CN113128555A - Method for detecting abnormality of train brake pad part - Google Patents
Method for detecting abnormality of train brake pad part Download PDFInfo
- Publication number
- CN113128555A CN113128555A CN202110258195.XA CN202110258195A CN113128555A CN 113128555 A CN113128555 A CN 113128555A CN 202110258195 A CN202110258195 A CN 202110258195A CN 113128555 A CN113128555 A CN 113128555A
- Authority
- CN
- China
- Prior art keywords
- detection
- hook
- component
- model
- cnn
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Abstract
The invention discloses a method for detecting the abnormality of a train brake pad component, which comprises the following steps: firstly, preprocessing the acquired brake lining data by an image; secondly, loading a trained weight file by using a Yo l oV4-t i ny target detection model, positioning a hook component and detecting the abnormality of a pin component on a shutter picture, and carrying out one-step in-place detection on the pin by using the Yo l oV4-t i ny target detection model and outputting a detection result; positioning the hook component by using a Yo l oV4-t i ny target detection model, cutting out a hook component area, and detecting the hook screenshot by using an improved OC-CNN anomaly detection network; fourthly, obtaining an abnormal score through the screenshot of the OC-CNN detection hook, judging whether the hook is abnormal or not and outputting a detection result; and step five, integrating the Yo l oV4-t i ny and improving the detection result of the OC-CNN to obtain the detailed abnormal information of the whole brake pad. The Yo l oV4-ti ny and the OC-CNN form a complete brake pad abnormity detection model, and abnormity detection of the brake pad can be completed quickly, accurately and efficiently.
Description
Technical Field
The invention relates to the field of train component abnormity detection, in particular to a method for detecting abnormity of a train brake pad component.
Background
With the rapid development of the economy of China, the track mileage in China stably occupies the first place in the world. Great progress has been made in both speed and efficiency. In train maintenance, traditional manual detection still occupies most parts. In traditional fault detection, the maintainer operational environment is abominable, the train is numerous, needs a large amount of manpower and materials, and inefficiency reliability difference etc. this and present constantly automatic, intelligent time frame do not get into. The technology with high efficiency, high accuracy and low cost is urgently needed to automatically detect train faults and ensure the safe operation of trains. Thanks to the development of machine learning and deep learning, computer vision-based anomaly detection has the potential to fall to the ground.
At present, the main applications of computer vision are in the fields of face recognition, object classification and the like. For the fault diagnosis, there are three main research directions: (1) the template matching method judges whether the abnormity exists or not according to the similarity between the picture to be detected and the standard template, and the method is extremely easy to be interfered by imaging environment, stains and the like and has a harsh application range. (2) The machine learning method based on statistics is to extract the characteristics of the samples and then classify the samples according to the characteristic distribution. The key point is that the algorithm of feature extraction needs to be designed according to different targets, and the robustness is poor. Meanwhile, when the detection object is in a complex environment, the detection object is easily influenced by the environment, so that the reliability is difficult to guarantee. For small defects, the overall characteristic difference may not reach the threshold, and there is a large constraint on the application target. (3) Based on the defect detection of deep learning, the model needs a large amount of data sets to train the neural network, and needs to learn about each type of abnormality to obtain a better network weight, and the neural network of the deep learning model can effectively inhibit the influence of environments such as illumination conditions and stains on the detection result. However, in practical engineering projects, the proportion difference between the positive sample and the negative sample is large, and the negative sample is few, so that a model with a satisfactory effect cannot be trained.
The four key points that must be solved for fault diagnosis using computer vision are clearly available from the above-mentioned background: (1) the algorithm model must be capable of effectively inhibiting the interference of environmental factors, such as illumination, stains and the like, and has strong robustness. (2) The algorithm model has engineering feasibility only by training a model with satisfactory effect under the condition of a small amount of negative samples. (3) The algorithm model has the advantages of high precision, low false detection and zero missing report rate, which is the key point for replacing manual maintenance and ensuring the safe running of the train. (4) The efficiency is high, and the time of train maintenance can only be the operation idle time, needs to accomplish inspection and maintenance to whole train in this idle time.
