CN111079820B - Image recognition-based rail wagon fire-proof plate fault recognition method - Google Patents
Image recognition-based rail wagon fire-proof plate fault recognition method Download PDFInfo
- Publication number
- CN111079820B CN111079820B CN201911272266.0A CN201911272266A CN111079820B CN 111079820 B CN111079820 B CN 111079820B CN 201911272266 A CN201911272266 A CN 201911272266A CN 111079820 B CN111079820 B CN 111079820B
- Authority
- CN
- China
- Prior art keywords
- picture
- fire
- fault
- pictures
- plate
- 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.)
- Active
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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- 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)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
A rail wagon fire-proof plate fault identification method based on image identification relates to the technical field of rail wagon fault identification. The invention aims to solve the problems that the failure of the fireproof plate of the truck is easy to miss detection and the efficiency is low in the manual detection. High-definition imaging equipment is arranged on two sides and the center of a rail of the truck, and after the truck passes through the equipment installation position, an image of a fireproof plate on the upper portion of the bogie is obtained. The method has the advantages that faults of the fireproof plate in loss, falling and damage states are detected through the fast rcnn network of the deep learning network, manual detection of the fireproof plate is replaced by an automatic image identification mode, fault identification and detection efficiency is improved, and labor cost is reduced; the deep learning is applied to the detection of the faults of the fire-proof plate, so that the robustness and the accuracy of the algorithm can be effectively improved.
Description
Technical Field
The invention belongs to the technical field of fault identification of rail wagons.
Background
Train fires are one of the inertial accidents in driving. To address this problem, the railroads have taken various measures, including the addition of fire protection panels to the freight cars. The fireproof board is also called as a fireproof board, is a high-pressure laminated board with thermosetting resin impregnated paper, is a fireproof building material, and can play a certain role in fire prevention and fire resistance. Although the fire-proof effect is enhanced with the improvement methods of increasing the use area of the fire-proof plate and increasing the distance between the fire-proof plate and the floor. However, because the protection measures are not comprehensive enough, the damage of the fireproof plate and other conditions can not be found in time, so that fire accidents are not completely eradicated all the time, the transportation safety is threatened, and economic loss and adverse effects are caused.
At present, the fault detection of the fireproof plate part of the truck generally adopts a manual checking mode. However, such a detection method is greatly affected by factors such as quality of business, responsibility, and labor intensity of the operator, and thus, the conditions such as missing detection or simplified operation are often likely to occur.
Disclosure of Invention
The invention provides a method for identifying the failure of a fireproof plate of a railway wagon based on image identification, aiming at solving the problems that the failure of the fireproof plate of the wagon is easy to miss detection and low in efficiency in manual mode detection.
A rail wagon fire-proof plate fault identification method based on image identification comprises the following steps:
the method comprises the following steps:
collecting pictures of fire-proof plates of different types of railway trucks under different time, place and environment to establish a sample library, wherein the pictures of the fire-proof plates comprise pictures of the fire-proof plates under a normal state and pictures of the fire-proof plates under a fault state,
the picture of the fireproof plate on the upper part of the same truck bogie is transversely and evenly divided into two parts, then the two parts are longitudinally spliced to form a new picture with the length-width ratio smaller than 3,
dividing each new picture into 8-10 areas, marking whether each area has a fault and the fault type, and generating a corresponding label file,
establishing a training data set of a Faster rcnn target detection network by the new picture and the corresponding label file;
step two:
substituting the training data set into a Resnet50 pre-training network model of the Faster rcnn target detection network, training the model to obtain a weight value of the Faster rcnn target detection network, and substituting the weight value into the Faster rcnn target detection network to finish training;
step three:
collecting a picture of the fireproof plate to be identified, transversely and evenly dividing the picture of the fireproof plate into two parts, longitudinally splicing the two parts to form the picture with the length-width ratio smaller than 3,
calculating the average gray value of the picture, adjusting the brightness and contrast of the picture according to the average gray value to make the whole gray value of the picture be the average gray value,
calculating the ratio of a preset standard gray value to the average gray value of the picture, multiplying the gray value of each pixel point in the picture by the ratio, and adding a gray adjustment value to each product to obtain a final detection picture, wherein the value range of the gray adjustment value is 0-255;
step four:
and inputting the detection picture into the fast rcnn target detection network trained in the step two to obtain a recognition result.
