CN116542971B - Vehicle wheel axle defect identification method - Google Patents
Vehicle wheel axle defect identification method Download PDFInfo
- Publication number
- CN116542971B CN116542971B CN202310806339.XA CN202310806339A CN116542971B CN 116542971 B CN116542971 B CN 116542971B CN 202310806339 A CN202310806339 A CN 202310806339A CN 116542971 B CN116542971 B CN 116542971B
- Authority
- CN
- China
- Prior art keywords
- edge
- level
- obtaining
- points
- value
- 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 57
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000009826 distribution Methods 0.000 claims abstract description 43
- 230000011218 segmentation Effects 0.000 claims abstract description 41
- 244000144730 Amygdalus persica Species 0.000 claims abstract description 24
- 235000006040 Prunus persica var persica Nutrition 0.000 claims abstract description 24
- 238000007670 refining Methods 0.000 claims abstract description 8
- 238000012512 characterization method Methods 0.000 claims description 22
- 241000251468 Actinopterygii Species 0.000 claims description 6
- 238000003708 edge detection Methods 0.000 claims description 5
- 238000005452 bending Methods 0.000 description 9
- 241000270708 Testudinidae Species 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000010438 heat treatment Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000010791 quenching Methods 0.000 description 1
- 230000000171 quenching effect Effects 0.000 description 1
- 238000004381 surface treatment 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
-
- 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
- G06T7/13—Edge detection
-
- 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
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image processing, in particular to a vehicle wheel axle defect identification method. The method comprises the steps of obtaining a gray level image of a convex peach surface of a vehicle wheel axle, obtaining edges in the gray level image of the convex peach surface, refining to obtain a refined edge binary image, obtaining corresponding extension characteristic values according to refined edge sections of a net area in the refined edge binary image, grading edge segmentation points in the net area to obtain distribution levels, obtaining connection levels of the refined edge sections according to the distribution levels, obtaining comparison characteristics according to the size difference and the length mean value difference of the connection levels, adjusting the extension characteristic value mean value to obtain an integral extension characteristic value by taking the size of the connection levels as a weight, obtaining category characteristic values according to all the comparison characteristics and the integral extension characteristic values, and identifying the types of crack defects according to the average category characteristic values. According to the invention, the network crack defects are divided according to the image characteristics by the image recognition method, so that the defect type is detected.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a vehicle wheel axle defect identification method.
Background
The vehicle axle generally refers to parts such as a crankshaft and a camshaft of a vehicle, wherein the camshaft is one of the most important parts in an engine of the vehicle, and the surface defects of the camshaft are various, and the identification of the defects through the extraction of the surface defect characteristics is a main quality detection means in the production process of the camshaft.
The working surface of the camshaft needs to have higher flatness and hardness, so that the working surface of the camshaft needs to be subjected to surface treatment in the production process of the camshaft so as to meet the requirements, the working surface of the camshaft is generally ground by utilizing a grinding wheel after the camshaft is subjected to carburization, quenching and the like, and after the grinding process, the working surface of the camshaft may have net-shaped grinding crack defects, which affect the strength of the surface of the camshaft and can lead to the scrapping of the camshaft, the causes of the net-shaped grinding cracks can be roughly divided into two causes, one is caused by improper heat treatment of a workpiece, one is caused by improper grinding process, and the net-shaped defects caused by two different causes are also different. The existing feature recognition method mainly divides defect types by bending and straightness degrees of edges of two net-shaped defect shapes, but the division form is too single and has larger classification errors, and the classification result is not accurate enough.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a vehicle wheel axle defect identification method, which adopts the following technical scheme:
the invention provides a vehicle wheel axle defect identification method, which comprises the following steps:
obtaining a gray image of the convex peach surface of a vehicle wheel axle; obtaining an edge binary image of the convex peach surface gray level image; refining the edges in the edge binary image to obtain a refined edge binary image containing refined edge points; the refined edge binary image comprises a netlike area formed by at least two edge segmentation points and corresponding refined edge sections;
obtaining extension directions of adjacent pixel points in the thinned edge section in the mesh region, and obtaining an extension characteristic value of the thinned edge section according to the frequency of occurrence of each extension direction;
selecting any edge segmentation point as an initial level point, and obtaining the distribution level of each edge segmentation point according to the connection level of other edge segmentation points in the mesh region and the initial level point; obtaining a connection level corresponding to the thinned edge section according to the distribution level of the edge segmentation points; obtaining the length average value of the thinned edge section in each connection level; obtaining an extension characteristic value average value of the thinned edge section in each connection level; obtaining a comparison characteristic according to the difference of any two different connection levels and the corresponding difference of the length mean values;
the size of the connection level is taken as weight to adjust the average value of the corresponding extension characteristic values, so that the integral extension characteristic value of each connection level is obtained; obtaining class characterization values corresponding to the initial level points according to all the comparison features and the integral extension feature values;
and obtaining the class representation value corresponding to each edge segmentation point, and identifying the type of the crack defect according to the average class representation value.
Further, the obtaining of the extension characteristic value includes:
obtaining the extending direction between two adjacent points on the thinned edge section, and counting the times of the extending direction in a preset extending direction type; and taking the ratio of the occurrence times of each extending direction to the total extending times on the thinned edge section as an extending duty ratio, calculating entropy of the extending directions according to all extending duty ratios, and taking an entropy value as an extending characteristic value of the thinned edge section.
Further, the obtaining each edge segmentation point level according to the connection hierarchy of the other edge segmentation points and the initial level point in the mesh region includes:
setting the distribution level of the initial level points to be level 0; setting the distribution level of the edge division points directly connected with the initial level point to be 1 level; setting the distribution level of the edge segmentation points of an undetermined level directly connected with a level 1 edge segmentation point to be level 2; continuing to extend to finish grading all the edge segmentation points, and obtaining the distribution grade of all the edge segmentation points.
Further, the obtaining the connection level corresponding to the thinned edge segment according to the edge segmentation point level includes:
for one thinned edge section, adding one to the distribution level of the edge segmentation point with smaller distribution level in the two corresponding edge segmentation points as the connection level of the thinned edge section; the connection levels of all the thinned edge segments are obtained.
Further, the method for acquiring the contrast characteristic comprises the following steps:
selecting any two different connection levels from all the connection levels as a comparison group, and obtaining all the comparison groups;
constructing a comparison feature according to the difference of the connection levels of the comparison group and the corresponding difference of the length mean values, wherein the comparison feature comprises:
wherein ,is->Contrast characteristics of each of said contrast groups, +.>Is->One of said connection levels in said comparison group,/or->Is->The connection level of the other of said comparison sets,/->Is->The length average value corresponding to the connection level, +.>Is->The length average value corresponding to the connection level, +.>Indicating that the initial level point is +.>Maximum value of the connection level at each edge segmentation point, +.>Representing the connection level,/->Indicating that the initial level point is +.>And each edge dividing point.
Further, the method for obtaining the integral extension characteristic value comprises the following steps:
and adjusting the corresponding average value of the extension characteristic values by taking the size of the connection level as a weight to obtain an overall extension characteristic value, wherein the overall extension characteristic value comprises:
wherein ,for the initial level point is +.>The>Said global extension characteristic value of the connection level, < >>For the initial level point is +.>The>Said extended eigenvalue mean of connection level,/->Indicating that the initial level point is +.>Maximum value of the connection level at each edge segmentation point, +.>Representing the connection level,/->Indicating that the initial level point is +.>And each edge dividing point.
Further, the method for obtaining the category characterization value comprises the following steps:
obtaining an accumulated value of all the comparison features, and obtaining an accumulated value of the overall extension feature values of all the connection levels; and taking the product of the accumulated value of the comparison characteristic and the accumulated value of the integral extension characteristic value as a class representation value corresponding to the initial level point.
Further, the discriminating the type of the crack defect according to the average class characterization value includes:
when the average class characterization value is greater than or equal to a preset class threshold, the type of the crack defect is tortoiseshell;
and when the average class representation value is smaller than a preset class threshold value, the type of the crack defect is fish scale.
Further, the method for acquiring the refined edge binary image comprises the following steps:
obtaining an edge binary image of the convex peach surface gray level image by adopting canny edge detection on the obtained convex peach surface gray level image; and refining the edges of the edge binary image by adopting an HSCP edge refining algorithm to obtain a refined edge binary image containing the refined edge points.
Further, the mesh region specifically includes:
performing region growth on non-refined edge points in the refined edge binary image by adopting a region growing algorithm to obtain a connected domain formed by the non-refined edge points, taking the connected domain surrounded by the refined edge points and a region surrounded by an image boundary and the edge of the mesh region as independent regions, and numbering each independent region to obtain a category number;
acquiring the category number of the category numbers adjacent to each refined edge point; the refined edge points with the category number more than or equal to 3 are used as edge segmentation points; the continuous thinned edge points between the two edge segmentation points form a thinned edge section;
and forming a net-shaped area by the edge dividing points and the thinned edge sections.
The invention has the following beneficial effects:
according to the method, firstly, the difference of the bending degree of the edge line segments in two crack images is considered, firstly, the bending degree and the straight degree of the thinned edge are analyzed, the extending direction of the pixel points in the thinned edge segment is analyzed, and an extending characteristic value is obtained, wherein the extending characteristic value represents the bending degree and the straight degree of the thinned edge segment. And grading edge division points and thinned edge sections in the thinned edge binary image to obtain connection levels of the thinned edge sections, wherein the connection levels represent distribution conditions of the thinned edge sections in the image, and image characteristics can be analyzed according to the distribution conditions. The smaller the refinement edge segment connection level, i.e. the closer to the initial level point, the more its extension feature value is considered for the refinement edge segment, thus constructing an overall extension feature value. And the length of the two net-shaped cracks is also characterized by variability when the positions of the thinned edge sections are far apart, the position lengths of the thinned edge sections are analyzed to obtain comparison characteristics, the comparison characteristics represent the sizes of the differences of the extension lengths of the two groups of thinned edge sections, which are influenced by the positions, and the larger the position difference is, the more the reference significance is given to the differences of the extension lengths. And obtaining class characterization values corresponding to the initial level points by the contrast characteristic and the integral extension characteristic values, obtaining class characterization values of all edge points, wherein the average class characterization values represent the straight bending condition of the thinned edge on the image and the difference condition of the length of the thinned edge with larger distance difference, and dividing the crack according to the average class characterization values. The embodiment of the invention not only distinguishes the two cracks according to the bending and straight degree of the edge line segment change in the image, but also considers the characteristic of the difference of the length of the thinned edge when the distance of the thinned edge is far, comprehensively considers the characteristic difference of the two cracks, ensures that the distinguishing result is more accurate and comprehensive, more accurately divides the net-shaped cracks and distinguishes the causes of the net-shaped grinding cracks.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for identifying defects of a vehicle axle according to an embodiment of the present invention;
FIG. 2 is a refined edge binary image provided by one embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description refers to the specific implementation, structure, characteristics and effects of a vehicle wheel axle defect identification method according to the invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the method for identifying the defects of the wheel axle of the vehicle provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a method for identifying a defect of a vehicle axle according to an embodiment of the invention is shown, where the method includes:
step S1: obtaining a gray image of the convex peach surface of a vehicle wheel axle; obtaining an edge binary image of the gray level image of the convex peach surface; thinning the edges in the edge binary image to obtain a thinned edge binary image containing thinned edge points; the refined edge binary image comprises a netlike region formed by at least two edge segmentation points and corresponding refined edge segments.
In the embodiment of the invention, a high-speed industrial camera is deployed above a production line, and the high-speed industrial camera is utilized to shoot the convex peach surface image of each vehicle wheel axle on the production line, so that the convex peach surface image is grayed to obtain the gray image of the convex peach surface of the vehicle wheel axle for facilitating subsequent processing.
Because the cracks on the convex peach surface of the vehicle wheel axle have obvious edge texture characteristics, the network crack defects can be identified according to the edge information acquired in the images, and therefore, a refined edge binary image of the convex peach surface is required to be acquired before the network crack defects are identified.
Preferably, the gray level image of the convex peach surface of the camshaft of the vehicle is obtained, when the convex peach surface is cracked, the gray level image can be detected in the image through edge detection, and the edge of the gray level image of the convex peach surface is obtained by using a canny edge detection algorithm, so that an edge binary image is obtained. It should be noted that, the convex peach surface of the vehicle wheel axle with qualified quality should be smooth, that is, no edge information exists in the gray level image of the corresponding convex peach surface, so the obtained edge information indicates that a defect exists in the image. Because the edge obtained by the canny edge detection algorithm is not formed by connecting single pixel points, the edge analysis is not facilitated, the edge binary image is subjected to edge refinement by adopting the HSCP edge refinement algorithm to obtain refined edge points, the refined edge points are set to be 0, and the non-refined edge points are set to be 1, so that the refined edge binary image is obtained.
Referring to fig. 2, a refined edge binary image is shown. The crack defect area forms a net area in the thinned edge binary image, a plurality of edge dividing points and a plurality of thinned edge sections exist in the net area, namely the net area is composed of the edge dividing points and the thinned edge sections, but edge lines without reference meaning and corresponding edge points exist in the directly acquired net area, so that the net area is divided so as to analyze the image, and the method specifically comprises the following steps:
and carrying out region growth on non-refined edge points in the refined edge binary image by adopting a region growth algorithm to obtain a connected domain consisting of the non-refined edge points, taking the connected domain surrounded by the refined edge points and the region surrounded by the image boundary and the edge of the mesh region as independent regions, and numbering each independent region to obtain class numbers of all the independent regions. Classifying the refined edge points of the mesh region according to the number of the class numbers to obtain the number of the class numbers of each refined edge point neighbor, if the number of the class numbers of the refined edge point neighbor is more than or equal to 3, indicating that the corresponding refined edge point is positioned at the junction of three or more independent regions, and classifying the refined edge point into edge division points, namely, class A points in fig. 2; if the number of class numbers of the neighbor of the thinned edge point is equal to 2, describing that the corresponding thinned edge point is a junction point of two independent areas, and taking the thinned edge point as one thinned edge point forming a thinned edge section, namely a class B point in fig. 2; if the number of class numbers of the neighbor of the thinned edge point is equal to 1, the thinned edge point corresponding to the thinned edge point is indicated to be the thinned edge point in the independent area, namely the thinned edge point is the edge point extending to the inside of the independent area on the thinned edge surrounding the independent area, the thinned edge point is marked as a screening point, the position of the screening point is replaced by a non-thinned edge point, and the class number of the independent area where the screening point is located is marked, namely the class C point in fig. 2. A line segment is formed by continuous thinned edge points between the two edge segmentation points, and the line segment is the thinned edge segment.
Step S2: and obtaining the extending directions of the adjacent pixel points in the thinned edge section in the network area, and obtaining the extending characteristic value of the thinned edge section according to the frequency of each extending direction.
Since each thinned edge section represents the boundary form of the grid formed by the mesh-shaped crack, the grid of the tortoise-shell-shaped crack, namely the form of each thinned edge section, is a locally rigid folded line section, and the grid boundary of the fish scale-shaped crack, namely the form of each thinned edge section, is a curved curve, the boundary extension characteristic of the tortoiseshell-shaped crack in the extension direction is that the extension direction of the point on the thinned edge section tends to be one or two, and the boundary extension characteristic of the fish scale-shaped crack is that the extension direction of the point on the thinned edge section tends to be different. Therefore, the extension characteristics of the thinned edges on each thinned edge section in the mesh region need to be analyzed, the extension directions of all adjacent pixel points in all the thinned edge sections are obtained in the mesh region, the boundary extension characteristics of the corresponding crack defects can be represented according to the distribution characteristics of the extension directions, namely, the more uniform the distribution of the extension directions, the closer the boundary extension characteristics in the mesh region are to the tortoise shell-shaped cracks; the more discrete the extension direction distribution, the closer the boundary extension characteristics in the grid region are to the scale-like cracks. The extension characteristic value of the thinned edge section is obtained according to the frequency of each extension direction, and the extension characteristic value specifically comprises:
and extending from one edge dividing point in one thinning edge section to the other edge dividing point by taking pixel points on two adjacent thinning edge points to obtain corresponding extending directions in preset extending directions, counting the times of occurrence of each extending direction in the whole thinning edge section, taking the ratio of the occurrence times of the extending directions to the total extending times as an extending ratio, wherein the extending ratio represents the trend of the extending directions, calculating entropy of the extending directions according to the extending ratio of each thinning edge section, and taking the entropy value as an extending characteristic value of the corresponding thinning edge section.
In the embodiment of the present invention, 8 extension directions are set for the preset extension direction types, and when there is a preset direction that does not appear, the setting may be performed according to a specific direction trend, for example, the extension directions are initially set to 8 types, but in actual statistics, the extension directions lack one type, and then the extension directions may be set to 7 types. It should be noted that, entropy calculation is a technical means well known to those skilled in the art, and because the extending directions of the embodiments of the present invention are set to 8 types, the corresponding entropy calculation formula is:
in the formula ,expressed as entropy->Represented as +.>Strip thinning edge section->Denoted as +.>Direction of extension of->Denoted as +.>Strip thinning edge section +.>The ratio of the number of occurrences of the extension direction to the total number of extensions. When->The more toward 1, the more the direction of extension tends to be, the +.>The more curved the strip-thinned edge section is, when +>The closer to 0, the more consistent the extension direction, the +.>The strip thinning edge Duan Yueping is straight.
Step S3: selecting any edge dividing point as an initial level point, and obtaining the distribution level of each edge dividing point according to the connection level of other edge dividing points and the initial level point in the mesh area; obtaining the connection level of the corresponding thinned edge section according to the distribution level of the edge segmentation points; obtaining the length average value of the thinned edge section in each connection level; obtaining an extension characteristic value average value of the thinned edge section in each connection level; and obtaining the comparison characteristic according to the difference of any two different connection levels and the difference of the corresponding length average values.
The fine edge sections in the mesh region are classified, so that the distribution characteristics of the fine edge sections in the mesh region can be conveniently analyzed, two cracks are better distinguished through the distribution characteristics, the lengths of the fine edge sections with the far distance of tortoiseshell cracks are larger, and the lengths of the fine edge sections with the far distance of fish scale cracks are smaller, so that the distribution characteristics of the fine edge sections can be analyzed.
The refined edge segments are classified by classifying the edge dividing points, the distribution level of the edge dividing points is firstly obtained, any one edge dividing point is selected as an initial level point, and the distribution level of each edge dividing point is obtained according to the connection level of other edge dividing points and the initial level point in the mesh region. The method specifically comprises the following steps:
presetting an edge dividing point as an initial level point, grading the initial level point to be 0, grading the distribution level of the edge dividing points directly connected with the initial level point to be 1, grading the distribution level of the edge dividing points directly connected with the 1-level edge dividing points and not graded to be 2, and the like, so that all the edge dividing points finish grading, and obtaining the distribution level of the edge dividing points.
Further, according to the distribution level of the edge dividing points, the connection level of the corresponding thinned edge section is obtained, which specifically comprises:
for one thinned edge section, two edge division points are arranged, a level with smaller distribution level in the two edge division points is added as a connection level of the thinned edge section, the connection levels of all the thinned edge sections are obtained, and the distribution distance condition of the thinned edge sections can be represented by the difference value of the connection levels.
For two different netlike crack defects, the difference of the corresponding edge lengths of the edges with larger distance difference in the image is also different, and if the difference of the corresponding edge lengths of the edges with larger distance difference is larger, the crack defect is closer to the tortoise shell-shaped crack; if the edges with larger differences are similar in the corresponding edge length, the crack defect is closer to the scale-shaped crack. Therefore, a comparison characteristic is required to be constructed according to the distribution condition of the thinned edge sections in the mesh region, and the specific quantity involved in comparison is the length of the thinned edge sections in the connection level and the size of the connection level, so that the length average value of the thinned edge sections in each connection level is obtained; obtaining contrast characteristics according to the size difference of any two different connection levels and the difference of corresponding length mean values, wherein the contrast characteristics specifically comprise:
selecting any two different connection levels from all connection levels as a comparison group, obtaining all the comparison groups of the connection levels, constructing comparison features through differences of the connection levels of the comparison groups and differences of corresponding length mean values, wherein the comparison features represent comparison of position distribution differences of the connection levels and the length mean values, and a comparison feature formula comprises:
in the formula ,indicate->Contrast characteristics of the individual contrast groups->Is->One connection level in a comparison group, < >>Is->Another connection level in the comparison group, +.>Is->Length mean value corresponding to connection level, +.>Is->Length mean value corresponding to connection level, +.>Maximum value representing connection level, +.>Representing the connection level +.>Indicating the initial level point as +.>The individual edges divide the points. />Representing the difference of the length mean values, comparing the difference of the length mean values of the two connection levels, the larger the length difference is, the larger the comparison feature is. />Representing the difference of the connection levels and carrying out normalization processing on the comparison features, wherein when the positions of two refinement edge segment sets in an image differ far more than the two connection levels differ in the comparison group, the comparison features are larger, and the comparison results of the comparison features have reference significance.
Step S4: the size of the connection level is used as a weight to adjust the average value of the corresponding extension characteristic values, and the integral extension characteristic value of each connection level is obtained; and obtaining a class characterization value corresponding to the initial level point according to all the comparison features and the integral extension feature values.
In the mesh region, when any one of the edge division points is taken as a start point, the degree of bending and flatness of the thinned edge section closer to the start point is taken as one of the characteristics of class distinction, and the difference in length of the thinned edge closer to and farther from the start point is taken as the other characteristic of class distinction.
One of the characteristics of the classification is to consider the bending degree and the flatness degree of the thinned edge section close to the starting point, so that the influence of the extension characteristic value of the thinned edge section with smaller connection level is larger, according to the obtained extension characteristic value average value of all the thinned edge sections in each connection level, the corresponding extension characteristic value average value is adjusted by taking the size of the connection level as a weight, and the overall extension characteristic value of each connection level is obtained, wherein the overall extension characteristic value represents the comprehensive calculation of the extension characteristic value of the corresponding thinned edge section according to the size of the connection level, and the overall extension characteristic value comprises:
in the formula ,is the initial level point is +>The>The overall extension characteristic value of the connection level,is the initial level point is +>The>Extension feature value mean of connection level, +.>Indicating the initial level point as +.>Maximum value of connection level under each edge division point, +.>Representing the connection level +.>Indicating the initial level point as +.>The individual edges divide the points. />The smaller the connection level is, the more attention is paid to the extension characteristic of the connection level and the final integral extension characteristic value is normalized>Indicating when->The smaller the extension characteristic is, the closer to straight, and the larger the overall extension characteristic value is.
All the integral extension characteristic values and all the comparison characteristics are obtained, and the characteristic distinction of two categories can be integrally considered, so that the category characterization value corresponding to the initial level point is obtained according to all the comparison characteristics and the integral extension characteristic values, wherein the category characterization value is a reference index for dividing two net-shaped crack defects, and the method specifically comprises the following steps:
firstly, obtaining an accumulated value of the comparison characteristic, then obtaining an accumulated value of the integral extension characteristic value of all the connection levels, taking the product of the accumulated value of the comparison characteristic and the accumulated value of the integral extension characteristic value as a class characterization value corresponding to an initial level point, namely, the expression is:
in the formula ,expressed as +.>The individual edge division points are class representation values corresponding to the initial level points, and the +.>Indicate->Contrast characteristics of the individual contrast groups->Expressed as +.>The individual edge division points are the integral extension characteristic values corresponding to the initial level points, and the +.>Maximum value representing connection level, +.>Representing the connection level +.>Indicating the initial level point as +.>Individual edge division points->Expressed as +.>The individual edge dividing points are the accumulated values of the contrast characteristics corresponding to the initial level points, ++>Indicating the number of all contrast groups selected by the connection level, +.>Expressed as +.>The individual edge segmentation points are accumulated values of the integral extension characteristic values corresponding to the initial level points, and when the comparison characteristic is larger, the integral extension characteristic values are larger, the class characterization values corresponding to the initial level points are larger. Because the accumulated value of the comparison feature value and the accumulated value of the overall feature value are both values within 0 and 1, the resulting class characterization value is also a value between 0 and 1.