Therefore, a method for detecting the item points accurately in a short time by an algorithm model with high detection efficiency is needed.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a method for detecting the abnormity of the train brake pad component, which can quickly complete the abnormity detection of the train brake pad component, has high accuracy, can complete the abnormity detection of the component when a negative sample is insufficient and the positive sample and the negative sample are unbalanced, and solves the problems in the background art.
In order to achieve the purpose, the invention provides the following technical scheme: a method for detecting the abnormality of train brake sheet parts comprises the following model detection steps:
firstly, preprocessing images of brake pad data acquired by data acquisition equipment to conform to a model input format;
secondly, loading a trained weight file by using a YoroV 4-tiny target detection model, positioning a hook component and detecting the abnormity of a pin component on a brake lining picture, and carrying out one-step in-place detection on the pin by using a YoroV 4-tiny target detection model and outputting a detection result;
positioning the hook component by using a YoloV4-tiny target detection model, intercepting a hook component area, and detecting the hook screenshot by using an improved OC-CNN anomaly detection network;
fourthly, obtaining an abnormal score through the screenshot of the OC-CNN detection hook, judging whether the hook is abnormal or not and outputting a detection result;
and step five, synthesizing the YoloV4-tiny and improving the detection result of the OC-CNN to obtain the detailed abnormal information of the whole brake pad.
Preferably, before the model detection, there is a model training step, and the model training step is as follows:
step S1, obtaining pictures of the train brake sheet component through data acquisition equipment;
s2, adopting Labelimg labeling software to label the acquired brake lining picture to prepare a data set in a VOC format;
s3, constructing a training set and a testing set of the brake pad abnormity detection model, and obtaining a YoloV4-tiny weight file after training; the training set comprises 1300 positive sample brake lining pictures and 938 abnormal brake lining pictures; the test set comprises 1338 brake lining pictures;
and S4, detecting a positive sample brake lining picture through a trained YoloV4-ting model, obtaining 1000 hook screenshots in a normal state through network positioning interception, then additionally adding 70 hook abnormal screenshot pictures to prepare an OC-CNN network data set, and finally finishing the training of the OC-CNN.
Preferably, the image preprocessing in the first step is to process the picture format into pil.image, and the number of channels is 3, so as to ensure that the image format conforms to the model input format.
Preferably, if the abnormality score obtained in the fourth step is higher than 0.25, the hook member is judged to be normal; if the value is less than 0.25, it is judged that the hook member is abnormal.
Preferably, the improved OC-CNN anomaly detection network refers to an OC-CNN introducing negative samples for supervised learning.
Preferably, the abnormal brake lining picture in step S3 includes pin missing, hook missing or hook deformation.
Preferably, the abnormal slice picture in step S3 is obtained through PS or data enhancement processing.
Preferably, the model detection speed in the detection method reaches 0.2 per second under the running of the CPU, and the real-time detection of the brake pad abnormity can be realized.
The invention has the beneficial effects that:
1) the invention utilizes the abnormality detection model of YoloV4-tiny + OC-CNN, can reach 0.2 piece/s under the running of CPU, can realize the real-time detection of brake pad abnormality, and has fast detection speed and high detection efficiency.
2) The invention improves the ordinary OC-CNN anomaly detection network, improves the training from only using the positive sample to the positive sample, and simultaneously utilizes a small amount of negative samples to carry out fine adjustment, thereby reducing the dependence on the negative samples, improving the feasibility of the model in engineering projects, improving the sensitivity of the network to the anomaly characteristics, reducing false alarms and simultaneously improving the accuracy and the robustness.
