training.
And thirdly, performing operations such as turning over and adding noise on the 53 negative samples of the electric box cover, and expanding the negative samples of the data set of the electric box cover.
And fourthly, positioning the data set of the electric box cover through the Nanodet and cutting out corresponding rectangular nameplates, triangular nameplates and negative samples of the locking parts.
And fifthly, expanding the number of the negative samples of the rectangular nameplate, the triangular nameplate and the locking part to 1036, 668 and 288 respectively by turning the obtained negative samples of the rectangular nameplate, the triangular nameplate and the locking part, adding noise and the like.
And sixthly, training the negative samples obtained by the expansion in the fourth step and corresponding positive samples through a VGG16 network based on transfer learning, wherein the number of the positive samples used for training by the rectangular nameplate, the triangular nameplate and the locking part is 859, 761 and 456 respectively. The learning rate of the VGG16 is set to be 0.0001, the iteration times are also set to be 100 epochs, and finally three VGG16 models corresponding to three key parts are obtained through training.
The model test procedure was as follows:
firstly, converting a sample picture into a pseudo-color picture to meet a model input format;
secondly, inputting the pseudo-color picture into a trained lightweight target detection Nanodet model, positioning and cutting out a rectangular nameplate, a triangular nameplate and a lock catch;
thirdly, respectively inputting the cut key part data into the corresponding VGG16 classification networks to obtain the total number of rectangular nameplates, triangular nameplates and latches in a fault-free mode;
and fourthly, comparing the counted number of the key parts without the fault form with the number of the key parts under the condition of the positive sample of the data set of the electric box cover, and if the number of the key parts is equal to the number of the key parts under the condition of the positive sample of the data set of the electric box cover, judging that the sample is a positive sample, otherwise, judging that the sample is a negative sample.
The present invention and its embodiments have been described above schematically, without limitation, and what is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Therefore, if the person skilled in the art receives the teaching, without departing from the spirit of the invention, the person skilled in the art shall not inventively design the similar structural modes and embodiments to the technical solution, but shall fall within the scope of the invention.
Claims (5)
1. A fault detection method of an electric box cover of a subway train based on image recognition is characterized by comprising the following steps: the method comprises the following steps:
1) acquiring a data set of the electrical box cover component by a line scanning camera, namely a data set containing a positive sample for subsequent training; a fault data set containing a negative sample is also obtained through the line scanning camera;
2) preprocessing the data set and converting the data set into a pseudo-color image;
3) using the data set of the positive sample for the Nanodet training;
4) when the target detection network training is finished, cutting the rectangular nameplate, the triangular nameplate and the lock catch from the data set of the electric box cover, and storing the rectangular nameplate, the triangular nameplate and the lock catch as a classified positive sample set; turning over the negative sample of the electric box cover, adding noise to enhance data, expanding the number of the negative sample, positioning parts on the expanded negative sample through a trained Nanodet neural network, cutting the positioned parts, and storing the parts as a classified negative sample set;
5) training classified negative sample sets and classified positive sample sets of various categories by using a pretrained VGG16 model carried by a keras, and finally obtaining three fault detection models respectively used for detecting a rectangular nameplate, a triangular nameplate and a lock catch;
6) and inputting the sample to be detected into the fault detection model for testing.
2. The method for detecting the fault of the electric box cover of the subway train based on the image recognition is characterized by comprising the following steps of: step 6) comprises the following steps:
6.1) converting the sample to be detected into a pseudo-color image;
6.2) inputting the pseudo-color image into a trained lightweight target detection Nanodet model, positioning and cutting out a rectangular nameplate, a triangular nameplate and a lock catch;
6.3) respectively inputting the cut rectangular nameplate, the triangular nameplate and the lock catch data into the corresponding VGG16 classification networks to respectively obtain the number of the rectangular nameplate, the triangular nameplate and the lock catch in a fault-free mode;
6.4) comparing the number of the rectangular nameplates, the triangular nameplates and the latches which are obtained through statistics and have no fault with the number of the rectangular nameplates, the triangular nameplates and the latches under the condition of the positive sample of the electric box cover data set, and if the number of the rectangular nameplates, the triangular nameplates and the latches is equal, judging that the sample to be detected is a positive sample, otherwise, judging that the sample to be detected is a negative sample.
3. The method for detecting the fault of the electric box cover of the subway train based on the image recognition is characterized by comprising the following steps of: in step 1), the number of positive samples is 1470, and the number of negative samples is 53.
4. The method for detecting the fault of the electric box cover of the subway train based on the image recognition is characterized by comprising the following steps of: in the step 2), the pseudo color image is 3 channels.
