CN111080601A - Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon - Google Patents
Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon Download PDFInfo
- Publication number
- CN111080601A CN111080601A CN201911272267.5A CN201911272267A CN111080601A CN 111080601 A CN111080601 A CN 111080601A CN 201911272267 A CN201911272267 A CN 201911272267A CN 111080601 A CN111080601 A CN 111080601A
- Authority
- CN
- China
- Prior art keywords
- fault
- image
- pull ring
- brake device
- railway wagon
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
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)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
A railway wagon derailment braking device pull ring grinding shaft fault image identification method belongs to the technical field of railway wagon braking device safety. The invention aims at the problems that the fault identification is carried out by adopting a manual vehicle inspection mode for the derailment fault of the existing railway wagon, the detection efficiency is low and the accuracy is poor. Acquiring image training samples, configuring label information for each image training sample, and forming an image training sample library; training all image training samples by adopting a Faster RCNN model based on acceptance _ resnet _ v2 to obtain pull ring fault weight coefficients; obtaining the probability that the fault prediction box determined by the image training sample belongs to the label fault; obtaining an image to be identified; and judging a suspected prediction frame which is possibly failed in the image to be identified based on the pull ring failure weight coefficient, obtaining the score of the suspected prediction frame as a shaft grinding area based on the probability, determining the pull ring shaft grinding failure when the score is greater than a preset score threshold value, and alarming. The method is used for identifying the pull ring grinding shaft fault.
Description
Technical Field
The invention relates to an image identification method for a pull ring grinding shaft fault of a railway wagon derailment braking device, belonging to the technical field of railway wagon braking device safety.
Background
In the operation process of the railway wagon, the derailment automatic braking device plays a great role in preventing the occurrence of derailment accidents, and the economic loss and the personal injury caused by the derailment accidents are effectively reduced. It is therefore of utmost importance to ensure safe operation of the derailment brake.
At present, the monitoring mode of the derailment fault of the railway wagon is manual inspection, namely, the acquired corresponding images are inspected in a manual mode, and whether the evidence indicates the occurrence of the derailment fault is judged. The method is limited by the efficiency and the precision of manual operation, the condition of missing detection and error detection of faults is very easy to occur, and once the fault pictures are detected in a missing manner, the safe operation of the railway wagon is seriously threatened.
Because derailing brake device can reliably brake when derailing fault happens, if the normal operation of derailing brake device can be ensured, dangerous accident caused by derailing can be effectively prevented. Therefore, the monitoring of the derailment accident can be realized through the monitoring of the state of the derailment brake device.
The main failures of derailing brakes include pull ring drop, which can cause shaft wear once it occurs. With the rapid development of deep learning technology in recent years, an image automatic identification mode is required to replace a traditional manual vehicle inspection mode to automatically identify the grinding marks generated by the falling of the pull ring of the derailment automatic braking device, so that the derailment accident is safely monitored, and the accuracy of the detection result is obviously improved.
Disclosure of Invention
The invention provides a railway wagon derailment braking device pull ring grinding shaft fault image identification method, which aims at the problems that fault identification is carried out by adopting a manual vehicle inspection mode for the conventional railway wagon derailment fault, the detection efficiency is low, and the accuracy is poor.
The invention discloses a method for identifying a railway wagon derailment braking device pull ring grinding shaft fault image, which comprises the following steps of:
the method comprises the following steps: acquiring a fault image of a pull ring grinding shaft of a derailing brake device to obtain image training samples, and configuring marking information for each image training sample to form an image training sample library; the marking information is used for recording a fault label value and a preliminary fault position coordinate;
step two: training all image training samples by adopting a Faster RCNN model based on acceptance _ resnet _ v2 to obtain pull ring fault weight coefficients; obtaining the probability that the fault prediction box determined by the image training sample belongs to the label fault;
step three: acquiring an image of a derailed brake device of a railway wagon in operation and preprocessing the image to obtain an image to be identified; and judging a suspected prediction frame which is possibly failed in the image to be identified based on the pull ring failure weight coefficient, obtaining the score of the suspected prediction frame as a shaft grinding area based on the probability, determining the pull ring shaft grinding failure when the score is greater than a preset score threshold value, and alarming.
