CN107146224B - The online image detecting system of train wheel thread defect and method - Google Patents
The online image detecting system of train wheel thread defect and method Download PDFInfo
- Publication number
- CN107146224B CN107146224B CN201710305046.8A CN201710305046A CN107146224B CN 107146224 B CN107146224 B CN 107146224B CN 201710305046 A CN201710305046 A CN 201710305046A CN 107146224 B CN107146224 B CN 107146224B
- Authority
- CN
- China
- Prior art keywords
- image
- tread
- wheel
- acquisition device
- image acquisition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000007547 defect Effects 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000003709 image segmentation Methods 0.000 claims abstract description 7
- 230000015556 catabolic process Effects 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims description 11
- 230000004927 fusion Effects 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 4
- 239000003550 marker Substances 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000004323 axial length Effects 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000002203 pretreatment Methods 0.000 claims description 2
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract description 13
- 230000002950 deficient Effects 0.000 abstract description 8
- 238000003672 processing method Methods 0.000 abstract description 2
- 238000004381 surface treatment Methods 0.000 abstract 1
- 238000013459 approach Methods 0.000 description 21
- 238000010586 diagram Methods 0.000 description 9
- 238000005096 rolling process Methods 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 5
- 238000005520 cutting process Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 230000003137 locomotive effect Effects 0.000 description 5
- 238000013461 design Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 2
- PCTMTFRHKVHKIS-BMFZQQSSSA-N (1s,3r,4e,6e,8e,10e,12e,14e,16e,18s,19r,20r,21s,25r,27r,30r,31r,33s,35r,37s,38r)-3-[(2r,3s,4s,5s,6r)-4-amino-3,5-dihydroxy-6-methyloxan-2-yl]oxy-19,25,27,30,31,33,35,37-octahydroxy-18,20,21-trimethyl-23-oxo-22,39-dioxabicyclo[33.3.1]nonatriaconta-4,6,8,10 Chemical compound C1C=C2C[C@@H](OS(O)(=O)=O)CC[C@]2(C)[C@@H]2[C@@H]1[C@@H]1CC[C@H]([C@H](C)CCCC(C)C)[C@@]1(C)CC2.O[C@H]1[C@@H](N)[C@H](O)[C@@H](C)O[C@H]1O[C@H]1/C=C/C=C/C=C/C=C/C=C/C=C/C=C/[C@H](C)[C@@H](O)[C@@H](C)[C@H](C)OC(=O)C[C@H](O)C[C@H](O)CC[C@@H](O)[C@H](O)C[C@H](O)C[C@](O)(C[C@H](O)[C@H]2C(O)=O)O[C@H]2C1 PCTMTFRHKVHKIS-BMFZQQSSSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
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
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- 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
- 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
- G06T2207/30236—Traffic on road, railway or crossing
-
- 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
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of online image detecting system of train wheel thread defect and methods, are related to image processing method technical field.Described method includes following steps：Wheel is to image zooming-out；Image segmentation；Image rectification；Image mosaic；Image calibration；Tread whether there is breakdown judge；Tread failure is extracted：If judging existing defects failure in tread image by upper surface treatment, thread defect region is then isolated from tread image with the method that seed point growth is combined using region growing, and the dimension information of thread defect part is calculated using the geometrical correspondence of image calibration.Spliced and obtained after being handled each complete wheel tread information to image by the method, and the tread is carried out judging to obtain whether defective, thread defect accuracy of detection is high, can effectively avoid detection mistake.
Description
Technical field
The present invention relates to image processing method and systems technology field more particularly to a kind of train wheel thread defect are online
Image detecting system and method.
Background technology
Wheel is fitted in same root vehicle securely to being the part being in contact with rail on rolling stock, by two wheels in left and right
It is formed on axis.The effect of wheel pair is the operation and steering for ensureing rolling stock on rail, is born from the complete of rolling stock
Portion is quiet, dynamic loading, passes it on rail, and give rolling stock each parts the load transmission generated by guideway irregularity.
Wheel tread defect is one of the major failure type of wheel pair, is developed with the high speed and heavy loading of railway, thread defect failure
Appearance it is increasingly frequent.There is also many problems, major defect is to examine the online test method of thread defect on railway at present
It is poor to survey precision, can not correctly identify thread defect.
Invention content
The technical problem to be solved by the present invention is to how provide a kind of train wheel that thread defect accuracy of detection is high to step on
The online image detecting system of planar defect.