5. The method for detecting the fault of the electric box cover of the subway train based on the image recognition is characterized by comprising the following steps of: in the step 3), the learning rate of the Nanodet training is set to be 0.0001, and the iteration times are set to be 100 epochs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110620998.5A CN113362302B (en) | 2021-06-03 | 2021-06-03 | Fault detection method of subway train electric box cover based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110620998.5A CN113362302B (en) | 2021-06-03 | 2021-06-03 | Fault detection method of subway train electric box cover based on image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113362302A true CN113362302A (en) | 2021-09-07 |
| CN113362302B CN113362302B (en) | 2023-04-07 |
Family
ID=77531867
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110620998.5A Active CN113362302B (en) | 2021-06-03 | 2021-06-03 | Fault detection method of subway train electric box cover based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113362302B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115240122A (en) * | 2022-09-22 | 2022-10-25 | 南昌工程学院 | Air preheater area identification method based on deep reinforcement learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN110569841A (en) * | 2019-09-02 | 2019-12-13 | 中南大学 | contact gateway key component target detection method based on convolutional neural network |
| CN111080608A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Method for recognizing closing fault image of automatic brake valve plug handle of railway wagon in derailment |
| CN111950630A (en) * | 2020-08-12 | 2020-11-17 | 深圳市烨嘉为技术有限公司 | Small sample industrial product defect classification method based on two-stage transfer learning |
| CN112613508A (en) * | 2020-12-24 | 2021-04-06 | 深圳市杉川机器人有限公司 | Object identification method, device and equipment |
-
2021
- 2021-06-03 CN CN202110620998.5A patent/CN113362302B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN110569841A (en) * | 2019-09-02 | 2019-12-13 | 中南大学 | contact gateway key component target detection method based on convolutional neural network |
| CN111080608A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Method for recognizing closing fault image of automatic brake valve plug handle of railway wagon in derailment |
| CN111950630A (en) * | 2020-08-12 | 2020-11-17 | 深圳市烨嘉为技术有限公司 | Small sample industrial product defect classification method based on two-stage transfer learning |
| CN112613508A (en) * | 2020-12-24 | 2021-04-06 | 深圳市杉川机器人有限公司 | Object identification method, device and equipment |
Non-Patent Citations (6)
| Title |
|---|
| 常沛等: "基于CNN的SAR车辆目标检测", 《雷达科学与技术》, vol. 17, no. 02, 15 April 2019 (2019-04-15), pages 220 - 224 * |
| 李伟等: "基于改进CNN的宫颈细胞自动分类算法", 《计算机系统应用》, vol. 29, no. 06, 15 June 2020 (2020-06-15), pages 137 - 145 * |
| 汤踊等: "深度学习在输电线路中部件识别与缺陷检测的研究", 《电子测量技术》 * |
| 汤踊等: "深度学习在输电线路中部件识别与缺陷检测的研究", 《电子测量技术》, no. 06, 23 March 2018 (2018-03-23) * |
| 王昕钰等: "基于三级级联架构的接触网定位管开口销缺陷检测", 《仪器仪表学报》, vol. 40, no. 10, 15 October 2019 (2019-10-15), pages 1 - 17 * |
| 项宇杰等: "基于深度卷积神经网络的木材表面缺陷检测系统设计", 《系统仿真技术》, vol. 15, no. 04, 28 November 2019 (2019-11-28), pages 253 - 257 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115240122A (en) * | 2022-09-22 | 2022-10-25 | 南昌工程学院 | Air preheater area identification method based on deep reinforcement learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113362302B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106526400B (en) | The earth fault diagnostic method and device of DC600V power supply system of train | |
| CN107123114A (en) | A kind of cloth defect inspection method and device based on machine learning | |
| CN106706653A (en) | High-speed wide board detection method | |
| US20230306577A1 (en) | Cross-scale defect detection method based on deep learning | |
| CN103837087B (en) | Pantograph automatic testing method based on active shape model | |
| CN111724290B (en) | Environment-friendly equipment identification method and system based on depth layering fuzzy algorithm | |
| CN111723774A (en) | Target identification method for power transmission equipment based on unmanned aerial vehicle inspection | |
| CN113362302B (en) | Fault detection method of subway train electric box cover based on image recognition | |
| CN102157024A (en) | System and method for on-line secondary detection checking of checking data of large-sheet checking machine | |
| CN111899216A (en) | Abnormity detection method for insulator fastener of high-speed rail contact network | |
| CN113128555B (en) | Method for detecting abnormality of train brake pad part | |
| Wu et al. | An end-to-end learning method for industrial defect detection | |
| CN113962308A (en) | Aviation equipment fault prediction method | |
| CN115546223A (en) | Method and system for detecting loss of fastening bolt of equipment under train | |
| He et al. | Defect detection of printed circuit board based on improved YOLOv5 | |
| You | PCB defect detection based on generative adversarial network | |
| CN106355187A (en) | Application of visual information to electrical equipment monitoring | |
| CN112184624A (en) | Picture detection method and system based on deep learning | |
| CN114359779B (en) | Belt tearing detection method based on deep learning | |
| CN112712055B (en) | Double-path deformable CNN coal mine gateway belt conveying foreign matter monitoring method | |
| Su et al. | Detection and State Classification of Bolts Based on Faster R-CNN | |
| Wan et al. | Fault detection of air-spring devices based on GANomaly and isolated forest algorithms | |
| Li et al. | Detection of component types and track damage for high-speed railway using region-based convolutional neural networks | |
| CN113012113A (en) | Automatic detection method for bolt looseness of high-speed rail contact network power supply equipment | |
| CN113591992B (en) | Hole detection intelligent detection auxiliary system and method for gas turbine engine |
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 |