According to the method for identifying the fault image of the pull ring grinding shaft of the derailment brake device of the railway wagon,
the image training sample is obtained after preprocessing of the pull ring grinding shaft fault image.
According to the method for identifying the pull ring grinding shaft fault image of the railway wagon derailment braking device, the preprocessing sequentially comprises data enhancement processing and median filtering processing of the pull ring grinding shaft fault image and improvement of image contrast.
According to the method for identifying the grinding fault image of the pull ring grinding shaft of the derailment brake device of the railway wagon, the data enhancement processing comprises image rotation, random cutting, horizontal turning, vertical turning, stretching and scaling.
According to the method for identifying the pull ring grinding shaft fault image of the derailment brake device of the railway wagon, the process of training all image training samples by adopting a Faster RCNN model based on acceptance _ resnet _ v2 comprises the following steps:
and (4) performing feature extraction on the image training sample by adopting the convolutional layer to obtain a feature map.
According to the method for identifying the railway wagon derailment brake device pull ring grinding shaft fault image, the RPN layer is adopted to generate an anchor frame for the characteristic diagram, and on one hand, the anchor frame is judged to belong to a foreground or a background by adopting a softmax loss function; on the other hand, the anchor frame is modified by bounding box regression; further obtaining a fault prediction frame determined as a foreground; and performing border-crossing elimination and overlapping elimination on the fault prediction frames, sorting the fault prediction frames with the mutual coverage rate not exceeding a set IoU threshold from small to large, and reserving the first N fault prediction frames.
According to the method for identifying the fault image of the pull ring grinding shaft of the railway wagon derailment brake device, the N fault prediction frames are processed by the Roi pooling layer, and the recommended characteristic diagram with a fixed size is obtained.
According to the method for identifying the image of the grinding fault of the pull ring shaft of the derailment brake device of the rail wagon, a label fault is determined by using a softmax loss function by adopting a full connection layer for the suggested characteristic diagram, and the position of an actual fault area is determined; and meanwhile, the probability that the recommended feature graph belongs to the label fault is obtained.
According to the method for identifying the pull ring grinding fault image of the railway wagon derailment braking device, in the third step, an error area of a suspected prediction frame is removed according to the score of the grinding area, and a final grinding area is determined.
The invention has the beneficial effects that: according to the method, a deep learning model is built by utilizing a convolutional neural network according to the condition that a circle of grinding marks are generated on an axle after a pull ring of a derailing brake device is pulled off, and a weight coefficient is obtained by training by adopting an image training sample. And automatically detecting an image acquired by the truck in actual operation based on the weight coefficient so as to determine the axle grinding fault caused by the dropping of a pull ring of the derailment automatic braking device. When the detection result is that a fault occurs, the specific fault information can be uploaded to the vehicle detection platform.
According to the method, the deep learning model is adopted to automatically detect the derailment braking device pull ring grinding shaft fault, and manual operation is replaced by artificial intelligence, so that the efficiency of train inspection operation can be improved, and the fault detection accuracy rate is improved; the method is beneficial to timely finding out the running fault of the freight car and ensuring the safety of railway freight.
Drawings
FIG. 1 is a flow chart of the method for identifying the fault image of the pull ring grinding shaft of the derailment brake device of the railway wagon.
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.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The invention is further described with reference to the following drawings and specific examples, which are not intended to be limiting.
