CN109740410A - A kind of train groups fault recognition method and device without presetting template - Google Patents
A kind of train groups fault recognition method and device without presetting template Download PDFInfo
- Publication number
- CN109740410A CN109740410A CN201811320495.0A CN201811320495A CN109740410A CN 109740410 A CN109740410 A CN 109740410A CN 201811320495 A CN201811320495 A CN 201811320495A CN 109740410 A CN109740410 A CN 109740410A
- Authority
- CN
- China
- Prior art keywords
- image
- compartment
- train
- target
- train groups
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
Abstract
The embodiment of the present invention provides a kind of train groups fault recognition method and device without presetting template, this method comprises: according to the train groups image of acquisition, identify the headstock and the tailstock of train groups, image after removing the image and the tailstock in the train groups image of acquisition before headstock, obtains complete train group image;According to complete train group image, the compartment junction of train groups is identified;According to compartment junction, complete train group image is spliced, obtains the complete image in each section compartment of train groups;Characteristic matching is carried out by target compartment image and with reference to compartment image, identifies target compartment abort situation, wherein target compartment is compartment to be detected, is other compartments in train groups in addition to target compartment with reference to compartment.Train groups fault recognition method and device provided in an embodiment of the present invention without default template completes the real-time analysis and identification of failure using the train groups image currently acquired under the premise of being not necessarily to history car modal library.
Description
Technical field
The present invention relates to fault identification field more particularly to it is a kind of without preset template train groups fault recognition method and
Device.
Background technique
In order to meet the growing transportation demand of people, the speed of service of railroad train is stepped up, and density is not yet
Disconnected to increase, railway transportation production is higher and higher to the reliability requirement of train operation.Motor train unit train is to complete railway high speed fortune
The defeated most important mobile device of task, EMU density is big, operating mileage is long, operation environment is complicated, the speed of service is fast, fortune
Row figure layout is intensive, and under high-speed cruising state, any tiny, subtle failure is likely to cause major accident, influences height
The operational safety of fast railway causes traffic interruptions, backlog, causes greater loss to national economy, therefore reinforce EMU
Fault identification accuracy rate is most important.
However current EMU failure automatic identifying method, it is mostly based on largely default template library, according to train vehicle
Its corresponding history template is retrieved in compartment number, characteristic matching is carried out, to judge whether there is incipient fault.
There are the following problems for existing EMU fault recognition method:
Firstly, default template library is since the time of shooting is different, there are the external disturbing factor such as illumination, running speed, from
And though causing is same train, acquisition image have biggish difference, to cause very high rate of false alarm.
Secondly, train will appear the change in location of some components when repairing, to largely need to increase newly
Otherwise template library will have very big wrong report to occur.
Summary of the invention
The embodiment of the present invention is to overcome above-mentioned technological deficiency, provides a kind of train groups fault identification side without presetting template
Method and device.
In a first aspect, the embodiment of the present invention provides a kind of train groups fault recognition method without presetting template, comprising:
According to the train groups image of acquisition, the headstock and the tailstock of train groups are identified, remove the train groups image of the acquisition
The image after image and the tailstock before middle headstock, obtains complete train group image;
According to the complete train group image, the compartment junction of train groups is identified;
According to the compartment junction, the complete train group image is spliced, obtains each section vehicle of train groups
The complete image in compartment;
Characteristic matching is carried out by target compartment image and with reference to compartment image, identifies target compartment abort situation, wherein institute
Stating target compartment is compartment to be detected, is other compartments in train groups in addition to the target compartment with reference to compartment.
Second aspect, the embodiment of the present invention provide a kind of train groups fault identification device without presetting template, comprising:
First processing module identifies the headstock and the tailstock of train groups for the train groups image according to acquisition, described in removal
The image after image and the tailstock in the train groups image of acquisition before headstock, obtains complete train group image;
Second processing module, for identifying the compartment junction of train groups according to the complete train group image;
Splicing module, for splicing to the complete train group image, obtaining train according to the compartment junction
The complete image in each section compartment of group;
Identification module is used to carry out characteristic matching, the event of identification target compartment by target compartment image and with reference to compartment image
Hinder position, wherein the target compartment is compartment to be detected, is its in train groups in addition to the target compartment with reference to compartment
His compartment.
The third aspect, the embodiment of the present invention provide a kind of electronic equipment, including memory and processor, the processor and
The memory completes mutual communication by bus;The memory, which is stored with, to be referred to by the program that the processor executes
It enables, the processor calls described program to instruct the method being able to carry out as described in relation to the first aspect.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating
Machine program realizes that the train groups failure as described in relation to the first aspect without default template is known when the computer program is executed by processor
Other method.
Train groups fault recognition method and device provided in an embodiment of the present invention without default template, without any default
Headstock tailstock template library without identification rain model, but utilizes this column train image currently acquired, first identification train
The headstock tailstock of group, then according to the correlation between image, the characteristic of image itself identifies the compartment junction of train groups,
Template matching is carried out with other compartment images of train groups to be detected again, without finding history car modal according to rain model
Fault identification is realized under the premise of library, completes real-time analysis and the automatic early-warning of the visual structure abnormal conditions of operation EMU.
Fault identification is carried out using the current acquired image of same train, not only largely avoids illumination, running speed etc.
The influence of external disturbing factor can not also be influenced by the artificial service mode such as components position adjustment, be efficiently solved
The rate of false alarm as caused by the external factors such as illumination, train running speed, the problems such as rate of failing to report is higher in current failure identification, mention
The high efficiency of EMU fault identification and upkeep operation improves the discovery to the hidden failure of EMU in operational process and pre-
Alert ability.