Further, in step two, after the training is completed, the trained Faster rcnn target detection network is optimized by using the tensorRT.
Furthermore, the sample library not only includes the acquired picture, but also includes the picture after stretching, rotating and mirroring the acquired picture.
Further, in step two, the whole new picture in the training data set is input into the Resnet50 pre-training network model for feature extraction.
Further, in the third step, the sizes of all the final detected pictures are unified, and then the fourth step is executed.
Further, for each train with 4 fire-proof plates, the fire-proof plate pictures of each truck are fused into a matrix with the size of (4,380,1000, 3).
High-definition imaging equipment is arranged on two sides and the center of a rail of the truck, and after the truck passes through the equipment installation position, an image of a fireproof plate on the upper portion of the bogie is obtained. The method has the advantages that faults of the fireproof plate in loss, falling and damage states are detected through the fast rcnn network of the deep learning network, manual detection of the fireproof plate is replaced by an automatic image identification mode, fault identification and detection efficiency is improved, and labor cost is reduced;
the deep learning is applied to the detection of the faults of the fireproof plate, so that the robustness and the accuracy of the algorithm can be effectively improved;
after the fast rcnn network is trained for deep learning, the trained network model is accelerated by adopting the tensorrT during prediction, so that the prediction accuracy rate is ensured, and the running efficiency is also ensured.
And finally, uploading the final detection result to a network for manual reference. And the staff performs corresponding processing according to the image recognition result to ensure the safe operation of the locomotive.
Drawings
Fig. 1 is a flowchart of a method for identifying a failure of a fire-proof plate of a railway wagon based on image identification according to the invention.
Detailed Description
With the great improvement of the processing performance of chip hardware, a foundation is provided for the complex computation of a deep network. The deep learning is widely applied to the field of image processing, and compared with the traditional mode, the deep learning integrates the feature learning into the process of establishing the model, so that the accuracy and efficiency of fault detection can be effectively improved.
In recent years, object detection and recognition have been popular fields in computer vision research, and various network models for object detection and recognition are used in deep learning. In these algorithm models, the Faster-Region Convolutional Neural Network (Faster-Region based Convolutional Neural Network) algorithm is improved on the basis of the RCNN algorithm and the Fast RCNN algorithm, and is a more classical algorithm in the field of target detection and recognition.
The embodiment integrates the deep learning method into the detection field of the freight train, so that the damage of the upper fireproof plate of the bogie of the freight train can be accurately judged. The method comprises the following specific steps:
the first embodiment is as follows: specifically describing the embodiment with reference to fig. 1, the method for identifying the failure of the fire-proof plate of the railway wagon based on image identification in the embodiment comprises the following steps:
the method comprises the following steps:
the method comprises the steps of collecting images of the upper fire-proof plate of the bogie of different types of railway trucks at different time, places and environments from a network big database or an actual application environment, then respectively carrying out stretching, rotation and mirror image transformation on each collected image, establishing a sample library by utilizing all images before and after transformation, wherein the images of the upper fire-proof plate of the bogie comprise the images of the fire-proof plate in a normal state and the images of the fire-proof plate in a fault state.
For example: the method comprises the steps of collecting a fireproof plate picture, copying three identical pictures, respectively stretching, rotating and carrying out mirror image transformation on the three copied pictures to obtain a stretched picture, a rotated picture and a mirrored picture, and then storing an original picture and the three transformed pictures into a sample library together.
The operation aims to amplify the sample, collect the fireproof plate pictures under different conditions, and is beneficial to enriching the sample data and increasing the robustness and the adaptability of the subsequent training result.