In order to solve the above technical problems, the technical solution used in the present invention is：A kind of train wheel thread defect is online
Image detecting system, it is characterised in that：Including computer, four image acquisition devices and four approach switch sensors, first
Image acquisition device and third image acquisition device are located at the outside of the first rail, the second image acquisition device and the 4th image acquisition device position
In the outside of the second rail, the first image acquisition device and third image acquisition device and the second image acquisition device and the 4th Image Acquisition
Using the rail center line between two rails as symmetry axis between device, between the first image acquisition device and third image acquisition device and
Using image acquisition device center line as symmetry axis between second image acquisition device and the 4th image acquisition device；Each image acquisition device corresponds to
One train wheel, the image for shooting corresponding wheel, the circle of the central axes and the wheel of each described image collector
The heart is in same level, and the horizontal field of view angle of described image collector is all θ, and the first image acquisition device and the second image are adopted
Storage is located at the right side of locomotive, and third image acquisition device and the 4th image acquisition device are located at the left side of locomotive, the first Image Acquisition
The horizontal field of view of device and third image acquisition device, which has, to partly overlap, the level of the second image acquisition device and the 4th image acquisition device
Visual field, which has, to partly overlap, and four approach switch sensors are equally spaced to be set to the second image acquisition device and the 4th image
The inside of equitant second rail of collector horizontal field of view, and the distance between approach switch sensor is 0.785φ, φFor
The rolling circular diameter of the wheel；When the second wheel triggers the first approach switch sensor, the first wheel is located at the first image
In the horizontal field of view of collector, third wheel is located in the horizontal field of view of third image acquisition device, and the second wheel is located at the second figure
As collector horizontal field of view in, the 4th wheel is located in the horizontal field of view of the 4th image acquisition device, four image acquisition devices point
Image Acquisition is not carried out to corresponding wheel not for the first time；When the second wheel triggers the second approach switch sensor, the first wheel
In the horizontal field of view of the first image acquisition device, third wheel is located in the horizontal field of view of third image acquisition device, the second vehicle
Wheel is in the horizontal field of view of the second image acquisition device, and the 4th wheel is located in the horizontal field of view of the 4th image acquisition device, four
Image acquisition device carries out Image Acquisition second to corresponding wheel respectively；When the second wheel triggers third approach switch sensor
When, the first wheel is located in the horizontal field of view of the first image acquisition device, and the level that third wheel is located at third image acquisition device regards
In, the second wheel is located in the horizontal field of view of the second image acquisition device, and the 4th wheel is located at the level of the 4th image acquisition device
In visual field, third time carries out Image Acquisition to four image acquisition devices to corresponding wheel respectively；When the second wheel triggering the 4th connects
When nearly switch sensor, the first wheel is located in the horizontal field of view of the first image acquisition device, and third wheel is located at third image and adopts
In the horizontal field of view of storage, the second wheel is located in the horizontal field of view of the second image acquisition device, and the 4th wheel is located at the 4th image
In the horizontal field of view of collector, four image acquisition devices distinguish the 4th time and carry out Image Acquisition to corresponding wheel；Computer connects
Four images for receiving the transmission of each image acquisition device splice image and obtain each complete wheel tread after being handled
Information, and the tread is carried out judging to obtain whether defective.
Preferably：Horizontal distance between described first image collector and third image acquisition device and the first rail is
2m, the horizontal distance between second image acquisition device and the 4th image acquisition device and the second rail are 2m.
Preferably：Horizontal field of view angle θ is 65 °.
Preferably：Described image collector is fixed by the bracket by the rail.
Preferably：Described image collector is high-speed cmos video camera.
The invention also discloses a kind of online image detecting methods of train wheel thread defect, it is characterised in that including as follows
Step：
Wheel is to image zooming-out：Wheel is extracted from the whole image that image acquisition device acquires to region；
Image segmentation：Wheel is split image using improved Based On Method of Labeling Watershed Algorithm, extracts tread image；
Image rectification：The transverse and longitudinal coordinate for dividing the image into the tread image extracted carries out geometric correction, and tight to distorting
The inoperative side of weight is cut, and tread expanded view is obtained；
Image mosaic：According to last acquisition moment tread image size, using cube convolution method to first three moment tread figure
As obtaining the unified tread image of size into row interpolation, using block matching method to the aforementioned four tread image adjacent acquisition moment
Sequentially splicing fusion two-by-two is carried out, the complete tread image of each wheel is obtained；
Image calibration：User's case marker determines paper and is placed in tread respectively to acquire in moment camera photographable position, completes to splicing
Tread image afterwards is demarcated, the correspondence after being spliced between the pixel distance and dimension information of image；
Tread whether there is breakdown judge：Splice comentropy and the empirical value of rear tread pattern image relatively by calculating to sentence
Disconnected tread whether there is major defect failure；
Tread failure is extracted：Using improved splitting and merging to judging that the tread of existing defects is handled, utilize
Seed point growth is filled the thread defect image after growth, obtains thread defect image, and grid is actually used to defect
Method is estimated that area is calculated using pixel in region in image, and the value of the Gao Yukuan of pixel is to utilize figure respectively in image
The max min of the row and column in the thread defect region as in seeks difference to be calculated.
Further technical solution is that the method further includes before image segmentation：Using MSRCR algorithms to wheel pair
The pre-treatment step that the color information of image optimizes.