3) According to the invention, the One-stage target detection model YoloV4-tiny is utilized to realize rapid positioning and intercepting of the hook component and One-step in-place abnormality detection of the pin with large variation difference, the hook realizes positioning and intercepting, and the influence of other part characteristics of the picture to be detected on OC-CNN abnormality detection is effectively avoided.
4) According to the invention, the YoloV4-tiny target detection model is combined with the improved OC-CNN to form a brake pad abnormity detection model, the respective advantages of two networks are fully exerted, and the data acquisition environment is effectively eliminated as follows: compared with the traditional component fault detection based on template matching and simple feature comparison, the detection method has the advantages that the yolk V4-tiny + OC-CNN anomaly detection model has higher robustness and higher accuracy on the conditions of illumination, stain interference and the like under the actual condition, and can have a very low false alarm rate of less than 10% under the condition of no missing report. Can let the testing personnel break away from the low adverse operational environment of train, reduce maintenance personal work load. 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 flow chart of the detection according to the present invention;
FIG. 2 is a diagram of the YoloV4-tiny network structure according to the present invention;
FIG. 3 is a diagram of an improved OC-CNN network architecture according to the present invention;
FIG. 4 is a schematic diagram of normal and abnormal samples according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-4, as shown in fig. 1, fig. 1 is a flowchart illustrating a train brake lining component abnormality detection based on a yoolov 4-tiny target detection model and an improved OC-CNN abnormality detection network. The model training process is as follows:
the method comprises the steps that firstly, a picture of a train brake pad component is obtained by utilizing component picture information acquisition equipment, and an image of the brake pad component is obtained;
secondly, labeling the acquired brake lining pictures by using Labelimg labeling software to prepare a data set in a VOC format;
thirdly, 938 abnormal brake lining pictures (pin missing, hook missing and hook deformation) and 1300 normal samples are made through PS and data enhancement to serve as a training set of a brake lining abnormal model, 1338 pictures serve as a model testing set, and a YoloV4-tiny weight file is obtained after training;
fourthly, detecting a positive sample by using a trained YoloV4-ting model, obtaining 1000 hook screenshots in a normal state through network positioning interception, additionally adding 70 hook abnormal screenshot pictures on the basis to prepare an OC-CNN network data set, and finally finishing the training of the OC-CNN;
the model test procedure was as follows:
firstly, preprocessing a brake pad data acquired by an inspection robot, wherein the acquired brake pad data is a gray scale image, the number of channels is 2, the format is jpg format, and the acquired brake pad data does not conform to the YoloV4-Tiny detection model input format, and the preprocessing is to convert a picture into 3 channels through a cv2.cvtColor function and an image. from array function, namely the PIL. image format, so that the model input format is met;
secondly, loading a trained YoloV4-tiny weight file by using a YoloV4-tiny target detection model, wherein the YoloV4-tiny network structure is shown in figure 2, positioning a hook component and detecting the abnormity of a pin component on a brake lining picture, and the YoloV4-tiny target detection model detects the pin in place in one step and outputs a detection result;
thirdly, positioning the hook component by using a YoloV4-tiny target detection model, intercepting a hook component area, and detecting the hook screenshot by using an improved OC-CNN anomaly detection network, wherein the OC-CNN network structure is shown in FIG. 3;
and fourthly, judging whether the abnormity exists or not through an abnormity score obtained by detecting the hook screenshot through the OC-CNN. The hook component is normal if the anomaly score is higher than 0.25, otherwise, the hook component is abnormal;
and fifthly, synthesizing the YoloV4-tiny and improving the detection result of the OC-CNN to obtain the detailed abnormal information of the whole brake pad.
Firstly, acquiring a component image of a train to be overhauled by utilizing data acquisition equipment such as an industrial digital camera (CCD); and (3) transmitting the collected picture data to a server, detecting the picture of the brake pad component through a brake pad abnormity detection model deployed in the server, outputting and reflecting the detection result to a maintenance staff, and finally performing fixed-point maintenance on the maintenance staff according to the abnormity information.