Since the width-to-height ratio of the fireproof plate picture exceeds 10, the characteristic extraction of deep learning is not facilitated. Therefore, the upper fireproof plate picture of the same bogie is transversely divided into two pictures, and the two pictures are spliced up and down, so that the length-width ratio of the images can be controlled within 3, and the spliced pictures are adjusted to be the same in size.
Dividing each adjusted new picture into 8-10 areas, respectively marking whether each area has a fault and the fault type, and generating a corresponding label file,
and establishing a training data set of the Faster rcnn target detection network by using the new picture and the corresponding label file.
Step two:
feature extraction of the Faster rcnn target detection network employs the Resnet50 pre-trained network model. Resnet50 has higher precision compared with the Inception feature extraction network and higher operation efficiency compared with Resnet 101.
Specifically, the training data set is input into the Resnet50 pre-training network model of the fast rcnn target detection network, that is: inputting a whole new image in a training data set into a Resnet50 pre-training network model for feature extraction, generating suggestion windows by using RPN (resilient packet network), reducing the number of the suggestion windows to 200 in order to improve the running speed of the network, training the model, obtaining a weight value of a Faster rcnn target detection network, and substituting the weight value into the Faster rcnn target detection network to finish training;
after the training is finished, the trained Faster rcnn target detection network is optimized by using the tensorrT, and the prediction speed of the network is improved. Data type accuracy in the prediction process is reduced by combining and replacing some network layers by the tensorRT, and the FP32 is changed into the FP 16.
Step three:
high-definition imaging equipment is arranged at the two sides and the center of a rail of the truck, after the truck passes through the equipment installation position, a picture of the fireproof plate to be identified can be obtained, the picture of the fireproof plate is transversely and evenly divided into two parts, the two parts are longitudinally spliced to form a picture with the length-width ratio smaller than 3,
calculating the average gray value of the picture, adjusting the brightness and contrast of the picture according to the average gray value to make the whole gray value of the picture be the average gray value,
calculating a ratio of a preset standard gray value to an average gray value of the picture, multiplying the gray value of each pixel point in the picture by the ratio, and adding a gray adjustment value to each product to obtain a final detection picture, wherein the value range of the gray adjustment value is 0-255, and specifically, the adjustment value can be 65 during actual application;
and step four is executed after the sizes of all the final detection pictures are unified.
Step four:
and inputting the final detection picture into the fast rcnn target detection network trained in the second step to obtain a recognition result, wherein the recognition result comprises whether a fault exists, a fault type and a fault position.
In practice, there are usually 4 fire protection plates per truck. Therefore, tens of fireproof plate images of the whole train are all read into the memory in parallel in the fault identification process. The images of the fire protection plates of each truck are then fused into a matrix of sizes (4,380,1000,3) and then input into the network for prediction.
Meanwhile, as the size of the image of the fireproof plate is large, the types of faults needing to be detected are large, the size of the image cannot be reduced in order to keep the characteristic of breakage of the fireproof plate, and the number of the fireproof plates of a train is not fixed. Therefore, in the embodiment, the Resnet50 is used as a feature extraction network to ensure the detection precision, and the tensorRT is used for optimizing the training model and reducing the number of the suggestion windows, so that the program running efficiency can be improved on the premise of ensuring the detection precision.
In the embodiment, the Faster rcnn deep learning network is adopted to detect whether the fireproof plate image has damage, loss and falling faults. Although the speed of the target detection network such as SSD, yolo and the like is higher, the detection effect is poor for the fault that the target is smaller in the image, such as the breakage of the fireproof plate. In order to solve the problem that the efficiency of the Faster rcnn is low in the operation process, the prediction process of the Faster rcnn network is accelerated by adopting the tensorrT, and the prediction speed of the network is improved.
