CN114998286A - Train wheel tread defect detection method and system - Google Patents

Train wheel tread defect detection method and system Download PDF

Info

Publication number
CN114998286A
CN114998286A CN202210693894.1A CN202210693894A CN114998286A CN 114998286 A CN114998286 A CN 114998286A CN 202210693894 A CN202210693894 A CN 202210693894A CN 114998286 A CN114998286 A CN 114998286A
Authority
CN
China
Prior art keywords
defect
network model
dimensional image
image data
wheel tread
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
Application number
CN202210693894.1A
Other languages
Chinese (zh)
Inventor
张渝
赵波
彭建平
黄炜
章祥
王小伟
马莉
彭华
韩明阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu Lead Science & Technology Co ltd
Original Assignee
Chengdu Lead Science & Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chengdu Lead Science & Technology Co ltd filed Critical Chengdu Lead Science & Technology Co ltd
Priority to CN202210693894.1A priority Critical patent/CN114998286A/en
Publication of CN114998286A publication Critical patent/CN114998286A/en
Pending legal-status Critical Current

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
    • 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
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • 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
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • 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
    • 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
    • G06T2207/30164—Workpiece; Machine component
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention discloses a method and a system for detecting wheel tread defects of a train, wherein the method comprises the following steps: collecting two-dimensional image data of a wheel tread; constructing a neural network model for extracting the defect area; inputting the two-dimensional image data into the neural network model; the neural network model divides the two-dimensional image data into a plurality of types of image areas; and extracting a plurality of pieces of defect area information based on the plurality of types of image areas and generating a detection report. According to the method, the semantic segmentation network is used for accurately dividing the frequently-defective areas, so that after image background interference is reduced, the information of the defective areas and the defect types can be accurately extracted through the target detection model, the defect detection accuracy is improved, the missing report rate and the false report rate of the detection result are reduced, and the problem of low reliability of the detection result in the traditional wheel tread defect detection method is solved.

Description

Train wheel tread defect detection method and system
Technical Field
The invention relates to the technical field of rail vehicle detection, in particular to a method and a system for detecting wheel tread defects of a train.
Background
The train wheel is a key part related to the running safety of the train, transmits the load of the train to the steel rail, rotates on the steel rail to complete the running of the train, and is a final stress part for the running of the train. The wheel set tread is the part of train wheel and rail top surface contact, and the integrity of wheel set tread is the important factor of driving safety, and in the operation process, wheel set tread can appear wearing and tearing transfinite, tread scotch, peel off, quality problems such as rim surface defect of bruising etc. these problems can directly lead to the emergence of derailment accident, influence EMUs operational safety. In view of this, need carry out daily dynamic inspection to train wheel tread, ensure train operation safety.
The traditional method for detecting the wheel tread defect at home and abroad still adopts the traditional image scheme, firstly, the gray difference between a defect region and a non-defect region is utilized to roughly extract the defect region of the filtered wheel image, and then, the characteristics of regular defect shape, obvious edge characteristics and the like are utilized to extract the minimum enclosing rectangle of the defect by adopting a segmentation scheme. And finally, establishing a positive defect sample set and a negative defect sample set, and classifying the extracted defect images by adopting an SVM classifier. However, the conventional defect detection method depends on the image capturing quality, and the artificially set scheme is easily interfered by the background information when the background information is extremely rich. Causing more false alarms and detection distortion.
In summary, the existing wheel tread defect detection method has the problem of low reliability of detection results.
Disclosure of Invention
In view of the above, the present invention provides a method and a system for detecting a tread defect of a wheel of a train, which solves the problem of low reliability of a detection result of a conventional method for detecting a tread defect of a wheel by improving a method for detecting a tread defect of a wheel and a method for processing detection data.
In order to solve the problems, the technical scheme of the invention is to adopt a method for detecting the tread defect of the wheel of the train, which comprises the following steps: collecting two-dimensional image data of a wheel tread; constructing a neural network model for extracting the defect area; inputting the two-dimensional image data into the neural network model; the neural network model divides the two-dimensional image data into a plurality of types of image areas; and extracting a plurality of pieces of defect area information based on the plurality of types of image areas and generating a detection report.
Optionally, constructing a neural network model for extracting the defect region includes: constructing an initialization network model, wherein the network model comprises a semantic segmentation model and a target detection model; acquiring a training data set and a testing data set which are formed by wheel tread sample images containing artificially marked defect characteristic regions; training and testing the network model based on the training dataset and the testing dataset.
Optionally, the neural network model segments the two-dimensional image data into multiple classes of image regions, including: the semantic segmentation model performs down-sampling based on the two-dimensional image data and generates a multi-scale feature map; the semantic segmentation model can fuse a plurality of feature maps and generate a plurality of types of fused image areas by gradually adding a low-resolution feature map sub-network in parallel to a high-resolution feature map main network.
Optionally, extracting information of a plurality of defect regions based on a plurality of classes of the image regions and generating a detection report includes: and the target detection model outputs the defect region information consisting of a plurality of defect detection frame data based on the plurality of types of image regions, wherein the defect detection frame data at least comprises a defect type, a confidence coefficient, a coordinate of the upper left corner of a detection frame, a length of the detection frame and a width of the detection frame.
Optionally, the semantic segmentation model is a SegNet segmentation model.
Accordingly, the present invention provides a system for detecting a tread defect of a train, comprising: the data acquisition unit is used for acquiring two-dimensional image data of the wheel tread; and the data processing unit is used for constructing a neural network model for extracting the defect area, dividing the two-dimensional image data into a plurality of types of image areas by the neural network model, and extracting a plurality of pieces of defect area information based on the plurality of types of image areas to generate a detection report.
Optionally, the data acquisition unit comprises: a light source assembly for providing illumination conditions required by the image pickup unit; the camera shooting unit is used for collecting the two-dimensional image data.
Optionally, the wheel tread defect detection system further comprises: and the data storage unit is used for storing the two-dimensional image data and the sample image.
The method has the main improvement that the method for detecting the wheel tread defect of the train is provided, the semantic segmentation network is used for accurately dividing the frequently-defective area, the information and defect types of the defective area can be accurately extracted through the target detection model after the image background interference is reduced, the defect detection accuracy is improved, the missing report rate and the false report rate of the detection result are reduced, and the problem of low reliability of the detection result of the traditional wheel tread defect detection method is solved.
Drawings
FIG. 1 is a simplified flow diagram of a method of detecting wheel tread defects in a train in accordance with the present invention;
fig. 2 is a simplified unit connection diagram of the train wheel tread defect detection system of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood by those skilled in the art, the present invention will be further described in detail with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, a method for detecting a tread defect of a train includes: collecting two-dimensional image data of a wheel tread; constructing a neural network model for extracting the defect area; inputting the two-dimensional image data into the neural network model; the neural network model divides the two-dimensional image data into a plurality of types of image areas; and extracting a plurality of pieces of defect area information based on the plurality of types of image areas and generating a detection report.
Further, constructing a neural network model for extracting the defect region includes: constructing an initialization network model, wherein the network model comprises a semantic segmentation model and a target detection model; acquiring a training data set and a testing data set which are formed by wheel tread sample images containing artificially marked defect characteristic regions; training and testing the network model based on the training dataset and the testing dataset. Wherein the semantic segmentation model can be composed of a SegNet segmentation model; the target detection model can select a residual model and adopt a classification network of ResNet101, so that the problems such as gradient dissipation, gradient explosion and the like can be solved, and the extraction capability of the features is improved by increasing the number of network layers.
Further, the object detection model can also use a free anchor form of object detection network.
Further, the neural network model segments the two-dimensional image data into a plurality of classes of image regions, including: the semantic segmentation model performs down-sampling based on the two-dimensional image data and generates a multi-scale feature map; the semantic segmentation model can fuse a plurality of feature maps and generate a plurality of types of fused image areas by gradually adding a low-resolution feature map sub-network in parallel to a high-resolution feature map main network. Therefore, the wheel is divided into regions such as a tread, a hub and the like, and the regions can be divided into image regions of various types according to the form of the labels.
Further, extracting information of a plurality of defect areas based on a plurality of types of the image areas and generating a detection report includes: and the target detection model outputs the defect region information consisting of a plurality of defect detection frame data based on the plurality of types of image regions, wherein the defect detection frame data at least comprises a defect type, a confidence coefficient, a coordinate of the upper left corner of a detection frame, a length of the detection frame and a width of the detection frame.
According to the method, the semantic segmentation network is used for accurately dividing the frequently-defective areas, so that after image background interference is reduced, the information of the defective areas and the defect types can be accurately extracted through the target detection model, the defect detection accuracy is improved, the missing report rate and the false report rate of the detection result are reduced, and the problem of low reliability of the detection result in the traditional wheel tread defect detection method is solved.
Accordingly, the present invention provides a system for detecting a tread defect of a train, comprising: the data acquisition unit is used for acquiring two-dimensional image data of the wheel tread; and the data processing unit is used for constructing a neural network model for extracting the defect area, dividing the two-dimensional image data into a plurality of classes of image areas by the neural network model, extracting a plurality of pieces of defect area information based on the plurality of classes of image areas and generating a detection report. Wherein, wheel tread defect detection system still includes: a data storage unit for storing two-dimensional image data and a sample image; the data acquisition unit includes: a light source assembly for providing illumination conditions required for the image pickup unit; the camera shooting unit is used for collecting the two-dimensional image data. Specifically, the camera units are arranged on two sides of the train wheel.
The method and the system for detecting the tread defect of the train provided by the embodiment of the invention are described in detail above. The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description. It should be noted that, for those skilled in the art, without departing from the principle of the present invention, it is possible to make various improvements and modifications to the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the technical solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

Claims (8)

1. A method for detecting defects of a tread of a wheel of a train is characterized by comprising the following steps:
collecting two-dimensional image data of a wheel tread;
constructing a neural network model for extracting the defect area;
inputting the two-dimensional image data into the neural network model;
the neural network model divides the two-dimensional image data into a plurality of types of image areas;
and extracting a plurality of pieces of defect area information based on the plurality of types of image areas and generating a detection report.
2. The method of detecting wheel tread defects according to claim 1, wherein constructing a neural network model for extracting defect regions comprises:
constructing an initialization network model, wherein the network model comprises a semantic segmentation model and a target detection model;
acquiring a training data set and a testing data set which are formed by wheel tread sample images containing artificially marked defect characteristic regions;
training and testing the network model based on the training dataset and the testing dataset.
3. The method of detecting wheel tread defects according to claim 2, wherein the neural network model segments the two-dimensional image data into multiple classes of image regions, including:
the semantic segmentation model performs down-sampling based on the two-dimensional image data and generates a multi-scale feature map;
the semantic segmentation model can fuse a plurality of feature maps and generate a plurality of types of fused image areas by gradually adding a low-resolution feature map sub-network in parallel to a high-resolution feature map main network.
4. The method according to claim 3, wherein extracting a plurality of defect region information based on a plurality of types of the image regions and generating a detection report comprises:
and the target detection model outputs the defect region information consisting of a plurality of defect detection frame data based on the plurality of types of image regions, wherein the defect detection frame data at least comprises a defect type, a confidence coefficient, a coordinate of the upper left corner of a detection frame, a length of the detection frame and a width of the detection frame.
5. The method according to claim 4, wherein the semantic segmentation model is a SegNet segmentation model.
6. A train wheel tread defect detection system, comprising:
the data acquisition unit is used for acquiring two-dimensional image data of the wheel tread;
and the data processing unit is used for constructing a neural network model for extracting the defect area, dividing the two-dimensional image data into a plurality of classes of image areas by the neural network model, extracting a plurality of pieces of defect area information based on the plurality of classes of image areas and generating a detection report.
7. The wheel tread defect detection system of claim 6, wherein the data acquisition unit comprises:
a light source assembly for providing illumination conditions required for the image pickup unit;
the camera shooting unit is used for collecting the two-dimensional image data.
8. The wheel tread defect detection system of claim 6, further comprising:
and the data storage unit is used for storing the two-dimensional image data and the sample image.
CN202210693894.1A 2022-06-19 2022-06-19 Train wheel tread defect detection method and system Pending CN114998286A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210693894.1A CN114998286A (en) 2022-06-19 2022-06-19 Train wheel tread defect detection method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210693894.1A CN114998286A (en) 2022-06-19 2022-06-19 Train wheel tread defect detection method and system

Publications (1)

Publication Number Publication Date
CN114998286A true CN114998286A (en) 2022-09-02

Family

ID=83035330

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210693894.1A Pending CN114998286A (en) 2022-06-19 2022-06-19 Train wheel tread defect detection method and system

Country Status (1)

Country Link
CN (1) CN114998286A (en)

Similar Documents

Publication Publication Date Title
EP2697738B1 (en) Method and system of rail component detection using vision technology
CN107545239B (en) Fake plate detection method based on license plate recognition and vehicle characteristic matching
TWI497422B (en) A system and method for recognizing license plate image
US9760789B2 (en) Robust cropping of license plate images
CN112686812B (en) Bank card inclination correction detection method and device, readable storage medium and terminal
CN111738342B (en) Pantograph foreign matter detection method, storage medium and computer equipment
CN110276295B (en) Vehicle identification number detection and identification method and device
CN111260629A (en) Pantograph structure abnormity detection algorithm based on image processing
CN111626170B (en) Image recognition method for railway side slope falling stone intrusion detection
CN110348360A (en) A kind of examining report recognition methods and equipment
Azad et al. New method for optimization of license plate recognition system with use of edge detection and connected component
CN107273802A (en) A kind of detection method and device of railroad train brake shoe drill ring failure
CN108664997A (en) High iron catenary equipotential line defective mode detection method based on cascade Faster R-CNN
CN103606221A (en) Fault automatic diagnostic method of counter and device
CN107909009A (en) A kind of obstacle detection method and device based on road surface study
CN112434555B (en) Key value pair region identification method and device, storage medium and electronic equipment
CN111079629A (en) Railway wagon center plate out-of-position fault image identification method
CN112488995A (en) Intelligent injury judging method and system for automatic train maintenance
Keong et al. Malaysian automatic number plate recognition system using Pearson correlation
JP2011060221A (en) Discriminator generation method, computer program, discriminator generating device and predetermined object detecting device
Shanthakumari et al. Mask RCNN and Tesseract OCR for vehicle plate character recognition
Shomee et al. License plate detection and recognition system for all types of bangladeshi vehicles using multi-step deep learning model
CN112651965B (en) Three-dimensional detection method and system for tread defect of wheel
CN111597992B (en) Scene object abnormity identification method based on video monitoring
CN114998286A (en) Train wheel tread defect detection method and system

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