CN115527018A - Fault identification method and device for parts formed by lower lock pins of railway wagon - Google Patents
Fault identification method and device for parts formed by lower lock pins of railway wagon Download PDFInfo
- Publication number
- CN115527018A CN115527018A CN202211313616.5A CN202211313616A CN115527018A CN 115527018 A CN115527018 A CN 115527018A CN 202211313616 A CN202211313616 A CN 202211313616A CN 115527018 A CN115527018 A CN 115527018A
- Authority
- CN
- China
- Prior art keywords
- image
- fault
- feature
- lock pin
- lower lock
- 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
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000001514 detection method Methods 0.000 claims abstract description 23
- 230000004927 fusion Effects 0.000 claims description 21
- 238000004590 computer program Methods 0.000 claims description 18
- 238000003860 storage Methods 0.000 claims description 13
- 238000007500 overflow downdraw method Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000007499 fusion processing Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 7
- 238000013135 deep learning Methods 0.000 abstract description 6
- 238000003384 imaging method Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 12
- 238000012549 training Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000011176 pooling Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 2
- KJLPSBMDOIVXSN-UHFFFAOYSA-N 4-[4-[2-[4-(3,4-dicarboxyphenoxy)phenyl]propan-2-yl]phenoxy]phthalic acid Chemical compound C=1C=C(OC=2C=C(C(C(O)=O)=CC=2)C(O)=O)C=CC=1C(C)(C)C(C=C1)=CC=C1OC1=CC=C(C(O)=O)C(C(O)=O)=C1 KJLPSBMDOIVXSN-UHFFFAOYSA-N 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
-
- 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
-
- 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
- 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
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
A fault identification method and device for accessories formed by a lower lock pin of a railway wagon relate to the technical field of railway wagon detection. The invention aims to solve the problems of false detection and high omission factor of the traditional method for manually detecting whether the accessories formed by the lower lock pins of the vehicle are lost. According to the invention, high-definition imaging equipment is respectively built around the rail of the truck, and the truck acquires a high-definition image after passing through the equipment. The method comprises the steps of firstly classifying and positioning the car coupler by adopting a deep learning network to obtain position information of the car coupler, and intercepting a lower lockpin to form a subgraph of an accessory according to position prior knowledge of the accessory part formed by the lower lockpin on different car couplers. And detecting the loss fault of the lower lock pin assembly component by using a deep learning network. And uploading and alarming the coupler with the loss of the accessory formed by the lower lock pin, and performing corresponding processing by the staff according to the recognition result to ensure the safe operation of the train.
Description
Technical Field
The invention belongs to the technical field of railway wagon detection, and particularly relates to detection of a fitting formed by a lower lock pin of a railway wagon.
Background
The railway freight car lower lock pin is formed into an accessory and plays an important role in firmly connecting a front car and a rear car by a car coupler. When a truck runs, the railway transportation frequency is high, the abrasion of parts is large, the condition that the lower lockpin assembly is lost is caused, and then the separation accident of the car coupler is caused, so that great potential safety hazards are brought to the running safety. Therefore, it is important for driving safety to check whether the lower lock pin assembly is complete.
The traditional manual detection mode causes the appearance of missed detection and false detection and influences the driving safety because the conditions such as fatigue and omission easily occur to vehicle detection personnel in the working process.
Disclosure of Invention
The invention provides a fault identification method and device for a railway wagon lower lock pin assembly part, aiming at solving the problems of false detection and high missing rate of the traditional method for manually detecting whether the vehicle lower lock pin assembly part is lost.
A fault identification method for a railway wagon lower lock pin assembly accessory specifically comprises the following steps:
acquiring an image of an accessory part formed by a lock pin assembly under a truck to be detected as a detected image, and inputting the detected image into a YOLOv5-SPD network to obtain an identification result;
the YOLOv5-SPD network adopts a non-step convolution network to extract a characteristic subgraph and adopts a method of weighting and fusing target frames to identify faults in the characteristic subgraph.
Further, a feature map is selected from the detected image by adopting a dynamic anchor frame, and the feature map is input into the Yolov5-SPD network.
Further, the dynamic anchor frame comprises an anchor improvement module, a bounded feature fusion module and a target detection module,
the anchor improving module is used for generating candidate frames and screening out the anchor frames which accord with the characteristic size from the candidate frames;
the bounded feature fusion module is used for fusing the edge feature of the anchor frame with the global feature to obtain a fusion feature;
and the target detection module is used for selecting a feature map in the detected image according to the fusion features.