Claims (6)
1. A rail wagon fire-proof plate fault identification method based on image identification is characterized by comprising the following steps:
the method comprises the following steps:
collecting pictures of fire-proof plates of different types of railway trucks under different time, place and environment to establish a sample library, wherein the pictures of the fire-proof plates comprise pictures of the fire-proof plates under a normal state and pictures of the fire-proof plates under a fault state,
the picture of the fireproof plate on the upper part of the same truck bogie is transversely and evenly divided into two parts, then the two parts are longitudinally spliced to form a new picture with the length-width ratio smaller than 3,
dividing each new picture into 8-10 areas, marking whether each area has a fault and the fault type, and generating a corresponding label file,
establishing a training data set of a Faster rcnn target detection network by the new picture and the corresponding label file;
step two:
substituting the training data set into a Resnet50 pre-training network model of the Faster rcnn target detection network, training the model to obtain a weight value of the Faster rcnn target detection network, and substituting the weight value into the Faster rcnn target detection network to finish training;
step three:
collecting a picture of the fireproof plate to be identified, transversely and evenly dividing the picture of the fireproof plate into two parts, longitudinally splicing the two parts to form the picture with the length-width ratio smaller than 3,
calculating the average gray value of the picture, adjusting the brightness and contrast of the picture according to the average gray value to make the whole gray value of the picture be the average gray value,
calculating the ratio of a preset standard gray value to the average gray value of the picture, multiplying the gray value of each pixel point in the picture by the ratio, and adding a gray adjustment value to each product to obtain a final detection picture, wherein the value range of the gray adjustment value is 0-255;
step four:
and inputting the detection picture into the fast rcnn target detection network trained in the step two to obtain a recognition result.
2. The method for identifying the fault of the fire-proof plate of the railway wagon based on the image identification is characterized in that in the second step, after the training is finished, the trained Faster rcnn target detection network is optimized by using tensorrT.
3. The method for identifying the fault of the fire protection plate of the railway wagon based on the image identification is characterized in that the sample library not only comprises the acquired pictures, but also comprises the pictures obtained by stretching, rotating and mirroring the acquired pictures.
4. The method for identifying the failure of the fire protection plate of the railway wagon based on the image identification is characterized in that in the second step, the whole new picture in the training data set is input into a Resnet50 pre-training network model for feature extraction.
5. The method for identifying the fault of the fire-proof plate of the railway wagon based on the image identification is characterized in that in the third step, the sizes of all final detection pictures are unified, and then the fourth step is executed.
6. The method for identifying the fault of the fire-proof plate of the railway wagon based on the image identification is characterized in that the pictures of the fire-proof plate of each wagon are fused into a matrix with the size of (4,380,1000,3) for each train with 4 fire-proof plates.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911272266.0A CN111079820B (en) | 2019-12-12 | 2019-12-12 | Image recognition-based rail wagon fire-proof plate fault recognition method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911272266.0A CN111079820B (en) | 2019-12-12 | 2019-12-12 | Image recognition-based rail wagon fire-proof plate fault recognition method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111079820A CN111079820A (en) | 2020-04-28 |
CN111079820B true CN111079820B (en) | 2020-08-25 |
Family
ID=70314015
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911272266.0A Active CN111079820B (en) | 2019-12-12 | 2019-12-12 | Image recognition-based rail wagon fire-proof plate fault recognition method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111079820B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
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 |
CN112488049B (en) * | 2020-12-16 | 2021-08-24 | 哈尔滨市科佳通用机电股份有限公司 | Fault identification method for foreign matter clamped between traction motor and shaft of motor train unit |
CN112907524B (en) * | 2021-02-04 | 2021-11-12 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting fault of fire-proof plate of rail wagon based on image processing |