In the top point, the abnormal characteristic of pin loss is obvious, whether the pin is abnormal or not can be accurately judged by using the target detection model, when the hook part is lost or defective, normal and abnormal samples are shown in fig. 4, the characteristics of the abnormal characteristic and the normal sample extracted by the yolo v4-tiny characteristic extraction network are in hard-to-divide areas in distribution, the difference is difficult to find a proper boundary, the target detection network cannot be used for detection in one step, and therefore the hook part needs to adopt an additional network model for detection. The method comprises the steps of firstly adopting a YoloV4-tiny target detection model to detect pins in one step to complete positioning and abnormity detection of the pins, simultaneously adopting YoloV4-tiny to position the hooks, cutting the areas to which the hooks belong, detecting pictures obtained by cutting the YoloV4-tiny by using an improved abnormity detection network OC-CNN introducing negative sample supervision and learning, and judging the abnormity condition of hook components by using abnormity score values obtained by OC-CNN detection. Compared with an OC-CNN network which only adopts a positive sample to train in an original version, the OC-CNN which introduces a negative sample to perform supervised learning has more accurate judgment on abnormal conditions, higher sensitivity and obviously improved accuracy and robustness. The yoloV4-tiny and the OC-CNN form a complete brake pad abnormity detection model, and the abnormity detection of the brake pad can be completed quickly, accurately and efficiently.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that various changes in the embodiments and/or modifications of the invention can be made, and equivalents and modifications of some features of the invention can be made without departing from the spirit and scope of the invention.
Claims (8)
1. A method for detecting the abnormity of a train brake pad component is characterized in that the model detection steps in the detection method are as follows:
firstly, preprocessing images of brake pad data acquired by data acquisition equipment to conform to a model input format;
secondly, loading a trained weight file by using a YoroV 4-tiny target detection model, positioning a hook component and detecting the abnormity of a pin component on a brake lining picture, and carrying out one-step in-place detection on the pin by using a YoroV 4-tiny target detection model and outputting a detection result;
positioning the hook component by using a YoloV4-tiny target detection model, intercepting a hook component area, and detecting the hook screenshot by using an improved OC-CNN anomaly detection network;
fourthly, obtaining an abnormal score through the screenshot of the OC-CNN detection hook, judging whether the hook is abnormal or not and outputting a detection result;
and step five, synthesizing the YoloV4-tiny and improving the detection result of the OC-CNN to obtain the detailed abnormal information of the whole brake pad.
2. The method for train brake lining component anomaly detection according to claim 1, characterized in that: before the model detection, the model training method also comprises the following steps:
step S1, obtaining pictures of the train brake sheet component through data acquisition equipment;
s2, adopting Labelimg labeling software to label the acquired brake lining picture to prepare a data set in a VOC format;
s3, constructing a training set and a testing set of the brake pad abnormity detection model, and obtaining a YoloV4-tiny weight file after training; the training set comprises 1300 positive sample brake lining pictures and 938 abnormal brake lining pictures; the test set comprises 1338 brake lining pictures;
and S4, detecting positive sample brake lining pictures through a trained YoloV4-Tiny model, obtaining 1000 hook screenshots in a normal state through network positioning interception, then additionally adding 70 hook abnormal screenshot pictures to prepare an OC-CNN network data set, and finally finishing the training of the OC-CNN.
3. The method for train brake lining component anomaly detection according to claim 1, characterized in that: the image preprocessing in the first step is to process the picture format into PIL image, the number of channels is 3, and the image is ensured to conform to the model input format.
4. The method for train brake lining component anomaly detection according to claim 1, characterized in that: if the obtained abnormal score in the fourth step is higher than 0.25, the hook component is judged to be normal; if the value is less than 0.25, it is judged that the hook member is abnormal.
5. The method for train brake lining component anomaly detection according to claim 1, characterized in that: the improved OC-CNN anomaly detection network refers to an OC-CNN which introduces negative samples for supervised learning.
6. The method for train brake lining component anomaly detection according to claim 2, characterized in that: the abnormal brake lining picture in the step S3 includes pin missing, hook missing or hook deformation.
7. The method for train brake lining component anomaly detection according to claim 2, characterized in that: the abnormal slice picture in step S3 is obtained through PS or data enhancement processing.
8. Method for train brake lining component anomaly detection according to claim 1 or 2, characterized in that: the model detection speed in the detection method reaches 0.2 per second under the running of the CPU, and the real-time detection of the brake pad abnormity can be realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110258195.XA CN113128555B (en) | 2021-03-09 | 2021-03-09 | Method for detecting abnormality of train brake pad part |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110258195.XA CN113128555B (en) | 2021-03-09 | 2021-03-09 | Method for detecting abnormality of train brake pad part |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113128555A true CN113128555A (en) | 2021-07-16 |
| CN113128555B CN113128555B (en) | 2022-05-31 |
Family
ID=76772949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110258195.XA Active CN113128555B (en) | 2021-03-09 | 2021-03-09 | Method for detecting abnormality of train brake pad part |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113128555B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113688171A (en) * | 2021-08-24 | 2021-11-23 | 北京工商大学 | Train braking fault detection method and system based on graph convolution neural network |
| CN116071423A (en) * | 2023-02-09 | 2023-05-05 | 哈尔滨市科佳通用机电股份有限公司 | Method, system and medium for positioning railway wagon brake adjuster component |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108152059A (en) * | 2017-12-20 | 2018-06-12 | 西南交通大学 | High-speed train bogie fault detection method based on Fusion |
| CN109685131A (en) * | 2018-12-20 | 2019-04-26 | 斑马网络技术有限公司 | Automobile vehicle device system exception recognition methods and device |
| CN110395402A (en) * | 2019-06-26 | 2019-11-01 | 北京晶品特装科技有限责任公司 | The low disturbance unmanned aerial vehicle ejecting frame separating mechanism of one kind and separation method |
| US20190339685A1 (en) * | 2016-05-09 | 2019-11-07 | Strong Force Iot Portfolio 2016, Llc | Methods and systems for data collection, learning, and streaming of machine signals for analytics and maintenance using the industrial internet of things |
| CN110530639A (en) * | 2019-10-08 | 2019-12-03 | 五邑大学 | A kind of bullet train axle box bearing failure diagnostic method |
| CN110929577A (en) * | 2019-10-23 | 2020-03-27 | 桂林电子科技大学 | Improved target identification method based on YOLOv3 lightweight framework |
| CN111239158A (en) * | 2020-03-13 | 2020-06-05 | 苏州鑫睿益荣信息技术有限公司 | Automobile instrument panel detection system and detection method based on machine vision |
| CN111337789A (en) * | 2019-10-23 | 2020-06-26 | 西安科技大学 | Method and system for detecting fault electrical element in high-voltage transmission line |
-
2021
- 2021-03-09 CN CN202110258195.XA patent/CN113128555B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190339685A1 (en) * | 2016-05-09 | 2019-11-07 | Strong Force Iot Portfolio 2016, Llc | Methods and systems for data collection, learning, and streaming of machine signals for analytics and maintenance using the industrial internet of things |
| CN108152059A (en) * | 2017-12-20 | 2018-06-12 | 西南交通大学 | High-speed train bogie fault detection method based on Fusion |
| CN109685131A (en) * | 2018-12-20 | 2019-04-26 | 斑马网络技术有限公司 | Automobile vehicle device system exception recognition methods and device |
| CN110395402A (en) * | 2019-06-26 | 2019-11-01 | 北京晶品特装科技有限责任公司 | The low disturbance unmanned aerial vehicle ejecting frame separating mechanism of one kind and separation method |