In a first specific embodiment, as shown in fig. 1, the invention provides a method for identifying a railway wagon derailment brake apparatus pull ring grinding shaft fault image, which is characterized by comprising the following steps:
the method comprises the following steps: acquiring a fault image of a pull ring grinding shaft of a derailing brake device to obtain image training samples, and configuring marking information for each image training sample to form an image training sample library; the marking information is used for recording a fault label value and a preliminary fault position coordinate;
step two: training all image training samples by adopting a Faster RCNN model based on acceptance _ resnet _ v2 to obtain pull ring fault weight coefficients; obtaining the probability that the fault prediction box determined by the image training sample belongs to the label fault;
step three: acquiring an image of a derailed brake device of a railway wagon in operation and preprocessing the image to obtain an image to be identified; and judging a suspected prediction frame which is possibly failed in the image to be identified based on the pull ring failure weight coefficient, obtaining the score of the suspected prediction frame as a shaft grinding area based on the probability, determining the pull ring shaft grinding failure when the score is greater than a preset score threshold value, and alarming.
In the embodiment, the acquisition of the image can be realized by the high-speed camera arrays arranged on two sides of the rail, the visual information of the derailment braking device is shot by the high-speed cameras, and the high-definition image is output.
In the embodiment, the fast RCNN model is built by utilizing a convolutional neural network, and can be used for automatically identifying the grinding shaft fault caused by the falling of the pull ring of the derailment automatic braking device after the training of the image training sample, so that the stability of the identification algorithm is improved on the basis of ensuring the identification precision of the algorithm.
For collected derailing brake device pull ring grinding shaft fault images, the collected vehicle shaft images have obvious difference in brightness and contrast due to the fact that the vehicle shaft is influenced by vehicle body color, new and old vehicles, rain stain interference and pollution factors of goods pulled by a carriage. In addition, images acquired by different hardware acquisition devices are different, and the detection results of the deep learning model can be obviously influenced. Therefore, when the image training sample library is established, in order to ensure the integrity of the sample library, the axle grinding trace images in various states need to be collected as many as possible.
And the marking information is used for marking the fault position of the preprocessed image training sample and establishing a VOC data set. Each image training sample can correspond to an xml file of the marking information, and the xml file stores the label value of the fault mark and the fault coordinate position information.
Further, the image training sample is obtained after preprocessing of the pull ring grinding shaft fault image.
And further, the preprocessing sequentially comprises data enhancement processing and median filtering processing of the split ring grinding shaft fault image and improvement of image contrast.
Since the pull ring grinding shaft fault is susceptible to noise interference, the images in the data set can be subjected to median filtering processing in order to resist interference from noise in the images. The median filtering is a nonlinear filter, and can effectively protect the edge information of the image and improve the anti-noise interference performance of the model while reducing the noise of the image.
The data enhancement processing includes, by way of example, image rotation, random cropping, horizontal flipping, vertical flipping, stretching, and scaling. Subject to objective factors, collected derailment braking device pull ring grinding shaft fault images cannot contain all conditions, so data enhancement processing is carried out on the grinding shaft fault images. After the grinding shaft fault image is amplified, the robustness of the data set can be enhanced.
Still further, the process of training all image training samples by using the Faster RCNN model based on the acceptance _ resnet _ v2 includes:
and (4) performing feature extraction on the image training sample by adopting the convolutional layer to obtain a feature map.
The fast RCNN model mainly comprises a convolutional layer, an RPN network, a Roi pooling layer and a full-link layer.
The feature map (feature maps) of the input image is extracted by using a set of basic convolution connected with a relu activation function and then connected with a pooling layer structure, and the feature maps (feature maps) are used for a subsequent RPN layer and a full connection layer.
The fast RCNN supports the input of images with any size, sets the normalized size of the images before entering the network, scales the images to a fixed size, and then sends the scaled images to the network. The convolutional layer comprises 13 conv layers, 13 relu layers and 4 pooling layers.
For 13 conv layers: kernel _ size is 3, pad is 1, stride is 1; convolution formula:
the conv layer does not change the image size; namely: the input image size is equal to the output image size;
for 13 relu layers: setting an activation function without changing the image size;
4 pooling layers: kernel _ size 2, stride 2; the pooling layer will let the output picture be 1/2 of the input picture.