CN113312961A (en) * | 2021-04-03 | 2021-08-27 | 国家计算机网络与信息安全管理中心 | Logo recognition acceleration method |
CN115331085B (en) * | 2022-08-15 | 2023-05-12 | 哈尔滨市科佳通用机电股份有限公司 | Bogie spring supporting plate detection method based on deep learning |
CN115439451B (en) * | 2022-09-09 | 2023-04-21 | 哈尔滨市科佳通用机电股份有限公司 | Denoising detection method for spring supporting plate of bogie of railway freight car |
CN115965915B (en) * | 2022-11-01 | 2023-09-08 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon connecting pull rod breaking fault identification method and system based on deep learning |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110490849A (en) * | 2019-08-06 | 2019-11-22 | 桂林电子科技大学 | Surface Defects in Steel Plate classification method and device based on depth convolutional neural networks |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8503539B2 (en) * | 2010-02-26 | 2013-08-06 | Bao Tran | High definition personal computer (PC) cam |
CN106952250B (en) * | 2017-02-28 | 2021-05-07 | 北京科技大学 | Metal plate strip surface defect detection method and device based on fast R-CNN network |
KR20190018798A (en) * | 2017-08-16 | 2019-02-26 | 강병수 | car noise sound with Convolution Nueral Network classification method |
CN110119686B (en) * | 2019-04-17 | 2020-09-25 | 电子科技大学 | Safety helmet real-time detection method based on convolutional neural network |
CN110211097B (en) * | 2019-05-14 | 2021-06-08 | 河海大学 | Crack image detection method based on fast R-CNN parameter migration |
CN110555842A (en) * | 2019-09-10 | 2019-12-10 | 太原科技大学 | Silicon wafer image defect detection method based on anchor point set optimization |
-
2019
- 2019-12-12 CN CN201911272266.0A patent/CN111079820B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110490849A (en) * | 2019-08-06 | 2019-11-22 | 桂林电子科技大学 | Surface Defects in Steel Plate classification method and device based on depth convolutional neural networks |
Also Published As
Publication number | Publication date |
---|---|
CN111079820A (en) | 2020-04-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111079820B (en) | Image recognition-based rail wagon fire-proof plate fault recognition method | |
CN111079819B (en) | Method for judging state of coupler knuckle pin of railway wagon based on image recognition and deep learning | |
CN109766746B (en) | Track foreign matter detection method for aerial video of unmanned aerial vehicle | |
CN111489339A (en) | Method for detecting defects of bolt spare nuts of high-speed railway positioner | |
CN111311567A (en) | Method for identifying fastener and steel rail diseases of track line image | |
CN111080608A (en) | Method for recognizing closing fault image of automatic brake valve plug handle of railway wagon in derailment | |
CN111080614A (en) | Method for identifying damage to rim and tread of railway wagon wheel | |
CN108846331B (en) | Video identification method for judging whether screw fastener of motor train unit chassis falls off or not | |
JP2022549541A (en) | Defect detection method, apparatus, electronic equipment and computer storage medium | |
CN111292432A (en) | Vehicle charging type distinguishing method and device based on vehicle type recognition and wheel axle detection | |
US11495022B2 (en) | Method for recognizing an object of a mobile unit | |
CN112464846B (en) | Automatic recognition method for abnormal faults of freight train carriage of station | |
WO2021004130A1 (en) | Logo defect detection method and device | |
CN109165541A (en) | Coding method for vehicle component in intelligent recognition rail traffic vehicles image | |
CN111080601A (en) | Method for identifying fault image of pull ring grinding shaft of derailment brake device of railway wagon | |
CN111080607B (en) | Rolling bearing oil slinging fault detection method based on image recognition | |
CN111091543A (en) | Railway wagon swing bolster spring loss fault target detection method | |
CN108648210B (en) | Rapid multi-target detection method and device under static complex scene | |
CN113221839A (en) | Automatic truck image identification method and system | |
CN115527170A (en) | Method and system for identifying closing fault of door stopper handle of automatic freight car derailing brake device | |
CN107247967A (en) | A kind of vehicle window annual test mark detection method based on R CNN | |
Liu et al. | A YOLOv3-based deep learning application research for condition monitoring of rail thermite welded joints | |
CN115359306B (en) | Intelligent identification method and system for high-definition images of railway freight inspection | |
CN113762247A (en) | Road crack automatic detection method based on significant instance segmentation algorithm | |
CN111652228A (en) | Railway wagon sleeper beam hole foreign matter detection 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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |