CN112085723A - Automatic detection method for spring jumping fault of truck bolster - Google Patents
Automatic detection method for spring jumping fault of truck bolster Download PDFInfo
- Publication number
- CN112085723A CN112085723A CN202010940916.0A CN202010940916A CN112085723A CN 112085723 A CN112085723 A CN 112085723A CN 202010940916 A CN202010940916 A CN 202010940916A CN 112085723 A CN112085723 A CN 112085723A
- Authority
- CN
- China
- Prior art keywords
- image
- spring
- fault
- bolster
- swing bolster
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 41
- 230000009191 jumping Effects 0.000 title claims abstract description 10
- 238000000034 method Methods 0.000 claims abstract description 44
- 230000008569 process Effects 0.000 claims abstract description 26
- 230000000877 morphologic effect Effects 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 238000001914 filtration Methods 0.000 claims description 15
- 238000004891 communication Methods 0.000 claims description 11
- 238000013459 approach Methods 0.000 claims description 10
- 230000011218 segmentation Effects 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 230000003321 amplification Effects 0.000 claims description 2
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 2
- 230000005465 channeling Effects 0.000 abstract description 6
- 229910000831 Steel Inorganic materials 0.000 description 7
- 239000010959 steel Substances 0.000 description 7
- 230000004927 fusion Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000003139 buffering effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30268—Vehicle interior
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
An automatic detection method for a truck bolster spring jumping fault belongs to the technical field of rail wagon component detection. The problem that manual detection efficiency is low and the problem that image automatic detection accuracy is low and unstable exist in the process of detecting the jumping fault of the truck bolster spring is solved. According to the method, the contour of a large image of the spring of the swing bolster is obtained by Canny detection, Hough transform detection is carried out to obtain a straight line to obtain the upper edge of the swing bolster, OTSU binaryzation is carried out on the large image of the spring of the swing bolster, two hollow areas are found through morphological calculation to position the lower edge of the spring of the swing bolster, so that a top sub-image, a tail end sub-image and an integral image of the spring are determined, and then channeling identification is carried out on the top sub-image, the tail end sub-image. The invention is suitable for automatic detection of the spring jumping fault of the truck bolster.
Description
Technical Field
The invention relates to an automatic detection method for a spring jumping fault of a truck bolster. Belongs to the technical field of railway wagon component detection.
Background
The truck bolster spring is used for a running part of a railway vehicle, plays a role in buffering and shock absorption, and is located on the running part, if the truck bolster spring is fleed out, the running safety is directly endangered, so that the detection of the fleeing fault of the truck bolster spring is very important in the detection process of parts of the railway vehicle.
The existing method for detecting the fault of the spring jumping of the swing bolster basically adopts a mode of manually checking images to detect the fault. The conditions of fatigue, omission and the like are easily caused by vehicle inspection personnel in the working process, so that the appearance of missed inspection and wrong inspection is caused, and the driving safety is influenced. With continuous maturity and development of image processing and technology, automatic detection of a truck bolster spring fleeing fault by adopting an automatic image identification mode is provided, but at present, the effect based on image processing is influenced by various factors including weather and the like, the detection accuracy is low, and the problems of missed detection rate and high false detection rate exist. And the detection effect of the image under different conditions is unstable.
Disclosure of Invention
The invention aims to solve the problems of low manual detection efficiency and low and unstable automatic image detection accuracy in the process of detecting the spring jumping fault of the truck bolster.
A method for automatically detecting a spring jumping fault of a truck bolster comprises the following steps:
s1, acquiring a large swing bolster spring image, wherein the large swing bolster spring image is a local truck image including a swing bolster spring;
s2, based on the large image of the swing bolster spring, obtaining the outline of the large image of the swing bolster spring by using Canny detection; carrying out Hough transform detection on the contour image to obtain the upper edge of the swing bolster, wherein the upper edge is used for positioning the upper edge of a swing bolster spring; after OTSU binaryzation is carried out on the large image of the swing bolster spring, two hollow areas are found through morphological calculation, and the lower edge of the swing bolster spring is located;
a subgraph between the upper boundary and the lower boundary is intercepted from a large image of the swing bolster spring to obtain an integral image of the spring; intercepting a top end sub-image and a tail end sub-image by combining the prior information with the upper edge information and the lower edge information;
s3, performing the play identification of the top sub-image, and if the play fault is identified, directly alarming by the image; if the top end is out of position and no fault is identified, the image at the tail end part is identified, and if the out of position fault is identified, the image directly alarms; if the top end and the tail end do not recognize the play fault, the left play fault and the right play fault are recognized, and if the fault is recognized, the image directly alarms.
Further, the process of performing top sub-image fleeing recognition at s3 includes the following steps:
extracting a template image corresponding to the top sub-image, and eliminating noise of the template image and the top sub-image through median filtering; and matching the top sub-image with the template image, wherein the maximum similarity of the normalized correlation coefficient matching of the top sub-image is less than a correlation coefficient threshold T1, and the maximum similarity of the normalized correlation matching is less than a correlation threshold T2 for direct alarm.
Further, the process of recognizing a tip section slip at s3 is the same as the process of recognizing a tip section slip.
Further, the process of performing left-right fleeing fault identification s3 is as follows:
performing median filtering on the whole image of the spring, then calculating an x-direction gradient Gx and a y-direction gradient Gy by using a sobel operator, and obtaining an edge profile I1(I, j) of the image;
meanwhile, the whole image of the spring is binarized, and a 2-valued image I2(I, j) of the difference between the reduction class and the amplification class is obtained through a global threshold value binarization filtering image;
segmenting by utilizing a curve evolution image to obtain a spring segmented contour, wherein an initial curve C is an I2(I, j) binary pixel intersection part in a 2-valued image, and then iterating by utilizing an evolution equation numerical solution to enable the initial curve to approach the spring contour and gradually approach the spring contour to a target edge, and finally finding the target edge;
performing linear fitting on the left and right side boundaries of each row of springs by combining the priori knowledge on the spring gap contour segmented by the level set; calculating the inclination angle of the boundary of the spring according to the fitted straight line, and judging whether the adjacent springs are crossed; and if the inclination angle of the boundary of the spring is obvious or two adjacent columns of springs are not parallel, directly alarming.
Further, the process of obtaining the 2-valued image I2(I, j) with the difference between the reduced class and the enlarged class by performing the global threshold binarization filtering includes the following steps:
globally binarizing the whole image of the spring, wherein the gray value is 255 when the gray value is larger than a gray threshold th1 and 0 when the gray value is smaller than or equal to th1, and obtaining a mask image;
assigning all pixel values of the filtered image to be 0, the size of the pixel values is the same as that of the whole image of the spring; if the (i, j) position is not 0 in the mask image, the following operation is performed: taking the position as a center, calculating a brightness mean value in a rectangular region with the length of W and the width of H in the whole image of the spring as a pixel value of a corresponding position in a filtering image, and only calculating a position other than 0 in a mask image in the process; the 2-valued image I2(I, j) of the difference between the reduced class and the enlarged class is obtained by binarizing the filtered image with the global threshold value th0, and the pixel value of the 2-valued image I2(I, j) is 0 or 255.
Further, the iteration of the numerical solution of the evolution equation is used for enabling the initial curve to approach the spring contour and gradually approach the spring contour to the target edge, and the process of finally finding the target edge comprises the following steps:
1) if the pixel value of the 2-valued image I2(I, j) of which the position corresponding to the initialized level set function phi (x, y) differs between the reduced class and the enlarged class is 255, the initial level set value is d, if the pixel value of the corresponding position in I2(I, j) is 0, the level set value is-d, and the black-white joint in the 2-valued image is an initial contour line;
2) at time t equal to 0, an initial symbol distance function phi is given, according to the equation of motion
Obtaining phi at any time t; according to phitThe evolution equation of + F | (Φ |) > 0 evolves a curve through a numerical solution; wherein the content of the first and second substances,
recording ^ phi, speed XtIs given by the normal vector of the surface and Xt＝F(X(t))n，
F is a velocity function, which is a function of the filtered image pixel values;
and (3) stopping iterating for a certain number of times N or stopping if the difference value of the level sets of the previous and subsequent times is smaller than an error threshold th, wherein the obtained curve is the final spring gap segmentation curve, and otherwise, continuing to iterate and evolve.
Further, the speed function F is specifically as follows:
where λ is a parameter that controls the penalty factor.
Further, the symbol distance function is as follows:
where d [ (x, y), C ] is the Euclidean distance of point (x, y) to curve C.
Further, in the step s2, after OTSU binarization is performed on the large image of the bolster spring, a specific process of finding out the two hollow areas through morphological operation to locate the lower edge of the bolster spring includes the following steps:
a. aiming at a large image of a swing bolster spring, obtaining a binary image by using an Otsu method, and then negating the binary image;
b. setting a structural unit with the width of w and the height of h, and carrying out morphological open operation to remove a spring gap communication area;
c. and positioning the hollow area by combining the position and the size information of the hollow area, and further obtaining the lower edge of the swing bolster spring.
Further, the method further comprises the steps of:
after the fault is identified, calculating the position of the fault in the original image according to the mapping relation from the subgraph to the large image of the bolster spring and the mapping relation between the large image of the bolster spring and the original image, and displaying the fault on a display interface;
the subgraphs are the whole image, the top end subimage or the end subimage of the spring.
Has the advantages that:
1. and the automatic image identification mode is used for replacing manual detection, so that the detection efficiency and accuracy are improved.
2. In the fault identification, the channeling fault is divided into three conditions of top channeling, bottom channeling and left-right inclined channeling, the end fault detection is firstly carried out on the channeling fault, and if the end fault is not detected, the left-right inclined detection is carried out, so that the detection speed and the accuracy can be improved.
3. The method has the advantages that the normal and fault-free template image of the spring is obtained by carrying out image fusion on different detection stations and different vehicle types, and the system universality is high.
4. The edge curve obtained by the curve evolution spring segmentation method has the advantages of being closed, smooth and the like, the interference of the edge fitting straight line of each subsequent row of springs is less, and the false alarm in fault detection is reduced.
5. The 2-valued image I2(I, j) of the difference between the reduction class and the enlargement class is used as an initial value of curve evolution, and the convergence rate of the curve evolution is high.
Drawings
FIG. 1 is a schematic view of an overall process for fault identification;
FIG. 2 is a schematic view of an end-play identification process;
FIG. 3 is a schematic diagram of a left-right oblique fleeing identification process.
Detailed Description
The first embodiment is as follows: the present embodiment is described in detail with reference to figure 1,
the automatic detection method for the spring fleeing fault of the truck bolster in the embodiment comprises the following steps of:
the hardware part comprises a camera acquisition unit, a magnetic steel unit, an image acquisition industrial personal computer unit, a control industrial personal computer unit and an image recognition unit, wherein the camera acquisition unit comprises a camera and a compensation optical module, the camera is used for acquiring images, and the compensation optical module is used for optical compensation during image acquisition; the magnetic steel unit comprises near-end magnetic steel and far-end magnetic steel, signals of the near-end magnetic steel and the far-end magnetic steel are transmitted to the control industrial personal computer unit, the control industrial personal computer unit calculates vehicle speed and wheel base information through the acquired signal information and transmits the vehicle speed and wheel base information to the image recognition unit, and the image recognition unit realizes an automatic recognition algorithm by utilizing the acquired wheel base information, image information and the like.
s1, acquiring a large image of the bolster spring:
the method comprises the steps of collecting an original image of a truck, roughly estimating the position of a swing bolster spring component by combining priori knowledge according to magnetic steel wheel base information obtained by a hardware part, and obtaining a large swing bolster spring image, wherein the large swing bolster spring image is a local image of the truck including a swing bolster spring.
s2, acquiring a small image of the bolster spring:
based on the large image of the swing bolster spring, utilizing Canny to detect and obtain the outline of the large image of the swing bolster spring; carrying out Hough transform detection on the contour image to obtain the upper edge of the swing bolster, wherein the upper edge is used for positioning the upper edge of a swing bolster spring; in view of the characteristic that the lower boundary of the swing bolster spring has two hollowed-out parts, the invention finds two hollowed-out areas through morphological calculation to position the lower edge of the swing bolster spring after OTSU binaryzation is carried out on a large image of the swing bolster spring.
The concrete positioning steps of the lower edge of the swing bolster spring are as follows:
a. and (3) binary image with the hollow area as the foreground:
aiming at a large image of the swing bolster spring, a binarization image is obtained by using the Otsu method, and the position of a hollow area is a black pixel. And then, inverting the binary image to obtain a pixel with a white hollow area value.
b. Eliminating spring clearance interference
And (b) the hollow area in the step (a) is white, and the slender contour line of the adjacent spring gap is also white. And a structural unit with the width of w and the height of h is arranged, and the spring gap communication area can be removed after morphological opening operation is carried out on the structural unit.
c. Positioning bolster spring lower edge according to hollowed-out region information
After the above steps, some images still have interference areas. The hollow area can be accurately positioned by combining the position and the size information of the hollow area. The center of the communication area is located at the lower left corner, the area of the communication area is larger than s, the communication area is a left hollow area, and the center of the communication area is located at the lower right corner, the area of the communication area is larger than s, the communication area is a right hollow area. And averaging the end rows of the two hollowed-out areas to obtain the lower edge of the swing bolster spring.
A subgraph between the upper boundary and the lower boundary is intercepted from a large image of the swing bolster spring to obtain an integral image of the spring; the image of the top end part and the image of the tail end part can be intercepted according to the proportion by combining the prior information with the upper edge information and the lower edge information; the entire spring image, the tip portion sub-image, and the tip portion sub-image are collectively referred to as sub-images.
s3, fault identification:
the spring play is divided into three conditions of top play, tail play and left and right inclined play. Because the spring itself receives influence such as goods light and heavy, camera angle, hardware installation environment, even like the trouble, the difference of different concrete trouble is also great, in order to improve the rate of accuracy of system, adopts the strategy of distinguishing respectively to three kinds of condition: firstly, carrying out play identification on top sub-images, and if a play fault is identified, directly alarming by the images; if the top end is out of position and no fault is identified, the image at the tail end part is identified, and if the out of position fault is identified, the image directly alarms; if the top end and the tail end do not recognize the play fault, the left play fault and the right play fault are recognized, and if the fault is recognized, the image directly alarms.
s3.1 end fleeing identification
The end part of the swing bolster spring is a head ring or a tail ring spring riding platform, and the end part of the swing bolster spring comprises a top end and a tail end; the play-out state is a change in the relative position with respect to the adjacent component. Because the springs are extruded and deformed mutually or dislocated with adjacent parts or leaked out of the bottom surfaces of the springs after the springs are shifted, edges in an image are increased, brightness smoothness of an image pixel value is reduced after the edges are increased, and shaking is increased, so that textures are richer than those of a normal spring, edges of the end parts of the springs in the image are increased than those of a normal fault-free image after shifting faults occur, and the textures are richer. Based on image feature transformation before and after the fault, the end part fleeing fault is identified by comparing the current image with the normal fault-free template image.
s3.1.1 building a template database
And performing multi-scale image fusion on images of different detection stations and images of different vehicle types, wherein the fused images have more and more valuable information. The process of multi-scale image fusion is as follows:
firstly, manually selecting an image as a template, then extracting characteristic point pairs for matching and calculating a transformation matrix according to the multi-scale of the template for other station images, and finally, weighting the corrected different station images to obtain a final template image.
s3.1.2 fault identification, as shown in figure 2,
extracting a template image corresponding to the top sub-image, and eliminating noise of the template image and the top sub-image through median filtering; and matching the top sub-image with the template image, wherein the maximum similarity of the normalized correlation coefficient matching of the top sub-image is less than a correlation coefficient threshold T1, and the maximum similarity of the normalized correlation matching is less than a correlation threshold T2 for direct alarm.
And when the top end is out of position and the fault is not identified, the tail end sub-image is identified, if the out of position fault is identified, the image directly alarms, and the processing process of the tail end sub-image is the same as that of the top end sub-image.
s3.2 left and right Tilt spring identification
For spring play failures of the left-right tilting type, a certain row of springs has obvious inclination on an image or two adjacent rows of springs are changed into cross from the original parallel relation. The spring overall image obtains the contour of the spring segmentation by evolving the contour of the level set of the surface S, and whether the swing bolster spring has a left-right inclination fault or not can be judged according to the contour of the spring segmentation. The curve evolution image segmentation firstly defines an initial curve C as the intersection part of black and white pixels of I2(I, j) in a 2-valued image, and then iterates by using a numerical solution of an evolution equation to enable the initial curve to approach to the spring contour, so that the initial curve gradually approaches to the target edge, and finally the target edge is found. The edge curve obtained by the dynamic approximation method has the advantages of being closed, smooth and the like, and the accurate spring gap profile can be obtained when evolution stops. Research and experiments show that the method has high accuracy in positioning the spring profile.
s3.2.1 data preparation before spring splitting
Although gaps among all layers of springs of the swing bolster spring are different, the brightness of most of the gaps of the spring is darker than that of the gaps of the adjacent springs, and parts of the gaps of the spring have shadows. The gradient image serving as a parameter in the evolution of the segmentation curve can accelerate the curve evolution speed.
The gradient image acquisition mode is as follows: firstly, performing median filtering on the whole spring image to eliminate noise; then, the sobel operator is used for calculating an x-direction gradient Gx and a y-direction gradient Gy, and after the x-direction gradient and the y-direction gradient are weighted, the edge contour I1(I, j) of the image is obtained if | Gx | 0.5+ | Gy | 0.5.
The 2-valued image I2(I, j) of the difference between the reduced class and the enlarged class provides an initial value for the curve evolution of the spring segmentation, which is obtained as follows:
firstly, the whole image of the spring is subjected to global binarization, the gray value is 255 when the gray value is larger than a gray threshold th1, and the gray value is 0 when the gray value is smaller than or equal to th1, and then a mask image can be obtained. The grayscale threshold is an empirical value that can divide a component region such as a spring and a shadow region such as a gap in an image obtained by hardware, and generally may be a value at a valley of an image grayscale histogram.
And next, acquiring a filtering image, wherein all pixel values of the filtering image are assigned to be 0, and the size of the filtering image is the same as that of the whole image of the spring. If the (i, j) position is not 0 in the mask image, the following operation is performed: taking the position as the center, calculating the brightness mean value in the rectangular area with the length W and the width H (only considering the position of non-0 pixel in the mask image) in the whole image of the spring as the pixel value of the corresponding position in the filter image. Considering that the spring gap is relatively dark, the filtered image is binarized by a global threshold value with a threshold value th0 to obtain a 2-valued image I2(I, j) of the difference between the reduced class and the enlarged class, th0 can be determined according to prior, for example, a numerical value separating the spring and other components from the background such as the gap, namely, a valley of the image gray level histogram; the pixel value of the 2-valued image I2(I, j) is 0 or 255.
Basic theory of evolution of s3.2.2 curve
The spring segmentation result is obtained by tracking contour points, and the implicit contour method is adopted in the invention in view of the defects that in the explicit contour tracking method, points need to be deleted when curves are combined and points need to be inserted when the curves are combined. Assuming that the curve on the two-dimensional plane is represented by y ═ f (x), the relationship can also be described by an implicit function y-f (x) of 0, where:
φ(x，y)＝y-f(x)(1)
then phi (x, y) being 0 is an implicit expression of the curve.
For the construction of phi (x, y), a symbol distance function is used, i.e.
Where d [ (x, y), C ] is the Euclidean distance of point (x, y) to curve C.
The point X (X, y) belongs to a curve that evolves with time, X (t) being its position at time t. At any time t, x (t) is for each point x (t) a point on the curve with height 0 of surface phi, i.e.:
φ(X(t),t)＝0 (3)
thus, a curved surface φ is obtained. X (X, y) is the contour line used to segment the spring. In the evolution process of the curved surface, the 0 level set, i.e., the contour lines for dividing the spring, is continuously evolving. What remains is how to evolve this curved surface. At time t equal to 0, an initial function phi is given, according to the equation of motion
Phi at any time t can be obtainedt. The velocity equation (3) of the curved surface evolution utilizes a chain rule, which has
φt+F|▽φ|＝0 (4)
Wherein the content of the first and second substances,
recording ^ phi, speed XtIs given by the normal vector of the surface and Xt＝F(X(t))n，
F is a speed function, which is a function of the pixel values of the filtered image in the previous step, the higher the frequency, the slower the evolution speed of the local curve, and the specific values of F are as follows:
where λ is a parameter that controls the penalty factor.
s3.2.3 Curve evolution procedure:
1) if the pixel value of the 2-valued image I2(I, j) of which the position corresponding to the initialized level set function phi (x, y) differs between the reduced class and the enlarged class is 255, the initial level set value is d, if the pixel value of the corresponding position in I2(I, j) is 0, the level set value is-d, and the black-white joint in the 2-valued image is an initial contour line;
2) evolving a curve by a numerical solution according to the evolution equation of the formula (4);
and (3) stopping iterating for a certain number of times N or stopping if the difference value of the level sets of the previous and subsequent times is smaller than an error threshold th, wherein the obtained curve is the final spring gap segmentation curve, and otherwise, continuing to iterate and evolve.
s3.2.4 combined with curve clearance to identify left-right inclined fleeing fault
And (3) performing linear fitting on the left and right side boundaries of each row of springs by combining the priori knowledge on the spring gap contour segmented by the level set. And calculating the inclination angle of the boundary of the spring according to the fitted straight line, and judging whether the adjacent springs are crossed. And if the inclination angle of the boundary of the spring is obvious or two adjacent columns of springs are not parallel, directly alarming.
In the following, two rows of springs are taken as an example, and how to identify the tilt spring fault according to the segmentation result is specifically described. In the spring gap split image, 4 sets of dot columns were sought. Searching a contour with the length being more than half of the length of the spring and the connected region being positioned on the left side of the whole image in the connected region of the segmented image, wherein if the x coordinate of the leftmost point of the contour is less than one third of the width of the image, the x coordinate of the leftmost point is a first group of point rows; searching a contour with the length being more than half of the length of the spring and the connected region being positioned on the right side of the whole image in the connected region of the segmented image, wherein if the x coordinate of the rightmost point of the contour is more than two-thirds of the width of the image, the x coordinate is a second group of point rows; searching a contour with the length being more than half of the length of the spring and the communication area being positioned on the left side of the whole image in the communication area of the segmented image, wherein if the x coordinate of the rightmost point of the contour is more than or equal to one third of the image width and less than or equal to two thirds of the image width, the x coordinate is a third group of point rows; and searching the outline with the length larger than half of the length of the spring and the connected area positioned on the right side of the whole image in the connected area of the divided image, wherein the x coordinate of the leftmost point of the outline is a fourth group of point rows if the x coordinate is larger than or equal to one third of the width of the image and smaller than or equal to two thirds of the width of the image. If the inclination angle of the first group or the fourth group of point column fitting straight line is larger than K, directly giving an alarm; and the included angle between the fitting straight lines of the second group and the third group is larger than K, and the alarm is directly given.
s4 failure handling
And after the fault is identified, calculating the position of the fault in the original image according to the mapping relation from the subgraph to the large image of the bolster spring and the mapping relation between the large image of the bolster spring and the original image, and displaying the fault on a display interface.
The present invention is capable of other embodiments and its several details are capable of modifications in various obvious respects, all without departing from the spirit and scope of the present invention.
Claims (10)
1. A method for automatically detecting a spring jumping fault of a truck bolster is characterized by comprising the following steps:
s1, acquiring a large swing bolster spring image, wherein the large swing bolster spring image is a local truck image including a swing bolster spring;
s2, based on the large image of the swing bolster spring, obtaining the outline of the large image of the swing bolster spring by using Canny detection; carrying out Hough transform detection on the contour image to obtain the upper edge of the swing bolster, wherein the upper edge is used for positioning the upper edge of a swing bolster spring; after OTSU binaryzation is carried out on the large image of the swing bolster spring, two hollow areas are found through morphological calculation, and the lower edge of the swing bolster spring is located;
a subgraph between the upper boundary and the lower boundary is intercepted from a large image of the swing bolster spring to obtain an integral image of the spring; intercepting a top end sub-image and a tail end sub-image by combining the prior information with the upper edge information and the lower edge information;
s3, performing the play identification of the top sub-image, and if the play fault is identified, directly alarming by the image; if the top end is out of position and no fault is identified, the image at the tail end part is identified, and if the out of position fault is identified, the image directly alarms; if the top end and the tail end do not recognize the play fault, the left play fault and the right play fault are recognized, and if the fault is recognized, the image directly alarms.
2. The automatic detection method for the spring fleeing fault of the truck bolster spring according to claim 1, wherein the process of performing the top sub-image fleeing recognition at s3 comprises the following steps:
extracting a template image corresponding to the top sub-image, and eliminating noise of the template image and the top sub-image through median filtering; and matching the top sub-image with the template image, wherein the maximum similarity of the normalized correlation coefficient matching of the top sub-image is less than a correlation coefficient threshold T1, and the maximum similarity of the normalized correlation matching is less than a correlation threshold T2 for direct alarm.
3. The automatic detection method for the spring fleeing fault of the truck bolster spring according to claim 2, characterized in that the process of identifying the terminal part image by s3 is the same as the process of identifying the top part image by fleeing.
4. The automatic detection method for the spring fleeing fault of the truck bolster spring according to claim 3, characterized in that the process of identifying the left and right fleeing faults in s3 is as follows:
performing median filtering on the whole image of the spring, then calculating an x-direction gradient Gx and a y-direction gradient Gy by using a sobel operator, and obtaining an edge profile I1(I, j) of the image;
meanwhile, the whole image of the spring is binarized, and a 2-valued image I2(I, j) of the difference between the reduction class and the amplification class is obtained through a global threshold value binarization filtering image;
segmenting by utilizing a curve evolution image to obtain a spring segmented contour, wherein an initial curve C is an I2(I, j) binary pixel intersection part in a 2-valued image, and then iterating by utilizing an evolution equation numerical solution to enable the initial curve to approach the spring contour and gradually approach the spring contour to a target edge, and finally finding the target edge;
performing linear fitting on the left and right side boundaries of each row of springs by combining the priori knowledge on the spring gap contour segmented by the level set; calculating the inclination angle of the boundary of the spring according to the fitted straight line, and judging whether the adjacent springs are crossed; and if the inclination angle of the boundary of the spring is obvious or two adjacent columns of springs are not parallel, directly alarming.
5. The automatic detection method for the spring fleeing fault of the truck bolster spring according to claim 4, wherein the process of obtaining the 2-valued image I2(I, j) of the difference between the reduced class and the enlarged class by global threshold value binary filtering comprises the following steps:
globally binarizing the whole image of the spring, wherein the gray value is 255 when the gray value is larger than a gray threshold th1 and 0 when the gray value is smaller than or equal to th1, and obtaining a mask image;
assigning all pixel values of the filtered image to be 0, the size of the pixel values is the same as that of the whole image of the spring; if the (i, j) position is not 0 in the mask image, the following operation is performed: taking the position as a center, calculating a brightness mean value in a rectangular region with the length of W and the width of H in the whole image of the spring as a pixel value of a corresponding position in a filtering image, and only calculating a position other than 0 in a mask image in the process; the 2-valued image I2(I, j) of the difference between the reduced class and the enlarged class is obtained by binarizing the filtered image with the global threshold value th0, and the pixel value of the 2-valued image I2(I, j) is 0 or 255.
6. The automatic detection method for the spring leap fault of the truck bolster as claimed in claim 5, wherein the process of using the numerical solution of the evolution equation to iterate to make the initial curve approach to the spring contour and gradually approach to the target edge, and finally finding the target edge comprises the following steps:
1) if the pixel value of the 2-valued image I2(I, j) of which the position corresponding to the initialized level set function phi (x, y) differs between the reduced class and the enlarged class is 255, the initial level set value is d, if the pixel value of the corresponding position in I2(I, j) is 0, the level set value is-d, and the black-white joint in the 2-valued image is an initial contour line;
2) at time t equal to 0, an initial symbol distance function phi is given, according to the equation of motion
Obtaining phi at any time t; according to
Evolves the curve by numerical solution; wherein the content of the first and second substances,
record as
Speed XtIs given by the normal vector of the surface and Xt＝F(X(t))n，
F is a velocity function, which is a function of the filtered image pixel values;
and (3) stopping iterating for a certain number of times N or stopping if the difference value of the level sets of the previous and subsequent times is smaller than an error threshold th, wherein the obtained curve is the final spring gap segmentation curve, and otherwise, continuing to iterate and evolve.
7. The method for automatically detecting the fault of the spring leap of the truck bolster as claimed in claim 6, wherein the speed function F is specifically as follows:
where λ is a parameter that controls the penalty factor.
8. The automatic detection method for the spring-out fault of the truck bolster as claimed in claim 7, wherein the symbolic distance function is as follows:
where d [ (x, y), C ] is the Euclidean distance of point (x, y) to curve C.
9. The method for automatically detecting the running fault of the truck bolster spring according to one of claims 1 to 8, wherein the specific process of finding two hollowed-out areas to locate the lower edge of the truck bolster spring through morphological calculation after OTSU binarization is performed on the large image of the truck bolster spring in the step s2 includes the following steps:
a. aiming at a large image of a swing bolster spring, obtaining a binary image by using an Otsu method, and then negating the binary image;
b. setting a structural unit with the width of w and the height of h, and carrying out morphological open operation to remove a spring gap communication area;
c. and positioning the hollow area by combining the position and the size information of the hollow area, and further obtaining the lower edge of the swing bolster spring.
10. The automatic detection method for the spring leap fault of the truck bolster as claimed in claim 9, further comprising the steps of:
after the fault is identified, calculating the position of the fault in the original image according to the mapping relation from the subgraph to the large image of the bolster spring and the mapping relation between the large image of the bolster spring and the original image, and displaying the fault on a display interface;
the subgraphs are the whole image, the top end subimage or the end subimage of the spring.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010940916.0A CN112085723B (en) | 2020-09-09 | 2020-09-09 | Automatic detection method for spring jumping fault of truck bolster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010940916.0A CN112085723B (en) | 2020-09-09 | 2020-09-09 | Automatic detection method for spring jumping fault of truck bolster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112085723A true CN112085723A (en) | 2020-12-15 |
| CN112085723B CN112085723B (en) | 2021-04-02 |
Family
ID=73732987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010940916.0A Active CN112085723B (en) | 2020-09-09 | 2020-09-09 | Automatic detection method for spring jumping fault of truck bolster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112085723B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651343A (en) * | 2020-12-28 | 2021-04-13 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon brake beam breaking fault identification method based on image processing |
| CN112699794A (en) * | 2020-12-29 | 2021-04-23 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying dislocation fault images of middle rubber and upper and lower floor plates of wagon axle box rubber pad |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000079481A1 (en) * | 1999-06-23 | 2000-12-28 | Massachusetts Institute Of Technology | Mra segmentation using active contour models |
| CN101837351A (en) * | 2010-06-02 | 2010-09-22 | 天津大学 | Oil seal spring full-automatic sorting system and method based on image detection method |
| CN102332089A (en) * | 2011-06-23 | 2012-01-25 | 北京康拓红外技术股份有限公司 | Railway wagon brake shoe key going-out fault recognition method based on artificial neural network |
| CN102509300A (en) * | 2011-11-18 | 2012-06-20 | 深圳市宝捷信科技有限公司 | Defect detection method and system |
| KR101590552B1 (en) * | 2014-08-11 | 2016-02-02 | 대원강업주식회사 | Curved spring shape inspection method |
| CN105701818A (en) * | 2016-01-14 | 2016-06-22 | 辽宁师范大学 | Multi-target image segmentation C-V method based on area division and gradient guiding |
| CN106600581A (en) * | 2016-12-02 | 2017-04-26 | 北京航空航天大学 | Train operation fault automatic detection system and method based on binocular stereoscopic vision |
| CN110567680A (en) * | 2018-06-05 | 2019-12-13 | 成都精工华耀科技有限公司 | Track fastener looseness detection method based on angle comparison |
| CN111091548A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon adapter dislocation fault image identification method and system based on deep learning |
| CN111091558A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon swing bolster spring jumping fault image identification method |
-
2020
- 2020-09-09 CN CN202010940916.0A patent/CN112085723B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000079481A1 (en) * | 1999-06-23 | 2000-12-28 | Massachusetts Institute Of Technology | Mra segmentation using active contour models |
| CN101837351A (en) * | 2010-06-02 | 2010-09-22 | 天津大学 | Oil seal spring full-automatic sorting system and method based on image detection method |
| CN102332089A (en) * | 2011-06-23 | 2012-01-25 | 北京康拓红外技术股份有限公司 | Railway wagon brake shoe key going-out fault recognition method based on artificial neural network |
| CN102509300A (en) * | 2011-11-18 | 2012-06-20 | 深圳市宝捷信科技有限公司 | Defect detection method and system |
| KR101590552B1 (en) * | 2014-08-11 | 2016-02-02 | 대원강업주식회사 | Curved spring shape inspection method |
| CN105701818A (en) * | 2016-01-14 | 2016-06-22 | 辽宁师范大学 | Multi-target image segmentation C-V method based on area division and gradient guiding |
| CN106600581A (en) * | 2016-12-02 | 2017-04-26 | 北京航空航天大学 | Train operation fault automatic detection system and method based on binocular stereoscopic vision |
| CN110567680A (en) * | 2018-06-05 | 2019-12-13 | 成都精工华耀科技有限公司 | Track fastener looseness detection method based on angle comparison |
| CN111091548A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon adapter dislocation fault image identification method and system based on deep learning |
| CN111091558A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon swing bolster spring jumping fault image identification method |
Non-Patent Citations (5)
| Title |
|---|
| 周云燕: "基于图像分析理论的机械故障诊断研究", 《中国博士学位论文全文数据库信息科技辑》 * |
| 张益 等: "基于数字图像分析的铁路货车闸瓦插销窜出故障自动识别方法", 《铁路计算机应用》 * |
| 戴鹏: "基于计算机视觉的故障自动识别系统的设计和实现", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》 * |
| 李永: "基于Otsu阈值和水平集算法的非接触指纹背景分割研究", 《计算机与数字工程》 * |
| 王斌: "基于机器视觉的货车转向架典型故障检测方法研究", 《中国优秀硕士学位论文全文数据库工程科技II辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651343A (en) * | 2020-12-28 | 2021-04-13 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon brake beam breaking fault identification method based on image processing |
| CN112699794A (en) * | 2020-12-29 | 2021-04-23 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying dislocation fault images of middle rubber and upper and lower floor plates of wagon axle box rubber pad |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112085723B (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113781402B (en) | Method and device for detecting scratch defects on chip surface and computer equipment | |
| CN113160192B (en) | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background | |
| CN110119741B (en) | Card image information identification method with background | |
| CN109489724B (en) | Tunnel train safe operation environment comprehensive detection device and detection method | |
| CN102663760B (en) | Location and segmentation method for windshield area of vehicle in images | |
| CN111046872B (en) | Optical character recognition method | |
| CN109540925B (en) | Complex ceramic tile surface defect detection method based on difference method and local variance measurement operator | |
| CN112085723B (en) | Automatic detection method for spring jumping fault of truck bolster | |
| CN101122953A (en) | Picture words segmentation method | |
| CN107452035B (en) | Method and apparatus for analyzing lane line image and computer readable medium thereof | |
| CN110070523B (en) | Foreign matter detection method for bottle bottom | |
| CN110047063B (en) | Material drop detection method, device, equipment and storage medium | |
| CN112560538B (en) | Method for quickly positioning damaged QR (quick response) code according to image redundant information | |
| CN108765456B (en) | Target tracking method and system based on linear edge characteristics | |
| CN111476804A (en) | Method, device and equipment for efficiently segmenting carrier roller image and storage medium | |
| CN110648330A (en) | Defect detection method for camera glass | |
| CN111027544A (en) | MSER license plate positioning method and system based on visual saliency detection | |
| CN108537815B (en) | Video image foreground segmentation method and device | |
| CN111652228B (en) | Railway wagon sleeper beam hole foreign matter detection method | |
| CN108205678A (en) | A kind of nameplate Text region processing method containing speck interference | |
| CN112132824B (en) | Automatic detection method for failure of freight car axle box spring | |
| CN116363655A (en) | Financial bill identification method and system | |
| CN116486129A (en) | Deep learning-based railway wagon cover plate fault identification method and device | |
| CN115937719A (en) | Runway contour line obtaining method based on maximum area | |
| CN111402185A (en) | Image detection method and device |
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 |