Step S5: and obtaining a class representation value corresponding to each edge segmentation point, and identifying the type of the crack defect according to the average class representation value.
For thinning edgesIn the binary image, an average class representation value capable of dividing crack defects is obtained according to class representation values of all edge division points in the mesh region, namely, each edge division point corresponds to one class representation value, and the average class representation value is calculated so as to obtainIndicating that the average class representation value +.>And also ranges between 0 and 1.
When the average class characterization value is more 1, the comparison feature of the thinned edge section in the mesh region is more larger, the extension feature value is more 0, and the thinned edge section which is closer to the starting point is more straight by taking any one edge division point as the starting point, and the length difference of the thinned edge section which is closer to the starting point and the length difference of the thinned edge section which is farther from the starting point are more likely to be tortoise shell-shaped cracks. When the average class characterization value is more 0, the smaller the contrast characteristic of the thinned edge section in the mesh region is, the more 1 the extension characteristic value is, and the closer to the start point, the closer to the thinned edge section is to the curve, and the closer to the start point, the length of the thinned edge section is similar to the length of the thinned edge section, the more likely the mesh crack is a scale crack.
In the embodiment of the invention, the class threshold value is set to be 0.5, and when the average class characterization value is greater than or equal to the class threshold value, the crack defect type is tortoise shell-shaped, which indicates that the defect cause is improper heat treatment of the workpiece. When the average class characterization value is smaller than the class threshold value, the crack defect type is fish scale, which indicates that the defect is caused by improper grinding process.
In summary, according to the embodiment of the invention, the edge in the convex peach surface gray level image of the wheel axle is obtained by obtaining the edge in the convex peach surface gray level image and refining the edge to obtain a refined edge binary image, the extension characteristic value of the refined edge section is obtained according to the net region consisting of the edge dividing points and the refined edge section in the refined edge binary image, the edge dividing points in the net region are graded to obtain the distribution level of the edge dividing points, the connection level of the refined edge section is obtained according to the distribution level, the comparison characteristic is obtained according to the difference of the connection level and the length average value of the refined edge section, the average value of the extension characteristic value is adjusted by taking the size of the connection level as the weight to obtain the integral extension characteristic value, the category characteristic value is obtained according to the comparison characteristic and the integral extension characteristic value of all the connection levels, and the type of the crack defect is identified according to the average category characteristic value. The embodiment of the invention not only judges the defects through the straight and bending degrees of the edges, but also integrally considers the edge distribution characteristics, so that the judging standard is not single, the type of the netlike crack defects can be judged more comprehensively and accurately, and the causes of the crack defects can be better reflected.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.
Claims (10)
1. A method for identifying a vehicle axle defect, the method comprising:
obtaining a gray image of the convex peach surface of a vehicle wheel axle; obtaining an edge binary image of the convex peach surface gray level image; refining the edges in the edge binary image to obtain a refined edge binary image containing refined edge points; the refined edge binary image comprises a netlike area formed by at least two edge segmentation points and corresponding refined edge sections;
obtaining extension directions of adjacent pixel points in the thinned edge section in the mesh region, and obtaining an extension characteristic value of the thinned edge section according to the frequency of occurrence of each extension direction;
selecting any edge segmentation point as an initial level point, and obtaining the distribution level of each edge segmentation point according to the connection level of other edge segmentation points in the mesh region and the initial level point; obtaining a connection level corresponding to the thinned edge section according to the distribution level of the edge segmentation points; obtaining the length average value of the thinned edge section in each connection level; obtaining an extension characteristic value average value of the thinned edge section in each connection level; obtaining a comparison characteristic according to the difference of any two different connection levels and the corresponding difference of the length mean values;
the size of the connection level is taken as weight to adjust the average value of the corresponding extension characteristic values, so that the integral extension characteristic value of each connection level is obtained; obtaining class characterization values corresponding to the initial level points according to all the comparison features and the integral extension feature values;
and obtaining the class representation value corresponding to each edge segmentation point, and identifying the type of the crack defect according to the average class representation value.
2. The method for identifying a vehicle axle defect according to claim 1, wherein the obtaining of the extended feature value includes:
obtaining the extending direction between two adjacent points on the thinned edge section, and counting the times of the extending direction in a preset extending direction type; and taking the ratio of the occurrence times of each extending direction to the total extending times on the thinned edge section as an extending duty ratio, calculating entropy of the extending directions according to all extending duty ratios, and taking an entropy value as an extending characteristic value of the thinned edge section.
3. A vehicle wheel axle defect identification method according to claim 1, wherein said obtaining a distribution level of each of said edge division points from a connection hierarchy of other edge division points and said initial level point in said mesh region comprises:
setting the distribution level of the initial level points to be level 0; setting the distribution level of the edge division points directly connected with the initial level point to be 1 level; setting the distribution level of the edge segmentation points of an undetermined level directly connected with a level 1 edge segmentation point to be level 2; continuing to extend to finish grading all the edge segmentation points, and obtaining the distribution grade of all the edge segmentation points.
4. A vehicle wheel axle defect identification method according to claim 1 or 3, wherein said obtaining a connection level corresponding to said thinned edge section from a distribution level of said edge dividing points comprises:
for one thinned edge section, adding one to the distribution level of the edge segmentation point with smaller distribution level in the two corresponding edge segmentation points as the connection level of the thinned edge section; the connection levels of all the thinned edge segments are obtained.
5. The method for identifying defects of a vehicle wheel axle according to claim 1, wherein the method for acquiring the comparison features comprises:
selecting any two different connection levels from all the connection levels as a comparison group, and obtaining all the comparison groups;
constructing a comparison feature according to the difference of the connection levels of the comparison group and the corresponding difference of the length mean values, wherein the comparison feature comprises:
wherein ,is->Contrast characteristics of each of said contrast groups, +.>Is->One of said connection levels in said comparison group,/or->Is->The connection level of the other of said comparison sets,/->Is->The length average value corresponding to the connection level,is->The length average value corresponding to the connection level, +.>Indicating that the initial level point is +.>Maximum value of the connection level at each edge segmentation point, +.>Representing the connection level,/->Indicating that the initial level point is +.>And each edge dividing point.
6. The method for identifying a vehicle axle defect according to claim 1, wherein the method for acquiring the integral extension feature value comprises:
and adjusting the corresponding average value of the extension characteristic values by taking the size of the connection level as a weight to obtain an overall extension characteristic value, wherein the overall extension characteristic value comprises:
wherein ,for the initial level point is +.>The>Said global extension characteristic value of the connection level, < >>For the initial level point is +.>The>Said extended eigenvalue mean of connection level,/->Indicating that the initial level point is +.>Maximum value of the connection level at each edge segmentation point, +.>Representing the connection level,/->Indicating that the initial level point is +.>And each edge dividing point.
7. The method for identifying a vehicle axle defect according to claim 1, wherein the method for obtaining the class characterization value comprises:
obtaining an accumulated value of all the comparison features, and obtaining an accumulated value of the overall extension feature values of all the connection levels; and taking the product of the accumulated value of the comparison characteristic and the accumulated value of the integral extension characteristic value as a class representation value corresponding to the initial level point.
8. A method of identifying a vehicle axle defect according to claim 1, wherein identifying the type of crack defect based on the average class characterization value comprises:
when the average class characterization value is greater than or equal to a preset class threshold, the type of the crack defect is tortoiseshell;
and when the average class representation value is smaller than a preset class threshold value, the type of the crack defect is fish scale.
9. The method for identifying defects of a vehicle wheel axle according to claim 1, wherein the method for acquiring the thinned edge binary image comprises:
obtaining an edge binary image of the convex peach surface gray level image by adopting canny edge detection on the obtained convex peach surface gray level image; and refining the edges of the edge binary image by adopting an HSCP edge refining algorithm to obtain a refined edge binary image containing the refined edge points.
10. A method for identifying defects in a vehicle axle according to claim 1, wherein said meshed area comprises:
performing region growth on non-refined edge points in the refined edge binary image by adopting a region growing algorithm to obtain a connected domain formed by the non-refined edge points, taking the connected domain surrounded by the refined edge points and a region surrounded by an image boundary and the edge of the mesh region as independent regions, and numbering each independent region to obtain a category number;
acquiring the category number of the category numbers adjacent to each refined edge point; the refined edge points with the category number more than or equal to 3 are used as edge segmentation points; the continuous thinned edge points between the two edge segmentation points form a thinned edge section;
and forming a net-shaped area by the edge dividing points and the thinned edge sections.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310806339.XA CN116542971B (en) | 2023-07-04 | 2023-07-04 | Vehicle wheel axle defect identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310806339.XA CN116542971B (en) | 2023-07-04 | 2023-07-04 | Vehicle wheel axle defect identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116542971A CN116542971A (en) | 2023-08-04 |
| CN116542971B true CN116542971B (en) | 2023-08-29 |
Family
ID=87458122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310806339.XA Active CN116542971B (en) | 2023-07-04 | 2023-07-04 | Vehicle wheel axle defect identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116542971B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117195005B (en) * | 2023-11-03 | 2024-01-26 | 山东四季车网络科技有限公司 | Information data management system based on wisdom carwash |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2011103755A (en) * | 2011-02-02 | 2012-08-10 | Открытое акционерное общество "Новокузнецкий металлургический комбинат" (RU) | METHOD FOR DIAGNOSIS OF QUALITY OF SURFACE STRUCTURE OF METAL-ROLLING |
| RU2013158894A (en) * | 2013-12-30 | 2015-07-10 | Федеральное государственное бюджетное образовательное учреждение высшего профессионального образования "Московский государственный технический университет имени Н.Э. Баумана" (МГТУ им. Н.Э. Баумана") | METHOD FOR DIAGNOSIS OF CYCLIC MACHINES - METAL-CUTTING MACHINES BY PHASE-CHRONOMETRIC METHOD |
| CN111429403A (en) * | 2020-02-26 | 2020-07-17 | 北京航空航天大学杭州创新研究院 | Automobile gear finished product defect detection method based on machine vision |
| CN116309409A (en) * | 2023-02-28 | 2023-06-23 | 浙江工商大学 | Weld defect detection method, system and storage medium |
-
2023
- 2023-07-04 CN CN202310806339.XA patent/CN116542971B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2011103755A (en) * | 2011-02-02 | 2012-08-10 | Открытое акционерное общество "Новокузнецкий металлургический комбинат" (RU) | METHOD FOR DIAGNOSIS OF QUALITY OF SURFACE STRUCTURE OF METAL-ROLLING |
| RU2013158894A (en) * | 2013-12-30 | 2015-07-10 | Федеральное государственное бюджетное образовательное учреждение высшего профессионального образования "Московский государственный технический университет имени Н.Э. Баумана" (МГТУ им. Н.Э. Баумана") | METHOD FOR DIAGNOSIS OF CYCLIC MACHINES - METAL-CUTTING MACHINES BY PHASE-CHRONOMETRIC METHOD |
| CN111429403A (en) * | 2020-02-26 | 2020-07-17 | 北京航空航天大学杭州创新研究院 | Automobile gear finished product defect detection method based on machine vision |
| CN116309409A (en) * | 2023-02-28 | 2023-06-23 | 浙江工商大学 | Weld defect detection method, system and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 焊接缺陷磁光成像磁场分布数值模拟与试验分析;代欣欣;郑俏俏;季玉坤;高向东;;焊接学报(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116542971A (en) | 2023-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115829883B (en) | Surface image denoising method for special-shaped metal structural member | |
| CN117011297B (en) | Aluminum alloy automobile accessory die defect detection method based on image processing | |
| CN115311292B (en) | Strip steel surface defect detection method and system based on image processing | |
| CN107705314B (en) | Multi-object image segmentation method based on gray level distribution | |
| CN116542971B (en) | Vehicle wheel axle defect identification method | |
| CN115457041A (en) | Road quality identification and detection method | |
| CN115294158A (en) | Hot continuous rolling strip steel image segmentation method based on machine vision | |
| CN116823826B (en) | Numerical control machine tool tipping abnormity detection method | |
| CN115496692B (en) | Lubricating oil abrasive particle image enhancement method | |
| CN116092013B (en) | Dangerous road condition identification method for intelligent monitoring | |
| CN116777941B (en) | Profile contour detection method and system based on machine vision | |
| CN115578374A (en) | Mechanical part casting quality evaluation method and system | |
| CN115049651B (en) | Metal plate stamping abnormity detection method | |
| CN115861320A (en) | Intelligent detection method for automobile part machining information | |
| CN114723700A (en) | Iron furnace surface crack growth prediction method and system based on computer vision | |
| CN103500343A (en) | Hyperspectral image classification method based on MNF (Minimum Noise Fraction) transform in combination with extended attribute filtering | |
| CN116912234A (en) | Cable stranded wire quality detection method based on image features | |
| CN115082427B (en) | Aluminum material calendering process parameter control method based on artificial intelligence | |
| CN117911792B (en) | Pin detecting system for voltage reference source chip production | |
| CN109858544A (en) | The steel product quality detection method clustered based on section shade collection and density peaks | |
| CN116486091B (en) | Fan blade defect area rapid segmentation method and system based on artificial intelligence | |
| CN103268492B (en) | A kind of corn grain type identification method | |
| CN118115497B (en) | Quartz sand crushing and grinding detection method and device | |
| CN116051543A (en) | Defect identification method for peeling steel | |
| CN115018835A (en) | Automobile starter gear detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Method for Identifying Defects in Vehicle Wheel Axles Granted publication date: 20230829 Pledgee: Kexin Financial Guarantee Co.,Ltd. Pledgor: Shandong sijiche Network Technology Co.,Ltd. Registration number: Y2024980022664 |