Further technical solution is that the wheel is as follows to the method for image zooming-out：
Input pending image；
Pending image is filtered, uses canny operators to carry out edge detection, edge pixel point after grayscale equalization
As characteristic point；
[500,1000] W ∈ in scanning figure, H ∈ [1000,1600] region, statistical nature point establish accumulation structure, add up
Device uses two-dimensional array, two-dimensional array subscript to be characterized dot center's coordinate, and accumulator initial value is set as 0；
Each characteristic point in range is traversed, string midpoint Hough transform is carried out and scans other characteristic points, acquire in them
Point, corresponding accumulator add 1, and only string midpoint Hough transform is carried out to untreated characteristic point when calculating next characteristic point；
When accumulator count value is more than threshold value 5, then it is assumed that there are an ellipse, its corresponding parameters for this position（x0 i, y0 i）
It is exactly oval corresponding centre coordinate, subscript i indicates oval number；
Use each point（x0 i, y0 i）Elliptic equation is moved into origin, is found about origin symmetry under new coordinate system
3 pairs of characteristic points acquire each elliptic equation, compare transverse size, and it is the length of side to select the longest ellipse of axial length, and elliptical center is
The square region of the centre of form is as wheel to region；
Interception wheel carries out subsequent processing as final wheel to positive 120 ° of the region of image to image.
Further technical solution is that the block matching method is as follows：
Two adjacent time charts are sequentially loaded into as matrix；
Determine image block scanning window and threshold values；
Start block scan, calculates absolute value difference value between image-region；
Judge whether the difference value is more than threshold value, if it exceeds threshold value then continues to scan on, if being not above threshold value
Storage location；
Curve Matching is carried out, splicing result is exported after being weighted fusion to the curve after matching.
Further technical solution is that the improved splitting and merging includes the following steps：
Setting judges sentence P and minimum unit size min, builds discriminant function；
Judge whether each region of image meets discriminant function；
It is considered as if meeting discriminant function and meets splitting condition, the sub-district area image four etc. of splitting condition will be met
Point, and image is judged again；If being unsatisfactory for discriminant function, merging is set and judges sentence Q, and is judged by merging
Sentence Q judges whether have subregion to meet merging condition in image, if it is closes the subregion for meeting merging condition
And if not then terminating.
It is using advantageous effect caused by above-mentioned technical proposal：The system Computer receives each Image Acquisition
Four images of device transmission, then image is spliced by the method and obtains each complete wheel tread after being handled
Information, and the tread is carried out judging to obtain whether defective, thread defect accuracy of detection is high, can effectively avoid detection wrong
Accidentally.
Description of the drawings
Fig. 1 is the schematic layout diagram of system described in the embodiment of the present invention；
Fig. 2 is that an image acquisition device is illustrated with the side view structure of corresponding wheel in system described in the embodiment of the present invention
Figure；
Fig. 3 is approach switch sensor setting principle schematic diagram in system described in the embodiment of the present invention；
Fig. 4 is the flow chart of the method for the embodiment of the present invention；
Fig. 5 is the flow chart taken turns in the method for the embodiment of the present invention to image zooming-out；
Fig. 6-9 is four sub-pictures that the second image acquisition device acquires in the method for the embodiment of the present invention；
Figure 10-13 is the figure after Fig. 6-9 cuttings in the method for the embodiment of the present invention；
Figure 14 is the result figure of ellipses detection in the method for the embodiment of the present invention；
Figure 15 is to be taken turns in the method for the embodiment of the present invention to intercepting partial schematic diagram；
Figure 16-17 is to be enhanced comparison diagram before and after the processing to image to taking turns with MSRCR algorithms；
Figure 18-19 is the design sketch after being split to image to wheel using improved Based On Method of Labeling Watershed Algorithm；
Figure 20 is the tread expanded view in the method for the embodiment of the present invention；
Figure 21-24 is that tread image obtains the unified tread figure of size into row interpolation in the method for the embodiment of the present invention
Picture；
Figure 25 is to splice flow chart in the embodiment of the present invention；
Figure 26 is splicing result figure in the embodiment of the present invention:
Figure 27-28 is to divide tread image and plane rectification figure in the embodiment of the present invention；
Figure 29 is defective tread image processing flow figure in the embodiment of the present invention；
Figure 30 is the result figure handled using improved splitting and merging in the embodiment of the present invention；
Figure 31 is thread defect image in the embodiment of the present invention；
Wherein：1, the first image acquisition device 2, the second image acquisition device 3, third image acquisition device 4, the 4th image acquisition device
5, the first rail 6, the second rail 7, rail center line 8, image acquisition device center line 9, the first wheel 10, the first approach switch sensor
11, the second wheel 12, third wheel 13, the 4th wheel 14, the second approach switch sensor 15, three approach switch sensors 16,
The effective coverage of tread portions in 4th approach switch sensor 17, holder 18, wheel 19, rail 20, every image.
Specific implementation mode
With reference to the attached drawing in the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Ground describes, it is clear that described embodiment is only a part of the embodiment of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Many details are elaborated in the following description to facilitate a thorough understanding of the present invention, still the present invention can be with
Implemented different from other manner described here using other, those skilled in the art can be without prejudice to intension of the present invention
In the case of do similar popularization, therefore the present invention is not limited by following public specific embodiment.
Embodiment one
As shown in Figure 1, present embodiment discloses a kind of online image detecting system of train wheel thread defect, including calculate
Machine, four image acquisition devices and four approach switch sensors.First image acquisition device 1 and third image acquisition device 3 are located at
The outside of first rail 5, the second image acquisition device 2 and the 4th image acquisition device 4 are located at the outside of the second rail 6.First image
With between two articles of rails between collector 1 and third image acquisition device 3 and the second image acquisition device 2 and the 4th image acquisition device 4
Rail center line 7（Horizontal dotted line shown in Fig. 1）For symmetry axis；Between first image acquisition device 1 and third image acquisition device 3 and
With image acquisition device center line 8 between two image acquisition devices 2 and the 4th image acquisition device 4（Described image collector center line refers to
The first midpoint and the second image acquisition device between one image acquisition device and third image acquisition device and the 4th image acquisition device
Between the second midpoint between line, shown in Fig. 1 indulge dotted line）For symmetry axis；Each image acquisition device corresponds to a train
Wheel, the image for shooting corresponding wheel, the central axes of each described image collector are with the circle centre position of the wheel same
On one horizontal plane, the horizontal field of view angle of described image collector is all θ.
First image acquisition device 1 and the second image acquisition device are located at the right side of locomotive, third image acquisition device 3 and the 4th figure
As collector 4 is located at the left side of locomotive.The horizontal field of view of first image acquisition device 1 and third image acquisition device 3 has part weight
Folded, the horizontal field of view of the second image acquisition device 2 and the 4th image acquisition device 4, which has, to partly overlap.Described four pass close to switch
Sensor is equally spaced to be set to the second image acquisition device 2 and equitant second rail of 4 horizontal field of view of the 4th image acquisition device 6
Inside, and the distance between approach switch sensor is 0.785φ, φFor the rolling circular diameter of the wheel.
When the second wheel 11 triggers the first approach switch sensor 10, the first wheel 9 is located at the first image acquisition device 1
In horizontal field of view, third wheel 12 is located in the horizontal field of view of third image acquisition device 3, and the second wheel 11 is located at the second image and adopts
In the horizontal field of view of storage 2, the 4th wheel 13 is located in the horizontal field of view of the 4th image acquisition device 4, four image acquisition devices point
Image Acquisition is not carried out to corresponding wheel not for the first time；When the second wheel 11 triggers the second approach switch sensor 14, first
Wheel 9 is located in the horizontal field of view of the first image acquisition device 1, and third wheel 12 is located at the horizontal field of view of third image acquisition device 3
Interior, the second wheel 11 is located in the horizontal field of view of the second image acquisition device 2, and the 4th wheel 13 is located at the 4th image acquisition device 4
In horizontal field of view, four image acquisition devices carry out Image Acquisition second to corresponding wheel respectively；When the second wheel 11 triggers
When third approach switch sensor 15, the first wheel 9 is located in the horizontal field of view of the first image acquisition device 1, third wheel 12
In in the horizontal field of view of third image acquisition device 3, the second wheel 11 is located in the horizontal field of view of the second image acquisition device 2, and the 4th
Wheel 13 is located in the horizontal field of view of the 4th image acquisition device 4, four image acquisition devices respectively third time to corresponding wheel into
Row Image Acquisition；When the second wheel 11 triggers four approach switch sensors 16, the first wheel 9 is located at the first image acquisition device
In 1 horizontal field of view, third wheel 12 is located in the horizontal field of view of third image acquisition device 3, and the second wheel 11 is located at the second figure
As collector 2 horizontal field of view in, the 4th wheel 13 is located in the horizontal field of view of the 4th image acquisition device 4, four Image Acquisition
Device distinguishes the 4th time and carries out Image Acquisition to corresponding wheel；Computer receives four images of each image acquisition device transmission,
Spliced and obtained after being handled each complete wheel tread information to image, and the tread is carried out judging to obtain is
It is no defective.
Preferably, as shown in Figure 1, between described first image collector 1 and third image acquisition device 3 and the first rail 5
Horizontal distance be 2m, the horizontal distance between second image acquisition device, 2 and the 4th image acquisition device 4 and the second rail 6
For 2m.It should be noted that above-mentioned horizontal distance can carry out adjustment appropriate, the invention is not limited in above-mentioned specific
Numerical value.
Preferably, as shown in Figure 1, horizontal field of view angle θ is 65 °.Further, as shown in Fig. 2, described image collector is logical
Holder 17 is crossed to be fixed on by the rail.Preferably, described image collector is high-speed cmos video camera.It should be noted that
The occurrence of the horizontal field of view angle θ can also be other values, and the concrete form of described image collector can also be other shapes
Formula, such as high speed camera.
When a wheel is by one of approach switch sensor, the front and back wheel of bogie can trigger two in succession
A switching signal, it is assumed that current of traffic to the right when, then previous switching signal is used for controlling the first image acquisition device 1 and the
Two image acquisition devices 2 are shot, and the latter switching signal is used for controlling third image acquisition device 3 and the 4th image acquisition device 4 is clapped
It takes the photograph.Or, it is assumed that current of traffic to the right when, the first wheel pass through first sensor when trigger a switching signal, this is opened
OFF signal is carried out at the same time shooting for controlling first to fourth figure collector.
Tread region top and bottom part distort than more serious it is difficult to carry in the image taken due to image acquisition device
Effective information is taken out, at least needs 3 images on the picture theory of entire tread so if wanting to obtain.But it is examined in actual design
Consider necessary redundancy section in every image, present invention uses the images of 4 different locations to obtain entire tread image.
To keep the ratio that the tread region in every image accounts for entire tread identical and be evenly distributed on entire tread, in the present invention
By 4 approach switch sensor settings in order to which at equal intervals, interval width is the perimeter of 1/4 wheel rolling circle, that is, is spaced.Fig. 3 is setting and the shooting schematic diagram of sensor（Horizontal arrow refers to the traveling side of locomotive in Fig. 3
To）.
In invention centre is only used in order to reduce to the greatest extent in the every pictures of influence of the distortion critical regions to treatment effect
120 ° of tread region.Therefore using 4 approach switch sensors do every pictures that trigger source takes all with front and back figure
There are 15 ° of redundancy sections between piece, in being finally spliced into entirely from 4 images by these redundancy sections for processing
The image of tread.
Embodiment two
As shown in figure 4, the embodiment of the invention discloses a kind of online image detecting methods of train wheel thread defect, including
Following steps：
S101：Wheel is to image zooming-out：Wheel is extracted from the whole image that image acquisition device acquires to region；
It since the image of image acquisition device initial acquisition is larger, needs to cut initial pictures, extracts wheel to figure
Picture, wheel are as follows to the extracting method of image：
With second camera（Second image acquisition device）For acquiring trailing bogie front right tread, second camera exists in this system
The lower shooting of approach switch sensor triggering control obtains part tread image when wheel passes through 4 proximity sensor corresponding positions
Obtain complete tread image, as Figure 6-9.
In order to acquire 4 wheel proximity sensor corresponding position wheel tread images, used in second camera shooting process
Larger field angle, this causes to acquire image to include excessive background information, influences the image processing efficiency of tread portions target image.
It thus needs to cut tread image before carrying out tread image procossing, extraction wheel tread region.Camera acquisition figure
Picture size is 6000*4000, and the present invention uses I2=imcrop (1, [ABCD]) cuts function and cuts second camera acquisition image,
Wherein, 1 processing object diagram 6 is represented,（A, B）Indicate position of the top left corner pixel in original image after cutting；C indicates to scheme after cutting
The width of picture, D indicate the height of image after cutting.It is controlled by proximity sensor since camera image is acquired, Image Acquisition moment camera
It is substantially stationary to relative position with taking turns, thus present position is relatively fixed wheel tread image in the picture.To prevent vibration etc.
Influence of noise, it is [2,300 1,400 1000 that in the case where keeping certain corner redundancy condition, second camera, which corresponds to and cuts each parameter of function,
1100], [2,400 1,300 1,200 1300], [2,900 1,200 1,200 1400], [3,500 1,000 1,200 1600] Fig. 6-9
In each figure cutting effect as shown in figures 10-13.
The tread image known to Figure 10-13 still includes excessive background information, and the present invention uses before extraction tread region
From the graph further segmentation object wheel is to image for the method for ellipses detection, since wheel by wheel proximity sensor and makes it
It triggers in camera acquisition wheel tread image process, camera is fixed with wheel relative position, thus tread region is being schemed in image
Relative position is fixed as in, by taking Figure 13 as an example, directly according to lower right-most portion tread oval feature visibility point feature in image
Delimit the rectangular area for including part elliptical circular arc（W ∈ [500,1000], H ∈ [1000,1600]）Ellipses detection is carried out, favorably
The accuracy detected is improved in evading complex background interference while reducing calculation amount.Shown in concrete operations flow chart 5：
Input pending image；
Pending image is filtered, uses canny operators to carry out edge detection, edge pixel point after grayscale equalization
As characteristic point；
[500,1000] W ∈ in scanning figure, H ∈ [1000,1600] region, statistical nature point establish accumulation structure, add up
Device uses two-dimensional array, two-dimensional array subscript to be characterized dot center's coordinate, and accumulator initial value is set as 0；
Each characteristic point in range is traversed, string midpoint Hough transform is carried out and scans other characteristic points, acquire in them
Point, corresponding accumulator add 1, and only string midpoint Hough transform is carried out to untreated characteristic point when calculating next characteristic point；
When accumulator count value is more than threshold value 5, then it is assumed that there are an ellipse, its corresponding parameters for this position（x0 i, y0 i）
It is exactly oval corresponding centre coordinate, subscript i indicates oval number；
Use each point（x0 i, y0 i）Elliptic equation is moved into origin, is found about origin symmetry under new coordinate system
3 pairs of characteristic points acquire each elliptic equation, compare transverse size, and it is the length of side to select the longest ellipse of axial length, and elliptical center is
The square region of the centre of form is as wheel to region；
Interception wheel carries out subsequent processing as final wheel to positive 120 ° of the region of image to image.Figure 14 is oval
The result figure of detection.
Specifically, in order to remove influence of the distortion critical regions to segmentation tread region, to the wheel that extracts to region also
Need to further reduce the wheel extracted above to region.Only need positive 120 ° of tread region as final in the present invention
Wheel carries out subsequent processing to image, so the square area length of side finally intercepted is 0.866 Ф.Figure 15 is the embodiment of the present invention
Wheel is to intercepting partial schematic diagram in the method.
Further, for the ease of later image segmentation, the method further includes the steps that image preprocessing：It uses
MSRCR algorithms enhance image on wheel, and comparison diagram is as shown in figs. 16-17 before and after the processing.
S102：Image segmentation：Wheel is split image using improved Based On Method of Labeling Watershed Algorithm, extracts tread figure
Picture, as depicted in figs. 18-19；
S103：Image rectification：The transverse and longitudinal coordinate for dividing the image into the tread image extracted carries out geometric correction, and to abnormal
Become serious inoperative side to be cut, obtains tread expanded view, as shown in figure 20；
S104：Image mosaic：According to last acquisition moment tread image size, first three moment is stepped on using cube convolution method
Face image obtains the unified tread image of size into row interpolation, using block matching method to the adjacent acquisition of aforementioned four tread image
Moment carries out sequentially splicing fusion two-by-two, obtains the complete tread image of each wheel；
In the state of camera remains stationary, wheel sequentially passes through 4 position difference proximity sensor triggering camera acquisition figures
Picture, thus 4 moment, to correspond to wheel tread image pixel number different.The present invention is according to last acquisition moment tread image size
The unified tread image of size is obtained into row interpolation to first three moment tread image using cube convolution method, as further illustrated in figures 21-24.
The present invention to the aforementioned four tread image adjacent acquisition moment sequentially two-by-two splice using block matching method melts
It closes.It is as shown in figure 25 to splice flow.The specific method is as follows：
Two adjacent time charts are sequentially loaded into as matrix；
Determine image block scanning window and threshold values；
Start block scan, calculates absolute value difference value between image-region；
Judge whether the difference value is more than threshold value, if it exceeds threshold value then continues to scan on, if being not above threshold value
Storage location；
Curve Matching is carried out, splicing result is exported after being weighted fusion to the curve after matching.
Splicing result is as shown in figure 26.
S105：Image calibration：User's case marker determines paper and is placed in tread respectively to acquire in moment camera photographable position, completes pair
Spliced tread image is demarcated, the correspondence after being spliced between the pixel distance and dimension information of image；
User's case marker determines paper and is placed in tread respectively to acquire in moment camera photographable position, is with the 4th camera acquisition position
Example, segmentation tread image is with plane rectification image as shown in Figure 27-28：Demarcating each grid of paper isJust
It is rectangular, image detection scaling board grid pixel be about in 34*34, that is, image 1 pixel distance represent 1 mm distances, complete
Image calibration.
S106：Tread whether there is breakdown judge：Splice the comentropy and empirical value ratio of rear tread pattern image by calculating
Relatively come judge tread whether there is major defect failure；
The comentropy of 4 wheels in comparison diagram 21-24 can be seen that defective as shown in table 1 from the comparison of information entropy
Wheel information amount be obviously greater than normal wheels.
Whether there is or not the comparisons of the tread image information entropy of thread defect for table 1
| Tread type | Wheel 1 | Wheel 2 | Wheel 3 | Wheel 4 |
| Comentropy | 2.9213 | 3.3264 | 3.4781 | 2.8125 |
S107：Tread failure is extracted：Using improved splitting and merging to judging that the tread of existing defects is handled,
The thread defect image after growth is filled using seed point growth, obtains thread defect image, defect is actually used
Gridding method is estimated that area is calculated using pixel in region in image, and the value of the Gao Yukuan of pixel is profit respectively in image
Difference is sought with the max min of the row and column in the thread defect region in image to be calculated.
It is handled using improved splitting and merging, treated, and defective tread image flow chart is as shown in figure 29, place
It is as shown in figure 30 to manage design sketch.Processing is filled to the thread defect image after growth using seed point growth, to obtain
Thread defect image, as shown in figure 31.By image rectification above illustrate known to each pixel distance of image herein can be with
It is 1mm to be approximately considered, then can be estimated using gridding method defect real area, and area utilizes pixel in region in image
Point calculates.In image the value of the Gao Yukuan of pixel be respectively using the row and column in the thread defect region in image maximum value most
Small value seeks difference to be calculated.The results are shown in Table 2 for defects detection.
2 thread defect parameter of table compares
| Defect area (mm2) | The width (mm) of defect part | The height (mm) of defect part | |
| Actual defects | 1336（Estimation） | 37 | 96 |
| Divide rear tread pattern | 1298 | 42 | 94 |
It is using advantageous effect caused by above-mentioned technical proposal：The system Computer receives each Image Acquisition
Four images of device transmission, then image is spliced by the method and obtains each complete wheel tread after being handled
Information, and the tread is carried out judging to obtain whether defective, thread defect accuracy of detection is high, can effectively avoid detection wrong
Accidentally.
Claims (4)
1. a kind of carrying out the online image detecting method of train wheel thread defect using detecting system, it is characterised in that including as follows
Step：
Wheel is to image zooming-out：Wheel is extracted from the whole image that image acquisition device acquires to region；
Image segmentation：Wheel is split image using improved Based On Method of Labeling Watershed Algorithm, extracts tread image；
Image rectification：The transverse and longitudinal coordinate for dividing the image into the tread image extracted carries out geometric correction, and serious to distorting
Inoperative side is cut, and tread expanded view is obtained；
Image mosaic：According to last acquisition moment tread expanded view size, first three moment tread is unfolded using cube convolution method
Figure obtains the unified tread expanded view of size into row interpolation, using block matching method to the adjacent acquisition of aforementioned four tread expanded view
Moment carries out sequentially splicing fusion two-by-two, obtains the complete tread image of each wheel；
Image calibration：User's case marker determines paper and is placed in tread respectively to acquire in moment camera photographable position, completes to spliced
Tread image is demarcated, the correspondence after being spliced between the pixel distance and dimension information of image；
Tread whether there is breakdown judge：Splice comentropy and the empirical value of rear tread pattern image relatively by calculating to judge to step on
Face whether there is major defect failure；
Tread failure is extracted：Using improved splitting and merging to judging that the tread of existing defects is handled, seed is utilized
Point growth is filled the thread defect image after growth, the thread defect image after being filled, and grid is used to defect
Method is estimated that area is calculated using pixel in region in the thread defect image after filling, the thread defect figure after filling
The value of the Gao Yukuan of pixel is the row and column using the thread defect region in the thread defect image after filling respectively as in
Max min seeks difference to be calculated；
Wherein, the improved splitting and merging includes the following steps：
Setting judges sentence P and minimum unit size min, builds discriminant function；
Judge whether each region of image meets discriminant function；
It is considered as if meeting discriminant function and meets splitting condition, the sub-district area image quartering of splitting condition will be met, and
Image is judged again；If being unsatisfactory for discriminant function, merging is set and judges sentence Q, and judges sentence Q by merging
Judge whether there is subregion to meet merging condition in image, if it is merges the subregion for meeting merging condition, such as
Fruit is not to terminate.
2. the online image detecting method of train wheel thread defect as described in claim 1, it is characterised in that the method exists
Further include before image segmentation：The pre-treatment step that wheel optimizes the color information of image using MSRCR algorithms.
3. the online image detecting method of train wheel thread defect as described in claim 1, it is characterised in that the wheel is to figure
As the method for extraction is as follows：
Input pending image；The pending image is the whole image acquired from image acquisition device；
Pending image is filtered, uses canny operators to carry out edge detection after grayscale equalization, edge pixel point is
Characteristic point；
[500,1000] W ∈ in scanning figure, H ∈ [1000,1600] region, statistical nature point establish accumulation structure, and accumulator is adopted
With two-dimensional array, two-dimensional array subscript is characterized dot center's coordinate, and accumulator initial value is set as 0；
Each characteristic point in range is traversed, string midpoint Hough transform is carried out and scans other characteristic points, acquire their midpoints,
Corresponding accumulator adds 1, and only string midpoint Hough transform is carried out to untreated characteristic point when calculating next characteristic point；
When accumulator count value is more than 5, then it is assumed that there are an ellipse, its corresponding parameters for this position（x0 i, y0 i）It is exactly oval
Corresponding centre coordinate, subscript i indicate oval number；
Use each point（x0 i, y0 i）Elliptic equation is moved into origin, 3 couples of spies about origin symmetry are found under new coordinate system
Sign point acquires each elliptic equation, compares transverse size, and it is the length of side to select the longest ellipse of axial length, and elliptical center is the centre of form
Square region is as wheel to region；
Interception wheel carries out subsequent processing as final wheel to positive 120 ° of the region of image to image.
4. the online image detecting method of train wheel thread defect as described in claim 1, it is characterised in that：The Block- matching
Method is as follows：
Two adjacent time charts are sequentially loaded into as matrix；
Determine image block scanning window and discrepancy threshold；
Start block scan, calculates absolute value difference value between image-region；
Judge whether the difference value is more than discrepancy threshold, if it exceeds discrepancy threshold then continues to scan on, if being not above difference
Different threshold value then storage location；
Curve Matching is carried out, splicing result is exported after being weighted fusion to the curve after matching.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710305046.8A CN107146224B (en) | 2017-05-03 | 2017-05-03 | The online image detecting system of train wheel thread defect and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710305046.8A CN107146224B (en) | 2017-05-03 | 2017-05-03 | The online image detecting system of train wheel thread defect and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107146224A CN107146224A (en) | 2017-09-08 |
| CN107146224B true CN107146224B (en) | 2018-07-17 |
Family
ID=59775483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710305046.8A Active CN107146224B (en) | 2017-05-03 | 2017-05-03 | The online image detecting system of train wheel thread defect and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107146224B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921845A (en) * | 2018-07-10 | 2018-11-30 | 深圳大学 | Wheel tread rehabilitation plan generation method and device |
| CN111080632B (en) * | 2019-12-20 | 2023-05-26 | 石家庄铁道大学 | Wheel set tread polygon monitoring method and device and terminal equipment |
| CN113160176B (en) * | 2021-04-23 | 2022-10-18 | 歌尔股份有限公司 | Defect detection method and device |
| CN115439481B (en) * | 2022-11-09 | 2023-02-21 | 青岛平电锅炉辅机有限公司 | Deaerator welding quality detection method based on image processing |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105203552A (en) * | 2015-09-18 | 2015-12-30 | 南京拓控信息科技有限公司 | 360-degree tread image detecting system and method |
| CN106404791B (en) * | 2016-08-24 | 2019-05-24 | 北京主导时代科技有限公司 | A kind of train, online wheel tread image detection device and fault detection system |
-
2017
- 2017-05-03 CN CN201710305046.8A patent/CN107146224B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107146224A (en) | 2017-09-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107146224B (en) | The online image detecting system of train wheel thread defect and method | |
| CN105913396B (en) | A kind of image border holding mixing denoising method of noise estimation | |
| CN104835164B (en) | A kind of processing method and processing device of binocular camera depth image | |
| CN108230344B (en) | Automatic identification method for tunnel water leakage diseases | |
| CN102663760B (en) | Location and segmentation method for windshield area of vehicle in images | |
| CN109636732A (en) | A kind of empty restorative procedure and image processing apparatus of depth image | |
| CN105158258A (en) | Computer vision-based bamboo strip surface defect detection method | |
| CN107169979A (en) | A kind of method for detecting image edge of improvement Canny operators | |
| CN109325498B (en) | Vein extraction method for improving Canny operator based on window dynamic threshold | |
| CN109886935A (en) | A kind of road face foreign matter detecting method based on deep learning | |
| CN108710832B (en) | Reference-free iris image definition detection method | |
| CN111028169B (en) | Image correction method, device, terminal equipment and storage medium | |
| CN113506246B (en) | Concrete 3D printing component fine detection method based on machine vision | |
| CN110490886A (en) | A kind of method for automatically correcting and system for certificate image under oblique viewing angle | |
| CN106022268A (en) | Identification method and device of speed limiting sign | |
| CN106651837A (en) | White glass plate surface edge breakage defect detecting method | |
| CN109598736A (en) | The method for registering and device of depth image and color image | |
| CN114612410A (en) | Novel clothing detects device | |
| CN108256385A (en) | The front vehicles detection method of view-based access control model | |
| CN114212452B (en) | Coal flow detection method based on laser assistance and image processing and energy-saving control system | |
| CN106780529B (en) | Conference video mosaic detection method based on external rectangle | |
| CN105631868A (en) | Depth information extraction method based on image classification | |
| CN109493293A (en) | A kind of image processing method and device, display equipment | |
| CN107633505A (en) | A kind of undercarriage detection method based on target gray distribution character | |
| KR101696040B1 (en) | Apparatus and method for scalping defect detection |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210412 Address after: 215200 558 FENHU Road, Wujiang District, Suzhou, Jiangsu Patentee after: Suzhou Huayun Shichuang Intelligent Technology Co.,Ltd. Address before: 050043 No. 17, North Second Ring Road, Hebei, Shijiazhuang Patentee before: SHIJIAZHUANG TIEDAO University |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240109 Address after: Room 617, Building 1B, Launching Area of Yangtze River Delta International R&D Community, No. 286, Qinglonggang Road, High speed Railway New Town, Xiangcheng District, Suzhou City, Jiangsu Province, 215000 Patentee after: Suzhou Huichuang Yunlian Intelligent Technology Co.,Ltd. Address before: 215200 558 FENHU Road, Wujiang District, Suzhou, Jiangsu Patentee before: Suzhou Huayun Shichuang Intelligent Technology Co.,Ltd. |
|
| TR01 | Transfer of patent right |