Further, generating an anchor frame by adopting an RPN layer for the feature map, and judging whether the anchor frame belongs to the foreground or the background by adopting a softmax loss function on one hand; on the other hand, the anchor frame is modified by bounding box regression; further obtaining a fault prediction frame determined as a foreground; and performing border-crossing elimination and overlapping elimination on the fault prediction frames, sorting the fault prediction frames with the mutual coverage rate not exceeding a set IoU threshold from small to large, and reserving the first N fault prediction frames.
For the RPN layer: the rpn (region pro-posals) network is mainly used for generating region suggestions (region pro-posals), and first generates a stack of Anchor boxes (Anchor boxes), and after performing clipping and filtering on the Anchor boxes, judges whether Anchor points (anchors) belong to foreground (forego) or background (background) through a softmax loss function, that is, whether objects to be detected or not are objects to be detected, so that the process is a binary classification process. Meanwhile, the other branch corrects the Anchor box (Anchor box) by using bounding box regression (bounding box regression), and forms a more accurate suggestion (proposal).
After entering the RPN, the feature maps (feature maps) are first subjected to a 3 × 3 convolution, so as to further concentrate the feature information, followed by two full convolution structures, kernel _ size 1, p 0, and stride 1. Rpn _ cls and rpn _ bbox are obtained respectively, and are used for carrying out secondary classification on the Anchor box (Anchor box) and obtaining coordinate information of the Anchor box. After a more accurate prediction frame is obtained, the prediction frame is further subjected to border crossing elimination and nms (non-maximum suppression) utilization, and overlapped frames are eliminated. And (3) retaining coarse screening of which the coverage rate does not exceed a set IoU (interaction over Union) threshold, and then taking the first N boxes, so that the number of the prediction frames is only N when the next layer of Roi pooling is carried out.
And further, processing the N fault prediction frames by adopting a Roi pooling layer to obtain a fixed-size recommended feature map.
The Roi pooling layer obtains feature maps (feature maps) by using the prousals generated by the RPN and the interception _ resnet _ v2, obtains a fixed-size suggested feature map (feature map), and can perform target identification and positioning by using a full connection operation when entering the future. By designing the model based on the acceptance _ resnet _ v2 network structure, overfitting can be prevented. The network adopts a gradient descent optimization algorithm, and the training error is always reduced along with the increase of the network depth.
The allowance _ resnet _ v2 is the structural design of the allowance plus residual network. The inclusion network does not need to decide manually which filter to use or if pooling is needed, but rather the network determines these parameters itself, it can add all possible values of these parameters to the network, and then connect these outputs to let the network itself learn what parameters it needs and which filter combinations to use. The softmax branch is added in the network, and because even hidden units and intermediate layers also participate in feature calculation, the hidden units and the intermediate layers can predict the classification of pictures, so that the hidden units and the intermediate layers play a role in adjustment in the inclusion network and prevent overfitting. The residual network is a stack of residual blocks, so that the network can be designed to be deep. The difference between the residual network and the common network is that before nonlinear change, the data is copied and then accumulated with the appointed data, and nonlinear transformation is performed. For a common convolutional network, a gradient descent and other common optimization algorithms are used, as the depth of the network increases, the training error tends to decrease first and then increase, and the expected ideal result is that the training error gradually decreases as the depth of the network increases. The Resnet training error decreases with increasing network depth.
Further, determining a label fault by using a softmax loss function for the suggested characteristic diagram by using a full connection layer, and determining the position of an actual fault area; and meanwhile, the probability that the recommended feature graph belongs to the label fault is obtained.
At the full link layer, the classifier performs full link operation on a feature map (feature map) with a fixed size formed by the Roi pooling layer, performs classification of specific categories by using a softmax Loss function, and simultaneously performs bounding box regression (bounding box regression) operation by using L1 Loss to obtain an accurate position of the object.
That is, after the generic feature maps are obtained from the PoI Pooling, the functions mainly realized by the full connection layer include:
classifying the specific categories of the region explosals through full connection and softmax;
and carrying out bounding box regression on the region explosals again to obtain a prediction frame with higher precision.