Detailed description of the invention
Fig. 1 is the flow diagram of the train groups fault recognition method provided in an embodiment of the present invention without default template;
Fig. 2 is the flow diagram of the headstock tailstock recognizer of gauss hybrid models provided in an embodiment of the present invention;
Fig. 3 is the flow diagram of the compartment junction recognizer of symmetrical image detection provided in an embodiment of the present invention;
Fig. 4 is the process signal for the train groups fault recognition method without default template that further embodiment of this invention provides
Figure;
Fig. 5 is the structural schematic diagram of the train groups fault identification device provided in an embodiment of the present invention without default template;
Fig. 6 is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached in the embodiment of the present invention
Figure, is clearly and completely described the technical solution in the present invention, it is clear that described embodiment is one of the invention
Divide embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making
Every other embodiment obtained, shall fall within the protection scope of the present invention under the premise of creative work.
Existing train groups fault recognition method is mostly based on largely default template library, is retrieved according to compartment number
Its corresponding history template carries out characteristic matching, to judge whether there is incipient fault.Due to presetting template library shooting time
Difference, the external disturbing factor such as illumination, running speed can make the image difference of acquisition larger, and rate of false alarm is larger, while train
The change in location of components needs newly-increased template library.To solve the above-mentioned problems, the embodiment of the present invention provides a kind of train groups event
Hinder recognition methods, without default template library, but the directly collection in worksite train groups image that needs to detect, according to collection in worksite
Train groups image carries out fault identification, avoids the influence of the external factors such as illumination, driving and components change in location.
Fig. 1 is the flow diagram of the train groups fault recognition method provided in an embodiment of the present invention without default template,
As shown in Figure 1, comprising:
Step 11, according to the train groups image of acquisition, the headstock and the tailstock of train groups is identified, the train of the acquisition is removed
The image after image and the tailstock in group image before headstock, obtains complete train group image;
Step 12, according to the complete train group image, the compartment junction of train groups is identified;
Step 13, according to the compartment junction, the complete train group image is spliced, the every of train groups is obtained
The complete image in one section compartment;
Step 14, characteristic matching is carried out by target compartment image and with reference to compartment image, identifies target compartment abort situation,
Wherein, the target compartment is compartment to be detected, is other compartments in train groups in addition to the target compartment with reference to compartment.
Train groups fault recognition method without default template provided in an embodiment of the present invention, with existing train groups event
Hinder recognition methods the difference is that, existing train groups fault recognition method be based on a large amount of headstock tailstock preset template
Library needs to acquire the image of various types of train groups before fault identification, in fault identification according to be detected
The model of train groups finds corresponding history template according to the coach number of train groups to detect its headstock tailstock, thus into one
It walks and determines abort situation, and the train groups fault recognition method provided in an embodiment of the present invention without default template, without any
History template image, it is only necessary to according to the train groups image currently acquired, the headstock of automatic identification train groups and tailstock position, and
Image after image and the tailstock before the headstock for being not belonging to train image acquired in image is occurred occur removes;According to modification
Complete train group image afterwards identifies the compartment junction of train groups, and according to compartment junction position, splicing acquires image,
To which the image of acquisition is carried out splicing segmentation with compartment serial number, the complete image in each section compartment is formed.
Specific operation is train groups to be detected for one group, to be checked when measuring the train groups and occurring, and is acquired current
Train groups image, using acquired image as each frame of video image, carried out respectively from first image and tail an image from
Reason, the image that the image and the tailstock for finding headstock appearance occur, then judges specific position of the headstock in the image that headstock occurs
It sets and specific location of the tailstock in the image that the tailstock occurs, extracts the specific region of headstock and the tailstock.Determining headstock position
It sets with behind tailstock position, image after removing the image and the tailstock in train groups image before headstock, the purpose of this step is will to arrange
The image section unrelated with train removes in vehicle group image, obtains complete train group image.
After the headstock and the tailstock that identify train groups, need further to identify the compartment junction of train groups, the present invention is real
The train groups fault recognition method without default template for applying example offer is equally gone through without any when identifying compartment junction
History template library, and the train groups image for only needing currently to acquire can be completed.Train groups have the characteristics that it is certain, can be according to train
The characteristics of group itself, realizes the identification of compartment junction.By taking EMU as an example, EMU is by least two sections with driving force
Compartment and several sections are collectively constituted without the compartment of tractive force, wherein compartment of two sections with driving force is the dynamic of two section reverse symmetries
Power compartment, the i.e. headstock and the tailstock of EMU, and several sections are without the compartment that the compartment of tractive force is then among headstock and the tailstock,
Intermediate compartment is mutually symmetrical.Among any two section compartment, as compartment junction.In previous step by train groups
The image section unrelated with train removes in image, obtains complete train group image.Compartment, compartment junction and the back of EMU
The color image of scape be it is different, motor train unit carriage can be identified first on complete train group image, for it
His region can see whether the two sides image in the region is symmetrical, if right to differentiate compartment junction or background image
Claim, then it represents that the region be compartment junction, on the contrary it is then be background image.The compartment junction of every piece image is identified
Afterwards, image is spliced.If there is no compartment junction in the width image, by the width image direct splicing, if the width image
In have compartment junction, then half section of a left side for the width image is spliced as the section compartment, right half section is used as next section compartment
Starting point, in this way, complete EMU each section compartment splicing.
Behind the headstock tailstock and the compartment junction of identifying train groups, need to carry out fault identification, the embodiment of the present invention
In fault identification be still train groups image using collection in worksite, rather than default template.Specific operation is, to Mr. Yu
The possible faulty compartment of one section, its corresponding stitching image is taken out, as target compartment image, other compartments of train groups
Corresponding image, which is used as, refers to compartment image, carries out characteristic matching by target compartment image and with reference to compartment image, is matched
Value, matching value is compared with preset threshold.If matching value is higher than preset threshold, show the target compartment fault-free;Conversely,
Then show that the target compartment is faulty, according to matching result, exports location of fault.