| CN110530639A (en) * | 2019-10-08 | 2019-12-03 | 五邑大学 | A kind of bullet train axle box bearing failure diagnostic method |
| CN110929577A (en) * | 2019-10-23 | 2020-03-27 | 桂林电子科技大学 | Improved target identification method based on YOLOv3 lightweight framework |
| CN111337789A (en) * | 2019-10-23 | 2020-06-26 | 西安科技大学 | Method and system for detecting fault electrical element in high-voltage transmission line |
| CN111239158A (en) * | 2020-03-13 | 2020-06-05 | 苏州鑫睿益荣信息技术有限公司 | Automobile instrument panel detection system and detection method based on machine vision |
Non-Patent Citations (4)
| Title |
|---|
| DYLAN EINSIDLER等: "A DEEP LEARNING APPROACH TO TARGET RECOGNITION IN SIDE-SCAN SONAR IMAGERY", 《IEEE》 * |
| YAO YANG等: "Defect Detection of Axle Box Cover Device Fixing Bolts in Metro Based on Convolutional Neural Network", 《IEEE》 * |
| 吴志丹 等: "基于EEMD的高速列车横向减振器故障的排列熵特征分析", 《计算机科学》 * |
| 张甬成: "YOLOv3算法在闸片丢失及装反检测设备中的应用", 《铁道车辆》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113688171A (en) * | 2021-08-24 | 2021-11-23 | 北京工商大学 | Train braking fault detection method and system based on graph convolution neural network |
| CN116071423A (en) * | 2023-02-09 | 2023-05-05 | 哈尔滨市科佳通用机电股份有限公司 | Method, system and medium for positioning railway wagon brake adjuster component |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113128555B (en) | 2022-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108074231B (en) | Magnetic sheet surface defect detection method based on convolutional neural network | |
| CN104535356B (en) | A kind of roller steel wire rope rope fault monitoring method and system based on machine vision | |
| CN113128555B (en) | Method for detecting abnormality of train brake pad part | |
| CN107123114A (en) | A kind of cloth defect inspection method and device based on machine learning | |
| CN111784633A (en) | Insulator defect automatic detection algorithm for power inspection video | |
| Zhou et al. | Automatic detection of industrial wire rope surface damage using deep learning-based visual perception technology | |
| CN110490842A (en) | A kind of steel strip surface defect detection method based on deep learning | |
| CN115526515B (en) | Safety monitoring system of gate for water conservancy and hydropower | |
| CN116229052B (en) | Method for detecting state change of substation equipment based on twin network | |
| CN111899216A (en) | Abnormity detection method for insulator fastener of high-speed rail contact network | |
| Wu et al. | An end-to-end learning method for industrial defect detection | |
| CN107330440B (en) | Ocean state calculation method based on image recognition | |
| Chavan et al. | Quality control of PCB using image processing | |
| You | PCB defect detection based on generative adversarial network | |
| Kähler et al. | Anomaly detection for industrial surface inspection: Application in maintenance of aircraft components | |
| CN112102280B (en) | Method for detecting loosening and loss faults of small part bearing key nut of railway wagon | |
| CN113822836A (en) | Method of marking an image | |
| CN113362302B (en) | Fault detection method of subway train electric box cover based on image recognition | |
| Wang et al. | Visual defect detection for substation equipment based on joint inspection data of camera and robot | |
| CN115082850A (en) | Template support safety risk identification method based on computer vision | |
| Xia et al. | A multi-target detection based framework for defect analysis of electrical equipment | |
| Han et al. | Damage detection of quayside crane structure based on improved faster R-CNN | |
| CN117114420B (en) | Image recognition-based industrial and trade safety accident risk management and control system and method | |
| CN112465784A (en) | Method for detecting appearance abnormity of subway clamp | |
| CN117893789A (en) | High-precision method for detecting overhead transmission line insulator based on improved Faster R-CNN algorithm |
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 |