Further, the fusion process of the bounded feature fusion module is as follows:
wherein, f is a group of compounds represented by,representing a feature extraction function, M edge For edge features, M (i ) In order to be a global feature,it is indicated that the fusion is performed,represents the fused features after fusion of i features, i being the number of features fused, i =1, 2.
Furthermore, high-definition image acquisition equipment is built beside the rail of the truck to be measured, and the high-definition image acquisition equipment is used for acquiring high-definition images of the truck to be measured.
Further, the high-definition image is a grayscale image.
Further, when a fault exists in the detected image, obtaining the coordinate of the fault position and judging whether the coordinate belongs to a preset range, if so, sending out a fault alarm, otherwise, not sending out the alarm.
The utility model provides a railway freight car lower lockpin makes up accessory trouble recognition device, includes specifically that the following unit:
an acquisition unit: used for acquiring an image of the accessory part formed by the lock pin assembly under the tested truck as a tested image,
an identification unit: the system is used for inputting a detected image into a YOLOv5-SPD network to obtain an identification result;
the YOLOv5-SPD network adopts a non-step convolution network to extract a characteristic subgraph and adopts a method of weighting and fusing target frames to identify faults in the characteristic subgraph.
A computer-readable storage medium stores a computer program that, when executed, implements a method of identifying a fitting fault such as a rail wagon lower latch assembly.
Electronic equipment comprises a storage medium, a processor and a computer program which is stored in the storage medium and can run on the processor, wherein the processor executes the computer program to realize the locking pin assembly of the rail wagon, namely a rail wagon, and the accessory fault identification method. The invention has the following beneficial effects:
1. and the automatic image identification mode is used for replacing manual detection, so that the detection efficiency and accuracy are improved.
2. According to the TFDS, the coupler type and the coupler position provided by the frame are automatically identified, the position of a lower lock pin forming part is accurately positioned, and an accurate sub-picture range is provided for detecting a rear fault.
3. Histogram equalization is selected to enhance the image, and the generalization capability of the training model is improved.
4. The improved YOLOv5 network is used for composing the lost fault of the accessory by the lower lock pin, the improved YOLOv5 uses the dynamic anchor frame as input in the input stage, the larger formula in the regression training process of the bounding box is reduced, and the difficulty of model optimization is reduced.
5. The modified YOLOv5 network is utilized to group lock-down pins into an assembly loss fault, and the modified YOLOv5 uses a target box weighted fusion method as a method of a final predicted bounding box. The bounding box with smaller score but containing certain characteristic information is reserved, so that all the characteristic information is ensured, and the condition that the bounding box with high score is directly selected and part of the information is lost is avoided.
6. Aiming at an improved YOLOv5 network, the SPD Conv is used for replacing a traditional CNN and a pooling layer in a backbone network, no information is skipped for complex down-lock pin accessory images, fine-grained information and characteristics can be captured well, and the accuracy of fault detection is improved. The lower lock pin with a complex structure forms an accessory image, and the lower lock pin has higher robustness.
Drawings
FIG. 1 is a flow diagram of dynamic anchor block generation;
FIG. 2 is a schematic diagram of SPD-Conv when the variation factor is taken to be 2;
fig. 3 is a flow chart of fault identification.
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 of the present invention may be combined with each other without conflict.
The detection efficiency and stability can be improved by adopting an automatic image identification mode. In recent years, deep learning and artificial intelligence are continuously developed, and the technology is continuously mature. Therefore, the method adopts deep learning to perform the accessory loss fault identification by locking pin descending, and can effectively improve the detection accuracy. The method comprises the following specific steps:
the first embodiment is as follows: the method for identifying the fault of the accessory formed by the lower lock pin of the rail wagon in the embodiment specifically comprises the following steps:
acquiring an image of an accessory part formed by a lock pin under a truck to be detected as a detected image, and inputting the detected image into a YOLOv5-SPD network to obtain an identification result;
the YOLOv5-SPD network adopts a non-step convolution network to extract a characteristic subgraph and adopts a method of weighting and fusing target frames to identify faults in the characteristic subgraph.
The lower lock pin is complex in component part structure, multiple in component parts and compact in structure. The method comprises the steps that a lower lockpin is assembled to detect the loss fault of an accessory, the traditional YOLOv5 mainly utilizes CNN and a pooling layer to extract features, fine-grained information is easy to ignore, and an improved YOLOv5-SPD network is formed by combining an SPD Conv network and the YOLOv5 aiming at low resolution and small objects to train a training sample.
The YOLOv5-SPD uses non-stride convolution to replace the traditional CNN and pooling layer, and aiming at complex lower lock pin accessory images, no information is skipped, fine-grained information and characteristics can be captured well, and the accuracy of fault detection is improved. Where the SPD layer downsamples the feature map but retains all information in the channel dimensions so no information is lost.
The original fused feature map is S cp Size is S 1 ×S 2 ×C 1 In which S is 1 Is the height, S, of the image 2 Is the width of the image, C 1 Is the number of channels. The sub-feature map mapping relationship is as follows:
where scale is the variation factor. Giving a fused feature map S cp Sub feature map f x,y By all S cp (i, j) wherein i + x, i + y are each divisible by a varying factor. Each sub-feature map is passed through S cp And down-sampling the change factor. Fig. 2 is a SPD-Conv schematic diagram of scale = 2. When scale =2, four sub-feature maps f are obtained 0,0 ,f 1,0 ,f 0,1 ,f 1,1 Each sub-feature map has a size ofAnd is provided withTwo downsamplings are performed.
These sub-feature maps are then connected along the channel dimension, thereby obtaining a feature map X' having a spatial dimension reduced by scale and a spatial dimension reduced by scale 2 Increased channel dimensions. That is, the SPD will characterize graph X (S, S, C) 1 ) Conversion into intermediate feature mapsAfter SPD, add non-step convolution C 2 A filter, at this time C 2 <scale 2 C 1 The characteristic diagram is further changed intoThe reason we use non-strided convolutions is to preserve as much as possible of all the descriptive feature information. Otherwise, we would "downscale" the profile using a conventional stride =3,3 × 3 filter, but only sample once per pixel. stride =2 asymmetric sampling will occur in the case where even and odd rows or columns are to be sampled at different times.
YOLOv5-SPD replaces the stride convolution with stride 2 in YOLOv5 by SPD-Conv. Where there are 5 substitutions in the Backbone moiety; the portion of the neck (the connection network of the backbone network and the detector head) is replaced by 2, while after each step convolution in YOLOv5 there is a full link layer, and the YOLOv5-SPD continues to maintain a full link layer between SPD (Space-to-depth) and Conv (convolutional layer). The learning of fine-grained information and characteristics is focused, and the accuracy of fault detection is improved.
Improved YOLOv5 is a method of final predicted bounding box using a target box weighted fusion method. The method mainly comprises the steps of sequencing confidence degrees of all prediction frames of a model from top to bottom and placing the confidence degrees into a list, iteratively matching the predicted boundary frames with real boundary frames marked by image targets in a data set, when an IOU (interaction over Unit, a standard for measuring the accuracy of detecting corresponding objects in a specific data set) is larger than a set threshold value of 0.6, considering that matching is successful, placing the predicted boundary frames which are not successfully matched into a new iteration process to continue matching judgment processing, and if matching is successful, calculating coordinates and confidence degrees of the boundary frames. And fusing the predicted boundary frame with all the real boundary frames successfully matched with the predicted boundary frame to obtain the average confidence coefficient of the predicted boundary frame, wherein the coordinate is the weighted sum of the coordinates of all the boundary frames forming the predicted boundary frame, and the weight is the confidence coefficient proportion of the corresponding boundary frame. The bounding boxes that are incorrectly predicted by the model are corrected. The regression frame of the improved model trained by the YOLOv5 network is more appropriate to the real situation, and the accuracy of fault detection can be improved.
In the embodiment, in the improved YOLOv5-SPD network, a dynamic anchor frame is selected as an input, the backbone network uses a non-stride convolutional network to replace a traditional CNN layer and a traditional pooling layer for extracting features, and a method of weighting and fusing target frames is selected to replace a traditional non-maximum suppression (NMS) method for final prediction.
The second embodiment is as follows: the sizes of subgraphs obtained by different coupler types, different stations and different cameras are not unique, the sizes of lower lock pin rented accessories are also changed, the traditional generation mode of the YOLOv5 anchor frame easily causes that small targets are divided into the anchor frames which do not accord with the sizes of the targets, so that great loss is caused in the training process, and meanwhile, the anchor frames are used as input in the model training stage, so that the setting error is large, and the speed and the difficulty of model convergence are increased. In the embodiment, in order to improve the accuracy and speed of model training, a dynamic anchor frame is adopted to select a feature map from a detected image, and the feature map is input into a YOLOv5-SPD network.
The third concrete implementation mode: in this embodiment, the method for identifying a fault of a railway wagon lower lockpin assembly is further described, in which the dynamic anchor frame includes an anchor improvement module, a bounded feature fusion module and a target detection module,
the anchor improving module is used for generating candidate frames and screening the anchor frames which accord with the characteristic size from the candidate frames;
the bounded feature fusion module is used for fusing the edge feature of the anchor frame with the global feature to obtain a fusion feature;
and the target detection module is used for selecting a feature map in the detected image according to the fusion features.
As shown in fig. 1, the ARM generates a candidate frame, performs preliminary screening and fine tuning on the anchor frame, and then transmits the adjusted anchor frame to the ODM module through the connection of the BFF module. BFF can better fuse edge information with region information.
The fourth concrete implementation mode: the embodiment further describes a method for identifying a fault of a railway wagon lower lock pin assembly part, which is described in the third specific embodiment, and in the embodiment, the fusion process of the bounded feature fusion module is as follows:
wherein, f is a group of compounds represented by,representing a feature extraction function, M edge For edge features, M (i) In order to be a global feature,it is indicated that the fusion is performed,representing i characteristicsFused features after fusion, i being the number of features fused, i =1, 2.
The fifth concrete implementation mode: in the embodiment, a high-definition image acquisition device is built at the track side of the truck to be tested, and the high-definition image acquisition device is used for acquiring a high-definition image of the truck to be tested.
The sixth specific implementation mode is as follows: the present embodiment further describes a method for identifying a fault of a railway wagon lower locking pin assembly, where the method is described in the first embodiment, and in the present embodiment, the high-definition image is a grayscale image.
The seventh embodiment: in the embodiment, when a fault is identified in the detected image, the coordinate of the fault position is obtained and whether the coordinate belongs to a preset range is judged, if so, a fault alarm is given, and otherwise, no alarm is given.
In the method for identifying the fault of the lower lock pin assembly of the railway wagon according to the first to seventh embodiments, the YOLOv5-SPD network needs to be trained first. Specifically, high-definition equipment is firstly built around the rail of the truck respectively, and the truck acquires a high-definition image after passing through the equipment. The image is a sharp grayscale image. The truck parts can be influenced by rain, mud, oil, black paint and other natural conditions or artificial conditions. Also, there may be differences in the images taken at different sites. Therefore, the lower lock pin is very different from the accessory image. Therefore, in the process of collecting the image data of the lower lock pin assembly, the images of the lower lock pin assembly under various conditions are collected as completely as possible to ensure diversity.
Comprehensively judging the coupler type and the coupler position provided by a TFDS (vehicle safety monitoring system) system frame pilot module to obtain the coordinate position of the accessory formed by the lower lock pins, intercepting the sub-image of the accessory formed by the lower lock pins according to the obtained coordinate position, and finally obtaining a training sample image.
Although the creation of the sample data set includes images under various conditions, data amplification of the sample data set is still required to improve the stability of the algorithm. The histogram equalization image is used for enhancement, so that the diversity and the applicability of the sample can be ensured to the maximum extent. Meanwhile, the histogram equalization can improve the brightness of the image and enhance the characteristic information of the lower lock pin accessory.
And training the YOLOv5-SPD network by using the collected sample data set. After training, inputting the tested truck information into the trained YOLOv5-SPD network, and identifying the fault of the truck lower lockpin assembly, as shown in FIG. 3.
The specific implementation mode is eight: the accessory fault recognition device formed by combining the lower lock pins of the railway wagon in the embodiment specifically comprises the following units:
an acquisition unit: used for acquiring an image of the accessory part formed by the lock pin assembly under the tested truck as a tested image,
an identification unit: the system is used for inputting a detected image into a YOLOv5-SPD network to obtain an identification result;
the YOLOv5-SPD network adopts a non-step convolution network to extract a characteristic subgraph and adopts a method of weighting and fusing target frames to identify faults in the characteristic subgraph.
The specific implementation method nine: a computer-readable storage medium according to the present embodiment stores a computer program, and when the computer program is executed, the method according to any one of the first to seventh embodiments is implemented.
The specific implementation mode is ten: the electronic device according to this embodiment includes a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, and the processor executes the computer program to implement any one of the methods according to the first to seventh embodiments.
According to the invention, high-definition imaging equipment is respectively built around the rail of the truck, and the truck acquires a high-definition image after passing through the equipment. The method comprises the steps of adopting a deep learning network, firstly classifying and positioning the car coupler to obtain position information of the car coupler, and intercepting a lower lockpin to form a subgraph of an accessory according to position prior knowledge of the accessory part formed by the lower lockpin on different car couplers. And detecting the loss fault of the lower lock pin assembly accessory by using a deep learning network. And uploading and alarming the coupler with the lower lock pin assembly and part loss, and carrying out corresponding processing by the staff according to the identification result to ensure the safe operation of the train.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: although the present disclosure has been described in detail with reference to the above embodiments, those skilled in the art will appreciate that various changes, modifications and equivalents can be made in the embodiments of the invention without departing from the scope of the invention as set forth in the claims.
Claims (10)
1. A fault identification method for accessories formed by lower lock pins of a rail wagon comprises the following specific steps:
acquiring an image of an accessory part formed by a lock pin assembly under a truck to be detected as a detected image, and inputting the detected image into a YOLOv5-SPD network to obtain an identification result;
the method is characterized in that the YOLOv5-SPD network adopts a non-step convolution network to extract a characteristic subgraph and adopts a target frame weighting fusion method to identify faults in the characteristic subgraph.
2. The method for identifying the fault of the assembly of the locking pin of the rail wagon according to claim 1, wherein a dynamic anchor frame is adopted to select a feature map in the tested image, and the feature map is input into a Yolov5-SPD network.
3. The method of claim 2, wherein the dynamic anchor block comprises an anchor improvement module, a bounded feature fusion module and a target detection module,
the anchor improving module is used for generating candidate frames and screening out anchor frames which accord with the characteristic size from the candidate frames;
the bounded feature fusion module is used for fusing the edge feature of the anchor frame with the global feature to obtain a fusion feature;
and the target detection module is used for selecting a feature map in the detected image according to the fusion feature.
4. The method for identifying the fault of the assembly of the lower lock pin of the railway wagon as claimed in claim 3, wherein the fusion process of the bounded feature fusion module is as follows:
wherein, the f is a characteristic of the compound,representative feature extraction function, M edge For edge features, M (i) In order to be a global feature,it is indicated that the fusion is performed,represents the fused features after fusion of i features, i being the number of features fused, i =1, 2.
5. The method for identifying the fault of the fitting formed by the lower lock pin of the railway wagon as claimed in claim 1, wherein a high-definition image acquisition device is built beside a rail of the wagon to be tested, and the high-definition image acquisition device is used for acquiring a high-definition image of the wagon to be tested.
6. The method for identifying the fault of the assembly of the lower lock pin of the railway wagon as claimed in claim 5, wherein the high-definition image is a grayscale image.
7. The method for identifying the fault of the accessory assembled by the lower lockpin of the railway wagon as claimed in claim 1, wherein when the detected image is identified to have the fault, the coordinate of the fault position is obtained and whether the coordinate belongs to a preset range is judged, if so, a fault alarm is given out, and otherwise, the fault alarm is not given out.
8. The utility model provides a railway freight car lower lockpin makes up accessory trouble recognition device, includes specifically that the following unit:
an acquisition unit: used for acquiring an image of the accessory part formed by the lock pin assembly under the tested truck as a tested image,
an identification unit: the image recognition system is used for inputting the tested image into a YOLOv5-SPD network to obtain a recognition result;
the method is characterized in that the YOLOv5-SPD network adopts a non-step convolution network to extract a characteristic subgraph and adopts a target frame weighting fusion method to identify faults in the characteristic subgraph.
9. A storage medium readable by a computer, the storage medium storing a computer program, the computer program when executed implementing the method of any of claims 1 to 7.