Train groups fault recognition method provided in an embodiment of the present invention without default template, is not necessarily to any default headstock vehicle
Tail template library without identification rain model, but utilizes this column train image currently acquired, identifies the vehicle of train groups first
The head tailstock, then according to the correlation between image, the characteristic of image itself identifies the compartment junction of train groups, then with to
Other compartment images for detecting train groups carry out template matching, before without finding history car modal library according to rain model
Realization fault identification is put, real-time analysis and the automatic early-warning of the visual structure abnormal conditions of operation EMU are completed.Using same
The current acquired image of one train carries out fault identification, and it is external dry not only to largely avoid illumination, running speed etc.
The influence of factor is disturbed, can not also be influenced by the artificial service mode such as components position adjustment, efficiently solve current event
The rate of false alarm as caused by the external factors such as illumination, train running speed, the problems such as rate of failing to report is higher in barrier identification, improve dynamic
The efficiency of vehicle group fault identification and upkeep operation improves the discovery to the hidden failure of EMU in operational process and early warning energy
Power.
On the basis of the above embodiments, the headstock and the tailstock of the train groups are the headstock vehicles according to gauss hybrid models
Tail recognizer identifies to obtain.
The headstock tailstock recognizer of the gauss hybrid models, specifically includes:
Based on gauss hybrid models, learning model building is carried out to the image data of no train information, obtains background image;
In conjunction with the background image, background difference is carried out to the train groups image, obtains the image and vehicle of headstock appearance
The image that tail occurs;
The specific location that the headstock is determined in the image that the headstock occurs, extracts headstock image-region, described
The specific location that the tailstock is determined in the image that the tailstock occurs, extracts tailstock image-region.
Gauss model is exactly that Gaussian probability-density function (normal distribution curve) is used accurately to quantify things, by a things
It is decomposed into several models formed based on Gaussian probability-density function.Moving object detection extraction in, target context for
The identification and tracking of target are most important.And model the important link that exactly target context extracts.
The concept for lifting background and prospect is first had to, prospect refers to assuming that background is static, any intentional
The moving object of justice is prospect.The basic thought of modeling is that prospect is extracted from present frame, and the purpose is to keep background closer
The background of current video frame.It is weighted and averaged using the current background frame in present frame and video sequence to update background,
But due to the influence of illuminance abrupt variation and other external environments, the background after general modeling is not very clean clear, and
Gauss hybrid models (Gaussian Mixture Model, hereinafter referred to as GMM) are to model one of method the most successful, simultaneously
GMM can be used in monitor video and index and retrieve.
GMM characterizes the feature of each pixel in image using K (essentially 3 to 5) Gauss models, in a new frame
Image updates mixed Gauss model after obtaining, and is matched with each pixel in present image with mixed Gauss model, if at
Function then determines that the point is background dot, is otherwise foreground point.Several Gauss models predetermined are initialized first, to Gauss model
In parameter initialized, and parameter to be used after finding out.Secondly, being carried out for each of each frame pixel
Processing, sees whether it matches some model, if matching, is classified in the model, and to the model according to new pixel value
It being updated, if mismatching, a Gauss model, initiation parameter, instead of most can not in original model being established with the pixel
The model of energy.It finally selects the several most possible models in front as background model, lays the groundwork for target context extraction.
Fig. 2 is the flow diagram of the headstock tailstock recognizer of gauss hybrid models provided in an embodiment of the present invention, such as
Shown in Fig. 2, comprising:
Step 21, the image data without train information is obtained;
Step 22, background image is obtained to the image data learning model building of no train information based on gauss hybrid models;
Step 23, for headstock, successively progressively increase from first image and handled, for the tailstock, from last image
Successively successively decrease and is handled;
Step 24, in conjunction with background image, background difference is carried out using gauss hybrid models;
Step 25, after background difference, headstock image and tailstock image are judged whether there is, and if it exists, 26 are thened follow the steps,
If it does not exist, 23 are thened follow the steps;
Step 26, the specific location where headstock is exported, the specific location where the tailstock is exported.
The algorithm is not necessarily to default headstock tailstock template library, without identification rain model, but directly will currently acquire
The analogy of train groups image is operated for video frame, therefore firstly the need of background modeling, that is, when being directed to without acquiring any image,
Carry out background modeling;Subsequent background difference judges whether there is foreground image appearance, when foreground image occurs, positions foreground picture
The specific location of picture, so that it is determined that headstock/tailstock specific location.
Firstly, the image data without any train information that will acquire is modeled as background image based on GMM, make
For background image.
Then, respectively since first or the train image of last acquisition, background difference is carried out, is looked for whether
There is foreground image, if there is continuing to position foreground image position, and export the specific location of headstock and the tailstock.
Train groups fault recognition method provided in an embodiment of the present invention without default template, utilizes this column currently acquired
Train image carries out the identification of the headstock tailstock using gauss hybrid models, be not necessarily to any default headstock tailstock template library, also without
It need to identify rain model, then using the full column train image currently acquired, according to the correlation between image, image itself
Characteristic identifies the compartment junction of train groups, then carries out template matching with other compartment images of train groups to be detected, is being not necessarily to
Fault identification is realized under the premise of finding history car modal library according to rain model, the visual structure for completing operation EMU is different
The real-time analysis of reason condition and automatic early-warning carry out fault identification using the current acquired image of same train, not only very big
The influence of the external disturbing factor such as illumination, running speed is avoided in degree, it can not also be artificial by components position adjustment etc.
The influence of service mode efficiently solves in current failure identification since the external factors such as illumination, train running speed cause
Rate of false alarm, the problems such as rate of failing to report is higher, improve the efficiency of EMU fault identification and upkeep operation, improve to running
The discovery and pre-alerting ability of the hidden failure of EMU in journey.
On the basis of the above embodiments, the compartment junction of the train groups is connected according to the compartment of symmetrical image detection
The place's of connecing recognizer identifies to obtain.
Since the image that the headstock occurs, compartment company is carried out to each target image of the complete train group image
Place is met to determine;
To the either objective image in the complete train group image, if judgement knows that the either objective image includes vehicle
Compartment junction then splices the left-hand component of the compartment junction of the either objective image as this array carriage, described
Starting point of the right-hand component of the compartment junction of either objective image as next array carriage;
Otherwise, splice using the either objective image as complete piece image.