The model uses the softmax loss function in classification, the softmax function can output the probability of selecting the classification according to the relative size of the softmax loss function and the softmax loss functionThe sum of the probabilities is 1. The formula for Softmax is:
Sirepresented is the output of the ith neuron.
The model is classified by using a softmax loss function, the softmax function can output the probability of selecting the category according to the relative size of the softmax loss function, and the probability sum is 1. The formula for Softmax is:
Sirepresented is the output of the ith neuron.
And further, in the third step, removing an error area of the suspected prediction frame according to the score of the axis grinding area, and determining a final axis grinding area.
The method comprises the steps of obtaining a weight coefficient after training of a Faster RCNN model, predicting a region where a grinding shaft fault possibly occurs by using the obtained weight coefficient after preprocessing such as cutting, gray level stretching and median filtering of the position of a vehicle shaft in an acquired rail braking device image, and removing a misfound region by limiting scores of a prediction frame.
Although the invention herein has been described with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present invention. It is therefore to be understood that numerous modifications may be made to the illustrative embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that features described in different dependent claims and herein may be combined in ways different from those described in the original claims. It is also to be understood that features described in connection with individual embodiments may be used in other described embodiments.
Claims (9)
1. A railway wagon derailment brake device pull ring grinding shaft fault image identification method is characterized by comprising the following steps:
the method comprises the following steps: acquiring a fault image of a pull ring grinding shaft of a derailing brake device to obtain image training samples, and configuring marking information for each image training sample to form an image training sample library; the marking information is used for recording a fault label value and a preliminary fault position coordinate;
step two: training all image training samples by adopting a Faster RCNN model based on acceptance _ resnet _ v2 to obtain pull ring fault weight coefficients; obtaining the probability that the fault prediction box determined by the image training sample belongs to the label fault;
step three: acquiring an image of a derailed brake device of a railway wagon in operation and preprocessing the image to obtain an image to be identified; and judging a suspected prediction frame which is possibly failed in the image to be identified based on the pull ring failure weight coefficient, obtaining the score of the suspected prediction frame as a shaft grinding area based on the probability, determining the pull ring shaft grinding failure when the score is greater than a preset score threshold value, and alarming.
2. The method for image recognition of a railway wagon derailment brake device pull ring grinding shaft fault according to claim 1,
the image training sample is obtained after preprocessing of the pull ring grinding shaft fault image.
3. The method for identifying the fault image of the pull ring grinding shaft of the railway wagon derailment brake device according to claim 2, wherein the preprocessing comprises data enhancement processing, median filtering processing and image contrast improvement of the fault image of the pull ring grinding shaft in sequence.
4. The method for image recognition of the rail wagon derailment brake device pull ring grinding shaft fault according to claim 3, wherein the data enhancement processing comprises image rotation, random cropping, horizontal turning, vertical turning, stretching and scaling.
5. The method for image recognition of railway wagon derailment brake device pull ring grinding shaft failure according to claim 4, wherein the training of all image training samples by using the Faster RCNN model based on initiation _ resnet _ v2 comprises:
and (4) performing feature extraction on the image training sample by adopting the convolutional layer to obtain a feature map.
6. The method for identifying the railway wagon derailment brake device pull ring grinding shaft fault image as claimed in claim 5, wherein for the characteristic diagram, an RPN layer is adopted to generate an anchor frame, and on one hand, the anchor frame is judged to belong to a foreground or a background by a softmax loss function; on the other hand, the anchor frame is modified by bounding box regression; further obtaining a fault prediction frame determined as a foreground; and performing border-crossing elimination and overlapping elimination on the fault prediction frames, sorting the fault prediction frames with the mutual coverage rate not exceeding a set IoU threshold from small to large, and reserving the first N fault prediction frames.
7. The method for identifying the railway wagon derailment brake device pull ring grinding shaft fault image as claimed in claim 6, wherein the N fault prediction frames are processed by a Roi pooling layer to obtain a fixed-size suggested feature map.
8. The method for identifying the railway wagon derailment brake device pull ring grinding shaft fault image as claimed in claim 7, wherein a label fault is determined by using a softmax loss function for a full connection layer for the recommended feature map, and an actual fault area position is determined; and meanwhile, the probability that the recommended feature graph belongs to the label fault is obtained.
9. The method for image recognition of axle grinding fault of pull ring of the railway wagon derailment brake device according to claim 8, wherein in the third step, an error area of a suspected prediction frame is removed according to the score of the axle grinding area, and a final axle grinding area is determined.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911272267.5A CN111080601A (en) | 2019-12-12 | 2019-12-12 | Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911272267.5A CN111080601A (en) | 2019-12-12 | 2019-12-12 | Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111080601A true CN111080601A (en) | 2020-04-28 |
Family
ID=70314110
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911272267.5A Pending CN111080601A (en) | 2019-12-12 | 2019-12-12 | Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111080601A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111652296A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based rail wagon lower pull rod fracture fault detection method |
| CN111652294A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method and system for identifying falling fault of folding type manual brake shaft of railway wagon |
| CN111652094A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Railway flat car end plate falling fault image identification method based on deep learning |
| CN112613560A (en) * | 2020-12-24 | 2021-04-06 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying front opening and closing damage fault of railway bullet train head cover based on Faster R-CNN |
| CN112733747A (en) * | 2021-01-14 | 2021-04-30 | 哈尔滨市科佳通用机电股份有限公司 | Identification method, system and device for relieving falling fault of valve pull rod |
| CN115973125A (en) * | 2023-02-15 | 2023-04-18 | 慧铁科技有限公司 | Method for processing fault of automatic derailment braking device of railway wagon |
| CN116091652A (en) * | 2023-01-20 | 2023-05-09 | 哈尔滨市科佳通用机电股份有限公司 | Image generation and detection method for pull ring falling fault of derailment automatic device |
| CN116246114A (en) * | 2023-03-14 | 2023-06-09 | 哈尔滨市科佳通用机电股份有限公司 | Method and device for detecting pull ring falling image abnormality of self-supervision derailment automatic device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106226050A (en) * | 2016-07-15 | 2016-12-14 | 北京航空航天大学 | A kind of TFDS fault automatic identifying method |
| CN106778740A (en) * | 2016-12-06 | 2017-05-31 | 北京航空航天大学 | A kind of TFDS non-faulting image detecting methods based on deep learning |
| CN109800778A (en) * | 2018-12-03 | 2019-05-24 | 浙江工业大学 | A kind of Faster RCNN object detection method for dividing sample to excavate based on hardly possible |
-
2019
- 2019-12-12 CN CN201911272267.5A patent/CN111080601A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106226050A (en) * | 2016-07-15 | 2016-12-14 | 北京航空航天大学 | A kind of TFDS fault automatic identifying method |
| CN106778740A (en) * | 2016-12-06 | 2017-05-31 | 北京航空航天大学 | A kind of TFDS non-faulting image detecting methods based on deep learning |
| CN109800778A (en) * | 2018-12-03 | 2019-05-24 | 浙江工业大学 | A kind of Faster RCNN object detection method for dividing sample to excavate based on hardly possible |
Non-Patent Citations (4)
| Title |
|---|
| ALVARO ARCOS-GARCIA ET AL: "Evaluation of deep neural networks for traffic sign detection systems", 《NEUROCOMPUTING》 * |
| CHRISTIAN SZEGEDY ET AL: "Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning", 《ARXIV:1602.07261V2 [CS.CV]》 * |
| ZHEN ZHANG ET AL: "Comparison of multiple feature extractors on Faster RCNN for breast tumor detection", 《2019 8TH INTERNATIONAL SYMPOSIUM ON NEXT GENERATION ELECTRONICS (ISNE)》 * |
| 张江勇: "基于深度学习的动车关键部位故障图像检测", 《万方数据知识服务平台》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111652296A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based rail wagon lower pull rod fracture fault detection method |
| CN111652294A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method and system for identifying falling fault of folding type manual brake shaft of railway wagon |
| CN111652094A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Railway flat car end plate falling fault image identification method based on deep learning |
| CN111652294B (en) * | 2020-05-21 | 2021-05-07 | 哈尔滨市科佳通用机电股份有限公司 | Method and system for identifying falling fault of folding type manual brake shaft of railway wagon |
| CN112613560A (en) * | 2020-12-24 | 2021-04-06 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying front opening and closing damage fault of railway bullet train head cover based on Faster R-CNN |
| CN112733747A (en) * | 2021-01-14 | 2021-04-30 | 哈尔滨市科佳通用机电股份有限公司 | Identification method, system and device for relieving falling fault of valve pull rod |
| CN116091652A (en) * | 2023-01-20 | 2023-05-09 | 哈尔滨市科佳通用机电股份有限公司 | Image generation and detection method for pull ring falling fault of derailment automatic device |
| CN115973125A (en) * | 2023-02-15 | 2023-04-18 | 慧铁科技有限公司 | Method for processing fault of automatic derailment braking device of railway wagon |
| CN116246114A (en) * | 2023-03-14 | 2023-06-09 | 哈尔滨市科佳通用机电股份有限公司 | Method and device for detecting pull ring falling image abnormality of self-supervision derailment automatic device |
| CN116246114B (en) * | 2023-03-14 | 2023-10-10 | 哈尔滨市科佳通用机电股份有限公司 | Method and device for detecting pull ring falling image abnormality of self-supervision derailment automatic device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111080601A (en) | Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon | |
| CN111080598B (en) | Bolt and nut missing detection method for coupler yoke key safety crane | |
| CN111652227B (en) | Method for detecting damage fault of bottom floor of railway wagon | |
| CN111652296A (en) | Deep learning-based rail wagon lower pull rod fracture fault detection method | |
| CN111079747A (en) | Railway wagon bogie side frame fracture fault image identification method | |
| CN111079746B (en) | Railway wagon axle box spring fault image identification method | |
| CN115272334A (en) | Method for detecting micro defects on surface of steel rail under complex background | |
| CN111080617B (en) | Railway wagon brake beam pillar round pin loss fault identification method | |
| CN107273802B (en) | Method and device for detecting fault of brake shoe drill rod ring of railway train | |
| CN110047070B (en) | Method and system for identifying rail wear degree | |
| CN111080608A (en) | Method for recognizing closing fault image of automatic brake valve plug handle of railway wagon in derailment | |
| CN111080607B (en) | Rolling bearing oil slinging fault detection method based on image recognition | |
| CN111080600A (en) | Fault identification method for split pin on spring supporting plate of railway wagon | |
| CN111091543B (en) | Railway wagon swing bolster spring loss fault target detection method | |
| CN111079821A (en) | Derailment automatic braking pull ring falling fault image identification method | |
| CN110223266A (en) | A kind of Railway wheelset tread damage method for diagnosing faults based on depth convolutional neural networks | |
| CN111079629A (en) | Railway wagon center plate out-of-position fault image identification method | |
| CN111080612B (en) | Truck bearing damage detection method | |
| CN111626170A (en) | Image identification method for railway slope rockfall invasion limit detection | |
| CN111680613A (en) | Method for detecting falling behavior of escalator passengers in real time | |
| Zhao et al. | Image-based comprehensive maintenance and inspection method for bridges using deep learning | |
| CN113553916B (en) | Orbit dangerous area obstacle detection method based on convolutional neural network | |
| CN118506299B (en) | Rail train body abnormality detection method, device and storage medium | |
| CN115631146A (en) | Image-based pantograph carbon slide strip defect detection method | |
| CN112712552A (en) | Fault detection method for vehicle tread scratch |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200428 |