10. Electronic device comprising a storage medium, a processor and a computer program stored in said storage medium and executable on said processor, characterized in that said processor executes said computer program to implement the method according to any of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211313616.5A CN115527018A (en) | 2022-10-25 | 2022-10-25 | Fault identification method and device for parts formed by lower lock pins of railway wagon |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211313616.5A CN115527018A (en) | 2022-10-25 | 2022-10-25 | Fault identification method and device for parts formed by lower lock pins of railway wagon |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115527018A true CN115527018A (en) | 2022-12-27 |
Family
ID=84702693
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211313616.5A Pending CN115527018A (en) | 2022-10-25 | 2022-10-25 | Fault identification method and device for parts formed by lower lock pins of railway wagon |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115527018A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111080611A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon bolster spring fracture fault image identification method |
CN111652295A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon coupler yoke key joist falling fault identification method |
CN112906547A (en) * | 2021-02-09 | 2021-06-04 | 哈尔滨市科佳通用机电股份有限公司 | Railway train windshield breakage fault detection method based on E-YOLO |
CN114372949A (en) * | 2021-10-26 | 2022-04-19 | 桂林电子科技大学 | PCB surface defect detection method based on improved YOLOv5 algorithm |
CN114581860A (en) * | 2022-05-09 | 2022-06-03 | 武汉纺织大学 | Helmet detection algorithm based on improved YOLOv5 model |
CN114692826A (en) * | 2022-03-02 | 2022-07-01 | 华南理工大学 | Light-weight target detection system without prior frame |
-
2022
- 2022-10-25 CN CN202211313616.5A patent/CN115527018A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111080611A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon bolster spring fracture fault image identification method |
CN111652295A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon coupler yoke key joist falling fault identification method |
CN112906547A (en) * | 2021-02-09 | 2021-06-04 | 哈尔滨市科佳通用机电股份有限公司 | Railway train windshield breakage fault detection method based on E-YOLO |
CN114372949A (en) * | 2021-10-26 | 2022-04-19 | 桂林电子科技大学 | PCB surface defect detection method based on improved YOLOv5 algorithm |
CN114692826A (en) * | 2022-03-02 | 2022-07-01 | 华南理工大学 | Light-weight target detection system without prior frame |
CN114581860A (en) * | 2022-05-09 | 2022-06-03 | 武汉纺织大学 | Helmet detection algorithm based on improved YOLOv5 model |
Non-Patent Citations (1)
Title |
---|
RAJA SUNKARA 等: "No More Strided Convolutions or Pooling: A New CNN Building Block for Low-Resolution Images and Small Objects" * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP2018018461A (en) | Information processing apparatus, display device, information processing method, and program | |
CN111080611A (en) | Railway wagon bolster spring fracture fault image identification method | |
CN111080600A (en) | Fault identification method for split pin on spring supporting plate of railway wagon | |
CN111862029A (en) | Fault detection method for bolt part of vertical shock absorber of railway motor train unit | |
CN114742799B (en) | Industrial scene unknown type defect segmentation method based on self-supervision heterogeneous network | |
CN109543498B (en) | Lane line detection method based on multitask network | |
CN115527170A (en) | Method and system for identifying closing fault of door stopper handle of automatic freight car derailing brake device | |
CN113830136A (en) | Method for identifying malposition fault of angle cock handle of railway wagon | |
Abou Chacra et al. | Municipal infrastructure anomaly and defect detection | |
CN115409789A (en) | Power transmission line engineering defect detection method based on image semantic segmentation | |
JP2018017101A (en) | Information processing apparatus, information processing method, and program | |
CN115272850A (en) | Railway wagon BAB type brake adjuster pull rod head breaking fault identification method | |
CN115937736A (en) | Small target detection method based on attention and context awareness | |
CN115424128A (en) | Fault image detection method and system for lower link of freight car bogie | |
CN107247967A (en) | A kind of vehicle window annual test mark detection method based on R CNN | |
CN112085723B (en) | Automatic detection method for spring jumping fault of truck bolster | |
CN115527018A (en) | Fault identification method and device for parts formed by lower lock pins of railway wagon | |
CN114078106A (en) | Defect detection method based on improved Faster R-CNN | |
CN116596895A (en) | Substation equipment image defect identification method and system | |
KR102585146B1 (en) | Pothole detection device and method based on deep learning | |
CN111814582B (en) | Method and device for processing driver behavior monitoring image | |
CN114943869A (en) | Airport target detection method with enhanced style migration | |
CN115063771A (en) | Error correction method, system, storage medium and device for distance detection of obstacle | |
CN113643258A (en) | Method for detecting loss fault of skirt board at side part of train based on deep learning | |
CN116580277B (en) | Deep learning-based bottom electronic identification tag missing image identification method |
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: 20221227 |