Each target image to the complete train group image carries out the judgement of compartment junction, specifically includes:
For each target image, whether judgement wherein has vertical black region, and vertical black region model
It encloses whether in the first preset threshold range;
For with the vertical black region and the vertical black region is in first preset threshold range
Target image judges whether target image image at left and right sides of the vertical black region is symmetrical;
If judgement knows that target image image at left and right sides of the vertical black region is symmetrical, the mesh
The vertical black region of logo image is compartment junction.
Fig. 3 is the flow diagram of the compartment junction recognizer of symmetrical image detection provided in an embodiment of the present invention,
As shown in Figure 3, comprising:
Step 31, train groups headstock image is detected;
Step 32, headstock image is judged whether there is, and if it exists, execute step 33, if it does not exist, execute step 31;
Step 33, according to headstock/tailstock extracting method, headstock and the tailstock position of train groups is cut out, complete column is obtained
Vehicle group image;
Step 34, continue to read complete train group image;
Step 35, judge whether, if not having, execute containing vertical black region if so, thening follow the steps 36 in image
Step 37;
Step 36, judge whether the vertical black region left and right sides is symmetrical, if symmetrically, thening follow the steps 38, if asymmetric,
Then follow the steps 37;
Step 37, image co-registration;
Step 38, compartment junction is contained in the image;
Step 39, it is exported the train image merged as an array carriage.
It is primarily based on above-mentioned collected train head's location drawing picture to start, compartment junction is carried out to each target image
Judgement left half section is spliced as the section compartment if either objective image includes compartment junction, right half section of work
For next section compartment starting point, as if not if entire image spliced.
The identification of compartment junction provided in an embodiment of the present invention is not necessarily to any history image template, without identifying train
Vehicle, but based on the train groups image to be detected currently acquired, it is identified according to the characteristics of train groups compartment.Compartment connects
The specific determination method in the place of connecing is to judge that either objective image whether there is vertical black region, and vertical black regional scope exists
In first preset threshold range.If it is present judge at left and right sides of vertical black region image whether mirror symmetry, if right
Claim, the black region of the target image is compartment junction, and otherwise, which does not include compartment junction.It does so
The reason is that train groups have the characteristics that it is certain, can be according to the identification for realizing compartment junction the characteristics of train groups itself.With dynamic
For vehicle group, EMU is to be collectively constituted by compartment of at least two sections with driving force and several sections without the compartment of tractive force,
In compartment of two sections with driving force be two section reverse symmetries power compartment, i.e. the headstock and the tailstock of EMU, and several sections are not
Compartment with tractive force is then the compartment among headstock and the tailstock, and intermediate compartment is mutually symmetrical.In any two section compartment
Centre, as compartment junction.According to the train groups image of acquisition, the compartment of EMU, compartment junction and background color
Image be it is different, motor train unit carriage can be identified first on train groups image, for other regions, in order to point
It distinguishes compartment junction or background image, can see whether the two sides image in the region is symmetrical, if symmetrically, then it represents that the region
It is on the contrary then be background image for compartment junction.It is spliced with the train image spliced, forms the compartment image.
It is provided in an embodiment of the present invention to be not necessarily to default template train groups fault recognition method, it is not necessarily to any default headstock tailstock
Template library without identification rain model, but utilizes this column train image currently acquired, identifies the headstock of train groups first
The tailstock, then according to the correlation between image, the characteristic of image itself identifies the compartment junction of train groups, then with it is to be checked
Other compartment images for surveying train groups carry out template matching, in the premise without finding history car modal library according to rain model
Lower realization fault identification, using the full column train image currently acquired, according to the correlation between image, the spy of image itself
Property, real-time analysis and the automatic early-warning of the visual structure abnormal conditions of operation EMU are completed, currently adopting for same train is utilized
Collect image and carry out fault identification, not only largely avoids the influence of the external disturbing factor such as illumination, running speed,
It can not be influenced, be efficiently solved in current failure identification due to light by the artificial service mode such as components position adjustment
According to rate of false alarm caused by the external factors such as, train running speed, rate of failing to report is higher the problems such as, improve EMU fault identification and
The efficiency of upkeep operation improves discovery and pre-alerting ability to the hidden failure of EMU in operational process.
On the basis of the above embodiments, it is described by target compartment image and with reference to compartment image carry out characteristic matching it
Before, the method also includes:
The Scale invariant features transform for extracting target compartment image describes sub- SIFT feature, obtains preset failure point;
According to the regional location of the preset failure point, the preliminary fault zone in target compartment image is determined.
The regional location according to the preset failure point, determines the preliminary faulty section in target compartment image
Domain is obtained by non-maxima suppression algorithm.
Scale invariant features transform (Scale-Invariant Feature Transform, hereinafter referred to as SIFT) is to use
In a kind of description of field of image processing.This description has scale invariability, can detect key point in the picture, is a kind of
Local feature description's.
SIFT feature detection mainly includes following 4 basic steps:
1, scale space extremum extracting searches for the picture position on all scales.It is identified by gaussian derivative function latent
The point of interest for scale and invariable rotary.
2, crucial point location determines position and ruler by the fine model of a fitting on the position of each candidate
Degree.The selection gist of key point is in their degree of stability.
3, direction determines, the gradient direction based on image local distributes to each key point position one or more direction.
All subsequent operations to image data are converted both relative to the direction of key point, scale and position, thus offer pair
In the invariance of these transformation.
4, key point describes, and in the neighborhood around each key point, the ladder of image local is measured on selected scale
Degree.
Non-maxima suppression (Non-Maximum Suppression, hereinafter referred to as NMS), as the term suggests it is exactly to inhibit not
It is the element of maximum, it can be understood as local maxima search.What this was locally represented is a neighborhood, and there are two parameters for neighborhood
It is variable, first is that the dimension of neighborhood, second is that the size of neighborhood.Such as in object detection, the extracted feature of sliding window is categorized
After device Classification and Identification, each window can obtain a classification and score.But sliding window will lead to many windows and other
There is the case where including or largely intersecting in window.At this moment needing to use NMS just to choose score highest in those neighborhoods (is
The maximum probability of certain class object), and the window for inhibiting those scores low.
In object detection, NMS non-maxima suppression its purpose is to eliminate extra frame, is looked for using very extensive
To the position of optimal object detection.
It is described to carry out characteristic matching by target compartment image and with reference to compartment image, identify target compartment abort situation, tool
Body includes:
According to position of the preliminary fault zone in target compartment image, sought in the reference compartment image
Corresponding position is looked for, characteristic matching is carried out with the preliminary fault zone, obtains matching threshold;
If judgement knows that the matching threshold less than the second preset threshold, contains faulty section in the preliminary fault zone
Domain exports the position of the fault zone.
Fig. 4 is the process signal for the train groups fault recognition method without default template that further embodiment of this invention provides
Figure, as shown in Figure 4, comprising:
Step 401, detect that train groups vehicle occurs;
Step 402, train groups image data is obtained;
Step 403, the Scale invariant features transform for extracting the train groups image currently acquired describes sub- SIFT feature;
Step 404, it is based on non-maxima suppression algorithm (NMS), determines the preliminary fault zone of target compartment image;
Step 405, target compartment image and other compartments of the train groups image currently acquired are subjected to object matching, obtained
To matching threshold;
Step 406, judge matching threshold whether in preset threshold range, that is, whether matching threshold is default less than second
Threshold value thens follow the steps 407 if being less than, if being not less than, thens follow the steps 408;
Step 407, without failure, continue reading image and judged, and execute step 402;
Step 408, there are failures, and the location information of fault zone is saved, and continue reading image and judged;
Step 409, tailstock image is judged whether it is, if so, 410 are executed, if it is not, then executing 402;
Step 410, terminate breakdown judge, export fault zone position.
For the complete train group image currently acquired, based on the position of the point-of-interests such as preset failure point, Primary Location
The region of failure is likely to occur in image;Then using other image datas in the non-this section compartment of this train as image template
Its corresponding position is found in these fault zones by library in image template library, if can illustrate the image with successful match
Illustrate to contain fault zone in the image, and position fault zone if matching threshold is very low for normal region.Tool
Body includes:
The Scale invariant features transform for extracting the train operation image currently acquired first describes sub- SIFT feature;
Then use non-maximum restraining value (NMS), according to the regional location of the point-of-interests such as preset failure point in image and
Area fraction positions the preliminary fault zone in image;
Finally by the preliminary region of failure, compared with the different compartments of current train image carry out matching, therefore, it is determined that the figure
It seem no to be truly present fault zone, and positioning failure regional location.
Train groups fault recognition method provided in an embodiment of the present invention without default template, is not necessarily to any history Prototype drawing
Picture, without identifying rain model, but according to the full column train image currently acquired, using SIFT and NMS come Primary Location event
Hinder region, further determine the specific location of failure, according to the correlation between image, the characteristic of image itself completes fortune
The real-time analysis of the visual structure abnormal conditions of row EMU and automatic early-warning are carried out using the current acquired image of same train
Fault identification not only largely avoids the influence of the external disturbing factor such as illumination, running speed, can not also be by zero
The influence of the artificial service mode such as component locations adjustment, efficiently solves in current failure identification due to illumination, train operation
Rate of false alarm caused by the external factors such as speed, the problems such as rate of failing to report is higher improve EMU fault identification and upkeep operation
Efficiency improves discovery and pre-alerting ability to the hidden failure of EMU in operational process.
Fig. 5 is the structural schematic diagram of the train groups fault identification device provided in an embodiment of the present invention without default template,
As shown in figure 5, including first processing module 51, Second processing module 52, splicing module 53 and identification module 54, in which:
First processing module 51 is used for the train groups image according to acquisition, identifies the headstock and the tailstock of train groups, removes institute
Image after stating the image and the tailstock in the train groups image of acquisition before headstock obtains complete train group image;
Second processing module 52 is used to identify the compartment junction of train groups according to the complete train group image;
Splicing module 53 is used to splice the complete train group image according to the compartment junction, be arranged
The complete image in each section compartment of vehicle group;
Identification module 54 is for carrying out characteristic matching, the event of identification target compartment by target compartment image and with reference to compartment image
Hinder position, wherein the target compartment is compartment to be detected, is its in train groups in addition to the target compartment with reference to compartment
His compartment.
Existing train groups fault recognition method be based on a large amount of headstock tailstock preset template library, fault identification it
Before, need to acquire the image of various types of train groups, in fault identification according to the model of train groups to be detected come
Its headstock tailstock is detected, and corresponding history template is found according to the coach number of train groups, thus further determine that abort situation,
And the train groups fault identification device provided in an embodiment of the present invention without default template, it is not necessarily to any history template image, only
It needs according to the train groups image currently acquired, the headstock of automatic identification train groups and tailstock position, and will be in acquisition image
Image after image and the tailstock before being not belonging to the headstock appearance of train image occur removes;According to modified complete train groups
Image identifies the compartment junction of train groups, and according to compartment junction position, splicing acquisition image, thus by the figure of acquisition
As carrying out splicing segmentation with compartment serial number, the complete image in each section compartment is formed.
Specific operation is train groups to be detected for one group, to be checked when measuring the train groups and occurring, and acquires train groups
Image, first processing module 51 is using acquired image as each frame of video image, respectively from first image and tail figure
As being handled, then the image that the image and the tailstock for finding headstock appearance occur judges headstock in the image that headstock occurs
Specific location in the image that the tailstock occurs of specific location and the tailstock, extract the specific region of headstock and the tailstock.True
After determining headstock position and tailstock position, image after removing the image and the tailstock in train groups image before headstock, the mesh of this step
Be to remove image section unrelated with train in train groups image, obtain complete train group image.
After the headstock and the tailstock that identify train groups, need further to identify the compartment junction of train groups, the present invention is real
The train groups fault identification device without default template for applying example offer is equally gone through without any when identifying compartment junction
History template library, and the train groups image for only needing currently to acquire can be completed.Train groups have the characteristics that certain, second processing mould
Block 52 can be according to the identification for realizing compartment junction the characteristics of train groups itself.By taking EMU as an example, EMU is by least
Compartment of two sections with driving force and several sections are collectively constituted without the compartment of tractive force, wherein compartment of two sections with driving force is two
Save the power compartment of reverse symmetry, i.e. the headstock and the tailstock of EMU, and several sections without the compartment of tractive force be then headstock and
Compartment among the tailstock, intermediate compartment are mutually symmetrical.Among any two section compartment, as compartment junction.Upper
One step removes image section unrelated with train in train groups image, obtains complete train group image.The vehicle of EMU
The color image in compartment, compartment junction and background be it is different, can be on complete train group image first EMU car
Compartment identifies, for other regions, in order to differentiate compartment junction or background image, can see the two sides figure in the region
Seem it is no symmetrical, if symmetrically, then it represents that on the contrary the region is compartment junction, then be background image.Splicing module 53 will be each
After the compartment junction of width image identifies, image is spliced.If there is no compartment junction in the width image, splice
Module 53 is by the width image direct splicing, if there is compartment junction in the width image, splicing module 53 is by a left side for the width image
Half section is spliced as the array carriage, right half section of starting point as next section compartment, in this way, completing each section vehicle of EMU
The splicing in compartment.
Behind the headstock tailstock and the compartment junction of identifying train groups, need to carry out fault identification, the embodiment of the present invention
In fault identification be still train groups image using collection in worksite, rather than default template.Specific operation is to identify mould
The faulty compartment possible for a certain section of block 54, its corresponding stitching image is taken out, as target compartment image, train
The corresponding image in other compartments of group is used as, with reference to compartment image, to carry out feature by target compartment image and with reference to compartment image
Match, obtain matching value, matching value is compared with preset threshold.If matching value is higher than preset threshold, show the target compartment
Fault-free;Conversely, then showing that the target compartment is faulty, according to matching result, location of fault is exported.
Device provided in an embodiment of the present invention is for executing above-mentioned each method embodiment, specific process and in detail Jie
It continues and refers to above-mentioned each method embodiment, details are not described herein again.
Train groups fault identification device provided in an embodiment of the present invention without default template, is not necessarily to any default headstock vehicle
Tail template library without identification rain model, but utilizes this column train image currently acquired, identifies the vehicle of train groups first
The head tailstock, then according to the correlation between image, the characteristic of image itself identifies the compartment junction of train groups, then with to
Other compartment images for detecting train groups carry out template matching, before without finding history car modal library according to rain model
Realization fault identification is put, real-time analysis and the automatic early-warning of the visual structure abnormal conditions of operation EMU are completed.Using same
The current acquired image of one train carries out fault identification, and it is external dry not only to largely avoid illumination, running speed etc.
The influence of factor is disturbed, can not also be influenced by the artificial service mode such as components position adjustment, efficiently solve current event
The rate of false alarm as caused by the external factors such as illumination, train running speed, the problems such as rate of failing to report is higher in barrier identification, improve dynamic
The efficiency of vehicle group fault identification and upkeep operation improves the discovery to the hidden failure of EMU in operational process and early warning energy
Power.
Fig. 6 illustrates the entity structure schematic diagram of a kind of electronic equipment, as shown in fig. 6, the electronic equipment may include: place
Manage device (processor) 610, communication interface (Communications Interface) 620,630 He of memory (memory)
Bus 640, wherein processor 610, communication interface 620, memory 630 complete mutual communication by bus 640.Bus
640 can be used for the information transmission between electronic equipment and sensor.Processor 610 can call the logic in memory 630
Instruction, to execute following method: according to the train groups image of acquisition, identifying the headstock and the tailstock of train groups, remove the acquisition
Train groups image in image after image and the tailstock before headstock, obtain complete train group image;According to the complete train
Group image, identifies the compartment junction of train groups;According to the compartment junction, the complete train group image is spelled
It connects, obtains the complete image in each section compartment of train groups;Characteristic matching is carried out by target compartment image and with reference to compartment image,
Identify target compartment abort situation, wherein the target compartment is compartment to be detected, is in train groups except described with reference to compartment
Other compartments outside target compartment.
In addition, the logical order in above-mentioned memory 630 can be realized by way of SFU software functional unit and conduct
Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally
Substantially the part of the part that contributes to existing technology or the technical solution can be in other words for the technical solution of invention
The form of software product embodies, which is stored in a storage medium, including some instructions to
So that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation of the present invention
The all or part of the steps of example the method.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (ROM,
Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. it is various
It can store the medium of program code.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, the non-transient computer readable storage medium
Matter stores computer instruction, and it is a kind of without presetting template provided by above-described embodiment which execute computer
Train groups fault recognition method, for example, according to the train groups image of acquisition, identify the headstock and the tailstock of train groups, remove
The image after image and the tailstock in the train groups image of the acquisition before headstock, obtains complete train group image;According to described
Complete train group image, identifies the compartment junction of train groups;According to the compartment junction, to the complete train group image
Spliced, obtains the complete image in each section compartment of train groups;It is carried out by target compartment image and with reference to compartment image special
Sign matching, identifies target compartment abort situation, wherein the target compartment is compartment to be detected, is train groups with reference to compartment
In other compartments in addition to the target compartment.
The above description is only a preferred embodiment of the present invention, is not intended to restrict the invention.The technical field of the invention
Technical staff can make various modifications or additions to the described embodiments, but without departing from of the invention
Spirit surmounts the range that the appended claims define.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, and those skilled in the art is it is understood that it still can be right
Technical solution documented by foregoing embodiments is modified or equivalent replacement of some of the technical features;And this
It modifies or replaces, the spirit and model of technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (10)
1. a kind of train groups fault recognition method without presetting template characterized by comprising
According to the train groups image of acquisition, identifies the headstock and the tailstock of train groups, remove vehicle in the train groups image of the acquisition
The image after image and the tailstock before head, obtains complete train group image;
According to the complete train group image, the compartment junction of train groups is identified;
According to the compartment junction, the complete train group image is spliced, obtains each section compartment of train groups
Complete image;
Characteristic matching is carried out by target compartment image and with reference to compartment image, identifies target compartment abort situation, wherein the mesh
Mark compartment is compartment to be detected, is other compartments in train groups in addition to the target compartment with reference to compartment.
2. the method according to claim 1, wherein the headstock and the tailstock of the train groups are according to Gaussian Mixture
The headstock tailstock recognizer of model identifies to obtain, and specifically includes:
Based on gauss hybrid models, learning model building is carried out to the image data of no train information, obtains background image;
In conjunction with the background image, background difference is carried out to the train groups image, the image and the tailstock for obtaining headstock appearance go out
Existing image;
The specific location that the headstock is determined in the image that the headstock occurs, extracts headstock image-region, in the tailstock
The specific location that the tailstock is determined in the image of appearance, extracts tailstock image-region.
3. according to the method described in claim 2, it is characterized in that, the compartment junction of the train groups is according to symmetrical image
The compartment junction recognizer of detection identifies to obtain, and specifically includes:
Since the image that the headstock occurs, compartment junction is carried out to each target image of the complete train group image
Determine;
To the either objective image in the complete train group image, if judgement knows that the either objective image includes that compartment connects
Place is met, then is spliced the left-hand component of the compartment junction of the either objective image as this array carriage, it is described any
Starting point of the right-hand component of the compartment junction of target image as next array carriage;
Otherwise, splice using the either objective image as complete piece image.
4. according to the method described in claim 3, it is characterized in that, each target figure to the complete train group image
As progress compartment junction judgement, specifically include:
For each target image, whether judgement wherein has vertical black region, and the vertical black regional scope is
It is no in the first preset threshold range;
For the target with the vertical black region and the vertical black region in first preset threshold range
Image judges whether target image image at left and right sides of the vertical black region is symmetrical;
If judgement knows that target image image at left and right sides of the vertical black region is symmetrical, the target figure
The vertical black region of picture is compartment junction.
5. the method according to claim 1, wherein it is described by target compartment image and with reference to compartment image into
Before row characteristic matching, the method also includes:
The Scale invariant features transform for extracting target compartment image describes sub- SIFT feature, obtains preset failure point;
According to the regional location of the preset failure point, the preliminary fault zone in target compartment image is determined.
6. according to the method described in claim 5, it is characterized in that, the regional location according to the preset failure point, really
Preliminary fault zone in fixed target compartment image, is obtained by non-maxima suppression algorithm.
7. method according to claim 5 or 6, which is characterized in that described by target compartment image and with reference to compartment image
Characteristic matching is carried out, target compartment abort situation is identified, specifically includes:
According to position of the preliminary fault zone in target compartment image, phase is found in the reference compartment image
The position answered carries out characteristic matching with the preliminary fault zone, obtains matching threshold;
If judgement knows that the matching threshold less than the second preset threshold, contains fault zone in the preliminary fault zone,
Export the position of the fault zone.
8. a kind of train groups fault identification device without presetting template characterized by comprising
First processing module identifies the headstock and the tailstock of train groups, removes the acquisition for the train groups image according to acquisition
Train groups image in image after image and the tailstock before headstock, obtain complete train group image;
Second processing module, for identifying the compartment junction of train groups according to the complete train group image;
Splicing module, for splicing to the complete train group image, obtaining train groups according to the compartment junction
The complete image in each section compartment;
Identification module identifies target compartment fault bit for carrying out characteristic matching by target compartment image and with reference to compartment image
It sets, wherein the target compartment is compartment to be detected, is other vehicles in train groups in addition to the target compartment with reference to compartment
Compartment.
9. a kind of electronic equipment, which is characterized in that including memory and processor, the processor and the memory pass through always
Line completes mutual communication;The memory is stored with the program instruction that can be executed by the processor, the processor tune
The method as described in claim 1 to 7 is any is able to carry out with described program instruction.
10. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer
The train groups fault identification side as described in any one of claim 1 to 7 without default template is realized when program is executed by processor
Method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811320495.0A CN109740410B (en) | 2018-11-07 | 2018-11-07 | Train set fault identification method and device without preset template |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811320495.0A CN109740410B (en) | 2018-11-07 | 2018-11-07 | Train set fault identification method and device without preset template |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109740410A true CN109740410A (en) | 2019-05-10 |
CN109740410B CN109740410B (en) | 2020-11-13 |
Family
ID=66355666
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811320495.0A Active CN109740410B (en) | 2018-11-07 | 2018-11-07 | Train set fault identification method and device without preset template |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109740410B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110276258A (en) * | 2019-05-21 | 2019-09-24 | 智慧互通科技有限公司 | A kind of method and system identifying vehicle appearance |
CN110285870A (en) * | 2019-07-22 | 2019-09-27 | 深圳市卓城科技有限公司 | Vehicle spindle-type and wheel number determination method and its system |
CN111860384A (en) * | 2020-07-27 | 2020-10-30 | 上海福赛特智能科技有限公司 | Vehicle type recognition method |
CN111862623A (en) * | 2020-07-27 | 2020-10-30 | 上海福赛特智能科技有限公司 | Vehicle side map splicing device and method |
CN111860303A (en) * | 2020-07-17 | 2020-10-30 | 成都铁安科技有限责任公司 | Train pantograph image acquisition system and control method thereof |
CN112183786A (en) * | 2020-10-19 | 2021-01-05 | 邦邦汽车销售服务(北京)有限公司 | Method and device for determining damaged part of vehicle |
CN113837160A (en) * | 2021-11-29 | 2021-12-24 | 天津市中环系统工程有限责任公司 | Method for identifying normally-living people and vehicles without preset information |
CN115063903A (en) * | 2022-06-08 | 2022-09-16 | 深圳市永达电子信息股份有限公司 | Railway freight train compartment body abnormity monitoring method and system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0579133A1 (en) * | 1992-07-15 | 1994-01-19 | Hitachi, Ltd. | Superconducting magnet abnormality detection and protection apparatus |
EP0800153A3 (en) * | 1996-04-01 | 1998-05-13 | Hamamatsu Photonics K.K. | Smoke detecting apparatus and method |
CN104463235A (en) * | 2014-11-18 | 2015-03-25 | 中国铁道科学研究院电子计算技术研究所 | Fault recognition method and device based on operation images of motor train unit |
CN105787925A (en) * | 2016-02-03 | 2016-07-20 | 中国科学院空间应用工程与技术中心 | Push-scan type optical remote sensing load original image bad line automatic detection method and system |
-
2018
- 2018-11-07 CN CN201811320495.0A patent/CN109740410B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0579133A1 (en) * | 1992-07-15 | 1994-01-19 | Hitachi, Ltd. | Superconducting magnet abnormality detection and protection apparatus |
EP0800153A3 (en) * | 1996-04-01 | 1998-05-13 | Hamamatsu Photonics K.K. | Smoke detecting apparatus and method |
CN104463235A (en) * | 2014-11-18 | 2015-03-25 | 中国铁道科学研究院电子计算技术研究所 | Fault recognition method and device based on operation images of motor train unit |
CN105787925A (en) * | 2016-02-03 | 2016-07-20 | 中国科学院空间应用工程与技术中心 | Push-scan type optical remote sensing load original image bad line automatic detection method and system |
Non-Patent Citations (2)
Title |
---|
GIUSEPPE LISANTI ET AL.: "A multi-camera image processing and visualization system for train safety assessment", 《ARXIV》 * |
张望等: "动车组运行状态智能检测装备", 《铁路计算机应用》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110276258A (en) * | 2019-05-21 | 2019-09-24 | 智慧互通科技有限公司 | A kind of method and system identifying vehicle appearance |
CN110285870A (en) * | 2019-07-22 | 2019-09-27 | 深圳市卓城科技有限公司 | Vehicle spindle-type and wheel number determination method and its system |
CN111860303A (en) * | 2020-07-17 | 2020-10-30 | 成都铁安科技有限责任公司 | Train pantograph image acquisition system and control method thereof |
CN111860384A (en) * | 2020-07-27 | 2020-10-30 | 上海福赛特智能科技有限公司 | Vehicle type recognition method |
CN111862623A (en) * | 2020-07-27 | 2020-10-30 | 上海福赛特智能科技有限公司 | Vehicle side map splicing device and method |
CN111860384B (en) * | 2020-07-27 | 2021-12-14 | 上海福赛特智能科技有限公司 | Vehicle type recognition method |
CN112183786A (en) * | 2020-10-19 | 2021-01-05 | 邦邦汽车销售服务(北京)有限公司 | Method and device for determining damaged part of vehicle |
CN113837160A (en) * | 2021-11-29 | 2021-12-24 | 天津市中环系统工程有限责任公司 | Method for identifying normally-living people and vehicles without preset information |
CN113837160B (en) * | 2021-11-29 | 2022-04-22 | 天津市中环系统工程有限责任公司 | Method for identifying normally-living people and vehicles without preset information |
CN115063903A (en) * | 2022-06-08 | 2022-09-16 | 深圳市永达电子信息股份有限公司 | Railway freight train compartment body abnormity monitoring method and system |
Also Published As
Publication number | Publication date |
---|---|
CN109740410B (en) | 2020-11-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109740410A (en) | A kind of train groups fault recognition method and device without presetting template | |
CN104463235B (en) | Fault recognition method and device based on EMU operation image | |
CN108960266A (en) | Image object detection method and device | |
US10984588B2 (en) | Obstacle distribution simulation method and device based on multiple models, and storage medium | |
CN105160309A (en) | Three-lane detection method based on image morphological segmentation and region growing | |
Hadjidemetriou et al. | Automated detection of pavement patches utilizing support vector machine classification | |
CN105243381B (en) | Failure automatic identification detection system and method based on 3D information | |
CN113052159B (en) | Image recognition method, device, equipment and computer storage medium | |
JP6595375B2 (en) | Traffic condition analysis device, traffic condition analysis method, and traffic condition analysis program | |
Nguyen et al. | Vehicle re-identification with learned representation and spatial verification and abnormality detection with multi-adaptive vehicle detectors for traffic video analysis. | |
CN107909012B (en) | Real-time vehicle tracking detection method and device based on disparity map | |
CN109902610A (en) | Traffic sign recognition method and device | |
CN113808098A (en) | Road disease identification method and device, electronic equipment and readable storage medium | |
CN109427191A (en) | A kind of traveling detection method and device | |
CN113436157A (en) | Vehicle-mounted image identification method for pantograph fault | |
CN109740609A (en) | A kind of gauge detection method and device | |
CN113159024A (en) | License plate recognition technology based on improved YOLOv4 | |
CN110008900A (en) | A kind of visible remote sensing image candidate target extracting method by region to target | |
CN115439750A (en) | Road disease detection method and device, electronic equipment and storage medium | |
CN113393442B (en) | Train part abnormality detection method, system, electronic equipment and storage medium | |
CN112669615B (en) | Parking space detection method and system based on camera | |
Saha et al. | Road rutting detection using deep learning on images | |
CN115995075A (en) | Vehicle self-adaptive navigation method and device, electronic equipment and storage medium | |
US20230342937A1 (en) | Vehicle image analysis | |
CN115718702A (en) | Automatic driving test scene library construction 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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |