CN116758067B - Metal structural member detection method based on feature matching - Google Patents
Metal structural member detection method based on feature matching Download PDFInfo
- Publication number
- CN116758067B CN116758067B CN202311027608.9A CN202311027608A CN116758067B CN 116758067 B CN116758067 B CN 116758067B CN 202311027608 A CN202311027608 A CN 202311027608A CN 116758067 B CN116758067 B CN 116758067B
- Authority
- CN
- China
- Prior art keywords
- points
- point
- contour
- edge
- gradient
- 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
- 239000002184 metal Substances 0.000 title claims abstract description 105
- 238000001514 detection method Methods 0.000 title claims abstract description 33
- 238000012512 characterization method Methods 0.000 claims abstract description 167
- 238000012216 screening Methods 0.000 claims abstract description 9
- 238000000034 method Methods 0.000 claims description 52
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 4
- 238000004458 analytical method Methods 0.000 description 10
- 239000011159 matrix material Substances 0.000 description 5
- 238000003708 edge detection Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000001746 injection moulding Methods 0.000 description 1
- 238000005495 investment casting Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000007789 sealing Methods 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- 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
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention relates to the technical field of image data processing, in particular to a metal structure detection method based on feature matching, which comprises the steps of firstly analyzing gradient changes of pixel points in a preset neighborhood region of an edge point to be analyzed of a metal structure to be detected to obtain an optimized gradient direction of the pixel points in the neighborhood region; obtaining a direction area, combining pixel point positions and optimizing gradient directions to obtain gradient characteristic parameters, and screening out profile characterization points; and obtaining the angle area of the contour characterization point, and obtaining the characteristic information of the contour characterization point according to the number of the pixel points, the curvature characteristic and the pixel value. And constructing a cost function according to the characteristic information difference of the standard edge pixel points and the contour characterization points, and matching to obtain a matching pair. And detecting the outline of the metal structural part to be detected according to the difference of the characteristic parameters in all the matched pairs. The embodiment of the invention can realize automatic detection of the appearance outline of the metal structural part, can reduce the detection amount of the system and ensure the detection precision of the system.
Description
Technical Field
The invention relates to the technical field of image data processing, in particular to a metal structural member detection method based on feature matching.
Background
The metal structural part is a basic accessory in the fields of industrial production and machinery, and almost all industries are covered. In the production process of the metal structural part, a plurality of processes such as processing, stamping, precision casting, injection molding and the like are required, and each process flow is required to be strictly controlled so as to ensure the quality of the final metal structural part product. Appearance detection of the metal structural part is crucial to quality evaluation of the metal structural part, when appearance contours of the metal structural part deviate, problems that the structural part cannot be matched with equipment or sealing performance is poor in the using process are caused, and therefore the appearance contours of the metal structural part need to be detected after primary production of the metal structural part is completed, and size precision of the metal structural part is guaranteed.
Because the production process is different and the variety of metal structure spare is various, the size is different, and traditional measuring tool is in carrying out the outward appearance profile detection in-process of metal structure spare, can appear that the work load is big, the operation is complicated, detection efficiency is low and can't satisfy the problem that the appearance detected of multiple type and multiple type product. In the prior art, a shape context histogram matrix and a local appearance description matrix are combined to obtain a total similarity measurement matrix; and abstracting the distance matrix according to the total similarity measurement matrix, removing the outline edge corresponding to the minimum shortest distance from the plurality of shortest distances, and keeping other outline edges as defect outlines. The range of the self-contained contours obtained by the shortest distance screening is not accurate enough, so that the self-contained contours corresponding to a larger distance are easily divided into defect contours, and the accuracy is lacking. Or, the contours of the measured workpiece and the prototype workpiece are calculated by a method for calculating the distance from the point to the adjacent line segment, the obtained edge contours are not further screened, and the final detection work efficiency is low.
Disclosure of Invention
In order to solve the technical problems that the acquired edge profile is not further screened, the judgment condition is single, the identification accuracy is reduced and the working efficiency is reduced in the prior art, the invention aims to provide a metal structural member detection method based on feature matching, and the adopted technical scheme is as follows:
the invention provides a metal structural member detection method based on feature matching, which comprises the following steps:
obtaining a plurality of metal surface edge images of a metal structural member to be detected and a plurality of corresponding edge points to be analyzed;
obtaining an optimized gradient direction of each pixel point in a neighborhood region according to gradient changes of the pixel points in a preset neighborhood region of the edge point to be analyzed; dividing the neighborhood region of each edge point to be analyzed into a plurality of direction regions according to a preset angle range; obtaining gradient characteristic parameters of each edge point to be analyzed according to the positions of the edge points to be analyzed and the optimized gradient directions and positions of the pixel points in the direction area; screening out contour characterization points according to the gradient characteristic parameters;
taking each contour characterization point as a circle center, and obtaining a plurality of concentric circles of the contour characterization points according to a preset radius range; dividing each concentric circle at equal angles to obtain a plurality of angle areas; for any one of the contour characterization points, obtaining feature information of the contour characterization points according to the quantity information of edge pixel points in all the angle areas, curvature features of the edge points to be analyzed in the angle areas, pixel values of the contour characterization points and pixel values of neighbor contour characterization points of the contour characterization points;
obtaining all standard edge pixel points and corresponding characteristic information of a standard metal structural member; constructing a cost function according to the characteristic information difference of the standard edge pixel points and the contour characterization points, and obtaining a matching pair of each contour characterization point and the standard edge pixel points by combining a matching algorithm; and detecting the outline of the metal structural member to be detected according to the characteristic parameter differences of the outline characterization points and the standard edge pixel points in all the matching pairs.
Further, the method for obtaining the optimized gradient direction comprises the following steps:
for any neighborhood region of the edge points to be analyzed, taking each pixel point in the neighborhood region as a reference pixel point; obtaining the gradient direction of the reference pixel point; if the gradient direction of the reference pixel point is more than or equal to 0 degrees and less than or equal to 180 degrees, the optimized gradient direction of the corresponding reference pixel point is the gradient direction; if the gradient direction of the reference pixel point is larger than 180 degrees and smaller than or equal to 360 degrees, the difference value between the gradient direction and 180 degrees is used as the optimized gradient direction of the corresponding reference pixel point.
Further, the method for acquiring the gradient characteristic parameters comprises the following steps:
obtaining gradient amplitude values of each pixel point in the direction area according to the optimized gradient direction for any one of the direction areas of the edge points to be analyzed; obtaining a Gaussian weight of each pixel point in the direction region according to the positions of each pixel point in the direction region and the corresponding edge point to be analyzed; taking the product of the gradient amplitude and the Gaussian weight as a sub-gradient amplitude of each pixel point in the direction region; accumulating the sub-gradient amplitudes of all the pixel points in the direction area to obtain a combined gradient amplitude of the direction area; taking the total gradient amplitude range of the areas corresponding to all directions of each edge point to be analyzed as the neighborhood gradient amplitude of the edge point to be analyzed;
for any edge point to be analyzed, constructing a gradient histogram by taking an angle range corresponding to a direction interval as an abscissa and the number of pixel points in the direction interval as an ordinate; obtaining a second moment of the gradient histogram; taking the product of the positive correlation mapping value of the second moment and a preset transformation parameter as the gradient direction characteristic of the edge point to be analyzed;
taking the product of the gradient direction characteristic and the neighborhood gradient amplitude as a first gradient parameter of the edge point to be analyzed; normalizing the first gradient parameters to obtain gradient characteristic parameters of the edge points to be analyzed.
Further, the screening method of the profile characterization points comprises the following steps:
and taking the edge points to be analyzed, of which the gradient characteristic parameters are larger than a preset gradient characteristic threshold value, as contour characterization points.
Further, the method for acquiring the characteristic information comprises the following steps:
the feature information of each contour characterization point comprises a sub-block number histogram, a contour histogram and a contour continuity index;
regarding any one of the contour characterization points, taking the number of edge points to be analyzed in each angle area of the contour characterization point as the edge parameters of the angle areas; constructing a sub-block number histogram of the contour characterization point according to the edge parameters of all the angle areas of the contour characterization point;
for any one of the angle areas of the contour characterization points, obtaining the curvature of each edge point to be analyzed in the angle area, and taking the curvature average value of all the edge points to be analyzed in the angle area as the curvature parameter of the angle area; constructing a contour histogram according to the curvature parameters of all the angle areas of the contour characterization points;
sequencing all the contour characterization points according to the position sequence, and taking two other contour characterization points adjacent to the contour characterization points as adjacent points corresponding to the contour characterization points; and for any one of the contour characterization points, obtaining a first difference value of pixel values of a former adjacent contour characterization point and a corresponding contour characterization point, obtaining a second difference value of the corresponding contour characterization point and a latter adjacent contour characterization point, and taking the absolute value of the difference value of the first difference value and the second difference value as a contour continuity index of the contour characterization point.
Further, the formula of the cost function includes:
in the method, in the process of the invention,representing the outline characterization point +.>Cost function of->Representing outline characterization points +.>Representing any standard edge pixel point, < >>Representing a first weight parameter,/->Representing a second weight parameter,/->Representing the outline characterization point +.>And standard edge pixel point->Chi-square distance between histograms of number of sub-blocks, < >>Representing the outline characterization point +.>And standard edge pixel point->Chi-square distance between contour histograms of (2),>representing the outline characterization point +.>And standard edge pixel point->Profile continuity indicator difference of>Representing the minimum function.
Further, the detecting the profile of the metal structural member to be detected according to the characteristic parameter differences between the profile characterizing points and the standard edge pixel points in all the matching pairs includes:
for the contour characterization points and the standard edge pixel points in any matching pair, taking the chi-square distance between the sub-block quantity histograms of the contour characterization points and the standard edge pixel points as a first characteristic value; taking the chi-square distance between the contour characterization point and the contour histogram of the standard edge pixel point as a second characteristic value; taking the contour continuity index difference value between the contour characterization point and the standard edge pixel point as a third characteristic value; obtaining the average value of the first characteristic value, the second characteristic value and the third characteristic value, and normalizing the average value to obtain the characteristic index;
if the characteristic index is larger than a preset index threshold, the profile of the corresponding position of the profile characterization point corresponding to the characteristic index in the metal structural member to be detected is abnormal; otherwise, the profile of the corresponding position is considered normal.
Further, the index threshold is set to 0.5.
Further, the neighborhood region is set to 16×16.
Further, the preset radius ranges are set to 2, 4, 6, 8 and 10.
The invention has the following beneficial effects:
in the embodiment of the invention, the gradient change of the pixel points in the preset neighborhood region of the edge points to be analyzed of the metal structural member to be detected is analyzed, the optimized gradient direction of each pixel point in the neighborhood region is obtained, and the metal structural member is accurately detected through extraction of the appearance contour characteristic information of the metal structural member. And obtaining a direction region, obtaining gradient characteristic parameters according to the positions of the edge points to be analyzed and the optimized gradient directions and positions corresponding to the pixel points in all the direction region, and screening out the contour characterization points. The profile characterization points are edge feature points which are further screened out, can characterize the appearance profile features of the metal structural part to be detected, ensure the detection precision, reduce the calculated amount in the system detection process and improve the detection speed. And obtaining the angle area of the contour characterization point, and obtaining the characteristic information of the contour characterization point according to the number of the pixel points, the curvature characteristic and the pixel value. And constructing a cost function according to the characteristic information difference of the standard edge pixel points and the profile characterization points, combining a matching algorithm to obtain a matching pair, extracting multiple characteristics of the surface profile of the metal structural member, and constructing the cost function to ensure the matching precision of the profile characterization points based on characteristic matching. By carrying out point-by-point analysis through the matching of the cost function, the false judgment of the defect contour can be avoided, and the contour of the metal structural member to be detected is further detected according to the characteristic parameter differences in all matching pairs. The embodiment of the invention can realize automatic detection of the appearance outline of the metal structural part, can reduce the detection amount of the system and ensure the detection precision of the system.
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 flowchart of a method for detecting a metal structural member based on feature matching according to an 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 is a detailed description of specific implementation, structure, characteristics and effects of the metal structural member detection method based on characteristic matching according to the invention with reference to the attached drawings and the preferred embodiment. 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 metal structural member detection method based on feature matching provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a method for detecting a metal structural member based on feature matching according to an embodiment of the present invention is shown, where the method includes:
step S1: and obtaining a plurality of metal surface images of the metal structural member to be detected and a plurality of corresponding edge points to be analyzed.
According to the embodiment of the invention, the appearance outline features of the metal structural part are extracted mainly through the surface image of the metal structural part. Firstly, a plurality of metal surface images of a metal structural member to be detected need to be obtained, and in the embodiment of the invention, an image acquisition device is arranged and is used for extracting the surface images of the metal structural member to be detected. The image acquisition equipment comprises a camera, a light source, a fixed table and other devices, and equipment deployment operators such as a specific device arrangement and a camera visual angle range are set according to actual conditions. In the embodiment of the invention, the metal structural part to be detected is placed on the fixed table, and the metal surface image of each surface of the metal structural part to be detected is obtained through the overlooking view angle of the camera and is used for comprehensively detecting the appearance outline condition of the metal structural part to be detected. In the subsequent step, a plurality of metal surface images of the metal structural member to be detected are used as the basis for feature extraction and detection of the metal structural member to be detected.
In the embodiment of the invention, all the metal surface images of the metal structural member to be detected are analyzed so as to obtain the overall appearance outline characteristics of the metal structural member to be detected, and because the processing method of each metal surface image is the same, only any one metal surface image is taken as an example for detailed description in the follow-up description. The method comprises the steps of carrying out feature extraction analysis on a metal surface image of a metal structure to be detected, firstly, using an edge detection algorithm on the metal surface image to obtain a corresponding metal surface edge image, namely obtaining edge contour information corresponding to the metal surface image of the metal structure to be detected, and carrying out subsequent analysis on the metal surface edge image corresponding to the metal surface image, wherein the edge contour information is used for detecting the appearance contour of the metal structure to be detected. It should be noted that the specific edge detection algorithm may be specifically set according to the specific embodiment, and the edge detection algorithm is a technical means well known to those skilled in the art, and will not be described herein. After the edge detection algorithm, the detected edge points are the edge points to be analyzed.
Step S2: obtaining an optimized gradient direction of each pixel point in a neighborhood region according to gradient changes of the pixel points in a preset neighborhood region of the edge point to be analyzed; dividing a neighborhood region of each edge point to be analyzed into a plurality of direction regions according to a preset angle range; obtaining gradient characteristic parameters of each edge point to be analyzed according to the positions of the edge points to be analyzed and the optimized gradient directions and positions of the pixel points in the corresponding all-direction areas; and screening out contour characterization points according to the gradient characteristic parameters.
Obtaining a plurality of edge points to be analyzed in the metal surface edge image, considering that the number of the edge points to be analyzed in the metal surface edge image is large, if feature extraction and matching are performed through all the edge points to be analyzed, the system detection amount is large, and the calculation cost is high. Therefore, for the edge points to be analyzed, the embodiment of the invention screens all the edge points to be analyzed to obtain the contour characterization points, wherein the contour characterization points are the edge points to be analyzed for characterizing the appearance contour characteristic conditions of the metal structural part, and the problem of large calculation amount in the detection process of the metal structural part can be eliminated by analyzing the contour characterization points, and meanwhile, the appearance contour detection precision of the metal structural part is ensured. Therefore, in order to obtain the edge points to be analyzed, i.e. the contour characterization points, which characterize the appearance contour feature conditions of the metal structural member, the edge contour feature of each pixel point to be analyzed needs to be analyzed, wherein the gradient direction can characterize the gray scale change direction of the pixel point, the edge feature is obvious, the situation that the gradient direction of each pixel point in the neighborhood region of the edge point to be analyzed possibly has opposite conditions is considered, and in order to further obtain the edge contour feature of the pixel point to be analyzed, the gradient direction of the pixel point in the neighborhood region of the pixel point to be analyzed needs to be corrected, i.e. the optimal gradient direction of each pixel point in the neighborhood region is obtained according to the gradient change of the pixel point in the preset neighborhood region of the edge point to be analyzed, so that the contour characterization point can be conveniently screened in the pixel point to be analyzed. In the embodiment of the present invention, the neighborhood range is 16×16, the size is the size under the relative image size measurement unit, the unit is millimeter in the present invention, and the specific neighborhood range can be specifically set according to the specific implementation manner.
Preferably, the method for obtaining the optimized gradient direction specifically comprises the following steps:
and regarding each pixel point in the neighborhood region as a reference pixel point for the neighborhood region of any edge point to be analyzed. The method for obtaining the gradient direction of the reference pixel point is a technical means known to those skilled in the art, and will not be described herein. If the gradient direction of the reference pixel point is more than or equal to 0 degrees and less than or equal to 180 degrees, the optimized gradient direction of the corresponding reference pixel point is the gradient direction; if the gradient direction of the reference pixel point is larger than 180 degrees and smaller than or equal to 360 degrees, the difference value between the gradient direction and 180 degrees is used as the optimized gradient direction of the corresponding reference pixel point. The formula for optimizing the gradient direction specifically comprises the following steps:
in the method, in the process of the invention,indicate->Optimized gradient direction of the individual reference pixels, < >>Indicate->Gradient direction of each reference pixel, +.>The circumference ratio is indicated.
In the subsequent analysis, the characteristics of each edge point to be analyzed are analyzed according to the optimized gradient direction of the pixel points in the neighborhood region of each edge point to be analyzed.
And obtaining the optimized gradient direction of each pixel point in the neighborhood region of the edge point to be analyzed, and analyzing the neighborhood region of each edge point to be analyzed. Dividing a neighborhood region of each edge point to be analyzed into a plurality of direction regions according to a preset angle range, wherein the method is implemented in the inventionIn an embodiment, the preset angle range includes, centering on each edge point to be analyzed、/>、/>And->Dividing the neighborhood region of the edge point to be analyzed into four parts according to each angle range, and sequencing the direction regions according to the angle sizes, namely, each edge point to be analyzed has four direction regions.
The invention detects and analyzes the gradient characteristics, firstly analyzes the gradient conditions of the direction interval according to the gradient characteristics of the pixel points in the direction interval consistent with the gradient direction of the corresponding edge points to be analyzed, and obtains the gradient characteristic parameters of each edge point to be analyzed according to the positions of the edge points to be analyzed and the optimized gradient directions and positions of the pixel points in all the direction areas, wherein the embodiment of the invention specifically comprises the following steps:
1. and obtaining the gradient amplitude of each pixel point in the direction region according to the optimized gradient direction in any direction region of the edge point to be analyzed. It should be noted that, the method for obtaining the gradient magnitude is a technical means well known to those skilled in the art, and will not be described herein. And obtaining a Gaussian weight value of each pixel point in the direction area according to the positions of each pixel point in the direction area and the corresponding edge point to be analyzed, and taking the product of the gradient amplitude value and the Gaussian weight value as a sub-gradient amplitude value of each pixel point in the direction area. And accumulating the sub-gradient amplitude values of all the pixel points in the direction area to obtain the combined gradient amplitude value of the direction area. The formula of the gradient combining amplitude value of each direction interval of the edge point to be analyzed specifically comprises the following steps:
in the method, in the process of the invention,the +.o representing the edge points to be analyzed>The resultant gradient magnitude of the individual directional regions; />Indicate->The number of pixels in each direction region; />The +.o representing the edge points to be analyzed>The>Gradient magnitude of each pixel point; />The +.o representing the edge points to be analyzed>The>Gaussian weight of each pixel, < +.>Representing the size parameters of the Gaussian function, in the embodiment of the invention is set +.>The specific numerical value implementer can set up by himself according to specific situations; />The +.o representing the edge points to be analyzed>The>Coordinate information of the individual pixel points; />Representing the coordinate position of the edge point to be analyzed corresponding to the direction interval; />An exponential function based on a natural constant is represented.
In the formula for combining the gradient magnitudes,representing the edge points to be analyzed and the +.>The>Position difference of each pixel point, position difference and +.>In inverse proportion, the smaller the position difference, the description of +.>The>The more accurate the result that each pixel point is used for detecting the local gradient information of the edge point to be analyzed, namely the larger the corresponding Gaussian weight value is. />The +.o representing the edge points to be analyzed>The>Sub-gradient amplitude of each pixel, +.>And->All have a direct proportion relation with the combined gradient amplitude, so the sub-gradient amplitude has a direct proportion relation with the combined gradient amplitude, and the larger the sub-gradient amplitude is, the description is +.>The>The more accurate the result that each pixel point is used for detecting the local gradient information of the edge point to be analyzed.
The resultant gradient amplitude of the direction region of the edge point to be analyzed represents the accuracy of the local gradient information result of the pixel point in the corresponding direction region for detecting the edge point to be analyzed. If the combined gradient amplitude is larger, the more accurate the result of the local gradient information of the pixel points in the corresponding direction area for detecting the edge points to be analyzed is; if the combined gradient amplitude is smaller, the more inaccurate the local gradient information result of the pixel points in the corresponding direction area for detecting the edge points to be analyzed is.
2. Further, the total gradient amplitude range of each edge point to be analyzed corresponding to all the direction areas is used as the neighborhood gradient amplitude of the edge point to be analyzed. The neighborhood gradient amplitude of the edge point to be analyzed represents the accuracy of the local gradient information result of the pixel point in the neighborhood region corresponding to the edge point to be analyzed for detecting the edge point to be analyzed. If the neighborhood gradient amplitude is larger, the more accurate the result of detecting the local gradient information of the edge point to be analyzed is for the pixel point in the neighborhood region corresponding to the edge point to be analyzed; if the neighborhood gradient amplitude is smaller, the pixel points in the neighborhood region corresponding to the edge points to be analyzed are used for detecting the local gradient information result of the edge points to be analyzed more inaccurately.
3. In order to further improve the analysis accuracy of the gradient features of the edge points to be analyzed, a gradient histogram is constructed for any one edge point to be analyzed by taking the angle range corresponding to the direction interval as an abscissa and the number of pixel points in the direction interval as an ordinate. It should be noted that, the method for obtaining the second moment of the gradient histogram is a technical means well known to those skilled in the art, and will not be described herein. And taking the product of the positive correlation mapping value of the second moment and a preset transformation parameter as the gradient direction characteristic of the edge point to be analyzed. The formula of the gradient direction characteristic of the edge point to be analyzed comprises the following steps:
in the method, in the process of the invention,representing the edge points to be analyzed->Gradient direction features of>Representing the edge points to be analyzed->Second moment of the gradient histogram of>Representing model transformation parameters for controlling the range of gradient direction features, in embodiments of the inventionThe specific numerical value implementer can set up by himself according to specific situations; />Representing natural constant->Representing the positive correlation map of the second moment. The larger the gradient direction characteristic is, the more irregular the gradient change is in the local neighborhood range corresponding to the edge pixel point to be analyzed, and the more disordered the gradient change direction is.
4. Preferably, taking the product of the gradient direction characteristic and the neighborhood gradient amplitude as a first gradient parameter of the edge point to be analyzed; normalizing the first gradient parameters to obtain gradient characteristic parameters of the edge points to be analyzed. The formula of the gradient characteristic parameter of the edge point to be analyzed comprises the following steps:
in the method, in the process of the invention,representing the edge points to be analyzed->Gradient characteristic parameter of->Representing the edge points to be analyzed->Gradient direction features of>Representing the edge points to be analyzed->Neighborhood gradient amplitude,/,>representing natural constants.
In the formula of the gradient characteristic parameter of the edge point to be analyzed, the gradient direction characteristic and the neighborhood gradient amplitude are in a direct proportion relation with the gradient characteristic parameter, and the larger the gradient direction characteristic or the larger the neighborhood gradient amplitude is, the description that the edge point to be analyzed represents the appearance outline of the metal structural memberThe greater the likelihood of the feature condition, i.e. the greater the likelihood of the corresponding edge point to be analyzed being the profile characterizing point.Proportional relation with gradient characteristic parameter by +.>Will->Normalizing to obtain gradient characteristic parameter->And the numerical range of the gradient characteristic parameters is set in 0 to 1, so that the subsequent analysis is convenient.
The gradient characteristic parameter of the edge point to be analyzed represents the possibility that the edge point to be analyzed is a contour characterization point. If the gradient characteristic parameters are larger, the probability that the corresponding edge points to be analyzed represent the appearance outline characteristic conditions of the metal structural part is larger, namely the probability that the corresponding edge points to be analyzed are outline characteristic points is larger; if the gradient characteristic parameters are smaller, the probability that the corresponding edge points to be analyzed represent the appearance outline characteristic conditions of the metal structural part is smaller, namely the probability that the corresponding edge points to be analyzed are outline characteristic points is smaller.
Further, contour characterization points are screened out according to the gradient characteristic parameters, and edge points to be analyzed, of which the gradient characteristic parameters are larger than a preset gradient characteristic threshold value, are used as the contour characterization points. In the embodiment of the invention, the preset gradient characteristic threshold value is 0.6, and a specific numerical value implementation person can set the preset gradient characteristic threshold value according to specific situations.
And (3) obtaining contour characterization points capable of characterizing the appearance contour characteristic conditions of the metal structural part through the analysis of the step S2.
Step S3: taking each contour characterization point as a circle center, and obtaining a plurality of concentric circles of the contour characterization points according to a preset radius range; performing equal-angle division on each concentric circle to obtain a plurality of angle areas; and for any contour characterization point, obtaining the characteristic information of the contour characterization point according to the quantity information of edge pixel points in all the angle areas, the curvature characteristics of the edge points to be analyzed in the angle areas, the pixel values of the contour characterization point and the pixel values of the neighborhood contour characterization points.
And (3) analyzing the characteristics of each contour characterization point based on the contour characterization points obtained in the step (S2), and obtaining a plurality of concentric circles of the contour characterization points by taking each contour characterization point as a circle center according to a preset radius range. In the embodiment of the invention, the preset radius ranges are 2, 4, 6, 8 and 10. In the embodiment of the invention, each contour characterization point corresponds to five concentric circles. Each concentric circle is divided into a plurality of angle areas by equal angles, and in the embodiment of the invention, the concentric circles are divided into 12 parts, and the specific division number can be set by oneself. Thus, one concentric circle corresponds to 12 angular regions, and since each contour characterizing point corresponds to five concentric circles, each contour characterizing point corresponds to 60 angular regions. In the subsequent process, the angle area of each contour characterization point is analyzed to obtain the local appearance contour condition of the contour characterization point. For any contour characterization point, according to the quantity information of edge pixel points corresponding to all angle areas, the curvature characteristics of the edge points to be analyzed in the angle areas, the pixel values of the contour characterization point and the pixel values of the neighborhood contour characterization points, the characteristic information of the contour characterization point is obtained, wherein the characteristic information comprises a subblock quantity histogram, a contour histogram and a contour continuity index, and the method for obtaining the characteristic information in the embodiment of the invention specifically comprises the following steps:
1. and regarding any one contour characterization point, taking the number of edge points to be analyzed in each angle area of the contour characterization point as the edge parameters of the angle areas. And constructing a sub-block quantity histogram of the contour characterization point according to the edge parameters of all the angle areas of the contour characterization point. It should be noted that, the method of constructing the histogram is a technical means well known to those skilled in the art, and will not be described herein.
2. In order to improve the contour detection accuracy, it is necessary to detect the contour change condition in each angle region. And obtaining the curvature of each edge point to be analyzed in the angle area for any angle area of the contour characterization points, taking the curvature average value of all the edge points to be analyzed in the angle area as the curvature parameter of the angle area, wherein the curvature parameter can represent the contour change condition among the pixel points in the angle area. And constructing a contour histogram according to curvature parameters of all angle areas of the contour characterization points. It should be noted that, the methods for obtaining the curvature and constructing the histogram are all technical means well known to those skilled in the art, and will not be described herein.
3. Further, the local contour continuity of the contour characterization points is analyzed. And sequencing all the contour characterization points according to the position sequence, and taking two other contour characterization points adjacent to the contour characterization points as adjacent points of the corresponding contour characterization points. And for any one contour characterization point, obtaining a first difference value of pixel values of a former adjacent contour characterization point and a corresponding contour characterization point, obtaining a second difference value of the corresponding contour characterization point and a latter adjacent contour characterization point, and taking the absolute value of the difference value of the first difference value and the second difference value as a contour continuity index of the contour characterization point. The formula of the contour continuity index of the contour characterization point specifically comprises the following steps:
in the method, in the process of the invention,representing the outline characterization point +.>Outline continuity index of>Representing the outline characterization point +.>Pixel value of the previous neighboring contour characterizing point,/-for>Representing the outline characterization point +.>Profile characterizing points +.>Pixel value of>Pixel value representing the next adjacent contour characterizing point +.>Representing the absolute value function. />A first difference value is indicated and a second difference value is indicated,representing a second difference value, the difference between the first difference value and the second difference value being capable of representing a local contour continuity of the corresponding contour characterizing point.
And (3) through the analysis of the step (S3), obtaining the characteristic information of each contour characterization point, wherein the characteristic information comprises a sub-block number histogram, a contour histogram and a contour continuity index, and providing a data base for the subsequent step.
Step S4: obtaining all standard edge pixel points and corresponding characteristic information of a standard metal structural member; constructing a cost function according to the characteristic information difference of the standard edge pixel points and the contour characterization points, and obtaining a matching pair of each contour characterization point and the standard edge pixel points by combining a matching algorithm; and detecting the profile of the metal structural member to be detected according to the characteristic parameter difference between the inner profile characterization points and the standard edge pixel points.
In order to realize analysis and detection of the appearance outline of the metal structural part to be detected, the standard metal structural part of the corresponding model of the metal structural part to be detected is obtained from a standard library of the metal structural part. It should be noted that the standard library of metal structural members includes standard metal structural members of different types, which can be selected and built by the practitioner. And taking the edge pixel points of the standard metal structural part corresponding to the metal structural part to be detected as standard edge pixel points, and forming a standard point set by all the standard edge pixel points. And the feature information of each standard edge pixel point, namely the number of sub-blocks histogram, the contour histogram and the contour continuity index is obtained by the same method in the step S3.
Further, a cost function is constructed according to the characteristic information difference of the standard edge pixel points and the profile characterizing points and is used for matching the profile characterizing points of the metal structural part to be detected. The formula of the cost function in the embodiment of the invention comprises:
in the method, in the process of the invention,representing the outline characterization point +.>Cost function of->Representing outline characterization points +.>Representing any standard edge pixel point, < >>Representing a first weight parameter,/->Representing a second weight parameter,/->Representing the outline characterization point +.>And standard edge pixel point->Chi-square distance between histograms of number of sub-blocks, < >>Representing the outline characterization point +.>And standard edge pixel point->Chi-square distance between contour histograms of (2),>representing the outline characterization point +.>And standard edge pixel point->Profile continuity indicator difference of>Representing the minimum function.
In the formulation of the cost function,characterizing points for contours->And standard edge pixel point->Chi-square distance between histograms of number of sub-blocks, < >>The smaller, the description contour characterization point +.>And standard edge pixel point->The higher the similarity of the number of pixels in the neighborhood region. />Characterizing points for contours/>And standard edge pixel point->Chi-square distance between contour histograms of (2),>the smaller, the description contour characterization point +.>And standard edge pixel point->The higher the similarity of the profile change condition in the neighborhood region. It should be noted that, the method for obtaining the chi-square distance between the histograms is a technical means well known to those skilled in the art, and will not be described herein. />Representing the outline characterization point +.>And standard edge pixel point->Profile continuity indicator difference of>The smaller, the description contour characterization point +.>And standard edge pixel pointThe more similar the contour continuity of (c). In the embodiment of the invention, the first weight parameter +.>Is 0.5, the second weight parameter +.>For 0.5, the specific values may be specifically set according to the specific embodiment.
And combining the cost function with a matching algorithm to obtain a matching pair of each contour characterization point and the standard edge pixel point. And for any contour characterization point, when the cost function value corresponding to the contour characterization point is minimum, carrying out matching analysis by a matching algorithm, and obtaining the standard edge pixel point which is matched with the contour characterization point in the best mode in all standard edge pixel points. And finally, matching all contour characterization points of the metal structural member to be detected by combining a cost function through a matching algorithm to obtain a matching pair of each contour characterization point and a standard edge pixel point, and analyzing and evaluating the appearance contour condition of the metal structural member to be detected. It should be noted that, the matching algorithm includes many hungarian algorithms, DP matching algorithms, and the like, and the practitioner may select the matching algorithm by himself, and the specific matching process is a technical means well known to those skilled in the art, and is not described herein.
Further, detecting the profile of the metal structural part to be detected according to the difference of the characteristic parameters of the inner profile characterizing points and the standard edge pixel points, wherein the method specifically comprises the following steps:
and regarding the square-clamping distance between the contour characterization point and the sub-block number histogram of the standard edge pixel point as a first characteristic value for the contour characterization point and the standard edge pixel point in any matching pair. And taking the chi-square distance between the contour characterization point and the contour histogram of the standard edge pixel point as a second characteristic value. And taking the contour continuity index difference value between the contour characterization point and the standard edge pixel point as a third characteristic value. The first characteristic value, the second characteristic value and the third characteristic value are in a direct proportion relation with the characteristic index, the average value of the first characteristic value, the second characteristic value and the third characteristic value is obtained, the average value is normalized, the characteristic index is obtained, and the purpose of using normalization is to ensure that the numerical range of the obtained characteristic index is 0-1.
If the characteristic index is larger than a preset index threshold, the profile of the corresponding position of the profile characterization point corresponding to the characteristic index in the metal structural member to be detected is abnormal, and the profile characterization point needs to be processed again to meet the subsequent use requirement; otherwise, the profile of the corresponding position is considered normal. In the embodiment of the present invention, the preset index threshold is 0.5, and the specific value can be specifically set according to the specific implementation manner.
In summary, in the embodiment of the present invention, first, the gradient change of the pixel points in the preset neighborhood region of the edge point to be analyzed of the metal structure to be detected is analyzed, and the optimized gradient direction of each pixel point in the neighborhood region is obtained. And secondly, obtaining a direction region, obtaining gradient characteristic parameters according to the position of the edge point to be analyzed and the optimized gradient direction and position of the pixel points corresponding to all the direction regions, and screening out the contour characterization points. Then, an angle area of the contour characterization point is obtained, and feature information of the contour characterization point is obtained according to the number of pixel points, the curvature feature and the pixel value. And constructing a cost function according to the characteristic information difference of the standard edge pixel points and the contour characterization points, and combining a matching algorithm to obtain a matching pair. And finally, detecting the outline of the metal structural part to be detected according to the characteristic parameter differences in all the matched pairs. The embodiment of the invention can realize automatic detection of the appearance outline of the metal structural part, can reduce the detection amount of the system and ensure the detection precision of the system.
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.
Claims (6)
1. The method for detecting the metal structural part based on the feature matching is characterized by comprising the following steps of:
obtaining a plurality of metal surface edge images of a metal structural member to be detected and a plurality of corresponding edge points to be analyzed;
obtaining an optimized gradient direction of each pixel point in a neighborhood region according to gradient changes of the pixel points in a preset neighborhood region of the edge point to be analyzed; dividing the neighborhood region of each edge point to be analyzed into a plurality of direction regions according to a preset angle range; obtaining gradient characteristic parameters of each edge point to be analyzed according to the positions of the edge points to be analyzed and the optimized gradient directions and positions of the pixel points in the direction area; screening out contour characterization points according to the gradient characteristic parameters;
taking each contour characterization point as a circle center, and obtaining a plurality of concentric circles of the contour characterization points according to a preset radius range; dividing each concentric circle at equal angles to obtain a plurality of angle areas; for any one of the contour characterization points, obtaining feature information of the contour characterization points according to the quantity information of edge pixel points in all the angle areas, curvature features of the edge points to be analyzed in the angle areas, pixel values of the contour characterization points and pixel values of neighbor contour characterization points of the contour characterization points;
obtaining all standard edge pixel points and corresponding characteristic information of a standard metal structural member; constructing a cost function according to the characteristic information difference of the standard edge pixel points and the contour characterization points, and obtaining a matching pair of each contour characterization point and the standard edge pixel points by combining a matching algorithm; detecting the outline of the metal structural member to be detected according to the characteristic parameter differences of the outline characterization points and the standard edge pixel points in all the matching pairs;
the method for acquiring the optimized gradient direction comprises the following steps:
for any neighborhood region of the edge points to be analyzed, taking each pixel point in the neighborhood region as a reference pixel point; obtaining the gradient direction of the reference pixel point; if the gradient direction of the reference pixel point is more than or equal to 0 degrees and less than or equal to 180 degrees, the optimized gradient direction of the corresponding reference pixel point is the gradient direction; if the gradient direction of the reference pixel point is larger than 180 degrees and smaller than or equal to 360 degrees, the difference value between the gradient direction and 180 degrees is used as the optimized gradient direction of the corresponding reference pixel point;
the gradient characteristic parameter acquisition method comprises the following steps:
obtaining gradient amplitude values of each pixel point in the direction area according to the optimized gradient direction for any one of the direction areas of the edge points to be analyzed; obtaining a Gaussian weight of each pixel point in the direction region according to the positions of each pixel point in the direction region and the corresponding edge point to be analyzed; taking the product of the gradient amplitude and the Gaussian weight as a sub-gradient amplitude of each pixel point in the direction region; accumulating the sub-gradient amplitudes of all the pixel points in the direction area to obtain a combined gradient amplitude of the direction area; taking the total gradient amplitude range of the areas corresponding to all directions of each edge point to be analyzed as the neighborhood gradient amplitude of the edge point to be analyzed;
for any edge point to be analyzed, constructing a gradient histogram by taking an angle range corresponding to a direction interval as an abscissa and the number of pixel points in the direction interval as an ordinate; obtaining a second moment of the gradient histogram; taking the product of the positive correlation mapping value of the second moment and a preset transformation parameter as the gradient direction characteristic of the edge point to be analyzed;
taking the product of the gradient direction characteristic and the neighborhood gradient amplitude as a first gradient parameter of the edge point to be analyzed; normalizing the first gradient parameters to obtain gradient characteristic parameters of the edge points to be analyzed;
the screening method of the profile characterization points comprises the following steps:
taking the edge points to be analyzed, of which the gradient characteristic parameters are larger than a preset gradient characteristic threshold value, as contour characterization points;
the method for acquiring the characteristic information comprises the following steps:
the feature information of each contour characterization point comprises a sub-block number histogram, a contour histogram and a contour continuity index;
regarding any one of the contour characterization points, taking the number of edge points to be analyzed in each angle area of the contour characterization point as the edge parameters of the angle areas; constructing a sub-block number histogram of the contour characterization point according to the edge parameters of all the angle areas of the contour characterization point;
for any one of the angle areas of the contour characterization points, obtaining the curvature of each edge point to be analyzed in the angle area, and taking the curvature average value of all the edge points to be analyzed in the angle area as the curvature parameter of the angle area; constructing a contour histogram according to the curvature parameters of all the angle areas of the contour characterization points;
sequencing all the contour characterization points according to the position sequence, and taking two other contour characterization points adjacent to the contour characterization points as adjacent points corresponding to the contour characterization points; and for any one of the contour characterization points, obtaining a first difference value of pixel values of a former adjacent contour characterization point and a corresponding contour characterization point, obtaining a second difference value of the corresponding contour characterization point and a latter adjacent contour characterization point, and taking the absolute value of the difference value of the first difference value and the second difference value as a contour continuity index of the contour characterization point.
2. The feature matching-based metal structure detection method according to claim 1, wherein the formula of the cost function includes:
in the method, in the process of the invention,representing the outline characterization point +.>Cost function of->Representing outline characterization points +.>Representing any standard edge pixel point, < >>Representing a first weight parameter,/->Representing a second weight parameter,/->Representing the outline characterization point +.>And standard edge pixel point->Chi-square distance between histograms of number of sub-blocks, < >>Representing the outline characterization point +.>And standard edge pixel point->Chi-square distance between contour histograms of (2),>representing the outline characterization point +.>And standard edge pixel pointProfile continuity indicator difference of>Representing the minimum function.
3. The method for detecting a metal structural member based on feature matching according to claim 1, wherein detecting the profile of the metal structural member to be detected according to the feature parameter differences between the profile characterizing points and the standard edge pixel points in all the matching pairs comprises:
for the contour characterization points and the standard edge pixel points in any matching pair, taking the chi-square distance between the sub-block quantity histograms of the contour characterization points and the standard edge pixel points as a first characteristic value; taking the chi-square distance between the contour characterization point and the contour histogram of the standard edge pixel point as a second characteristic value; taking the contour continuity index difference value between the contour characterization point and the standard edge pixel point as a third characteristic value; obtaining the average value of the first characteristic value, the second characteristic value and the third characteristic value, and normalizing the average value to obtain a characteristic index;
if the characteristic index is larger than a preset index threshold, the profile of the corresponding position of the profile characterization point corresponding to the characteristic index in the metal structural member to be detected is abnormal; otherwise, the profile of the corresponding position is considered normal.
4. A method for detecting a metal structural member based on feature matching according to claim 3, wherein the index threshold is set to 0.5.
5. The feature matching-based metal structure detection method according to claim 1, wherein the neighborhood region is set to 16×16.
6. The feature matching-based metal structure detection method according to claim 1, wherein the preset radius ranges are set to 2, 4, 6, 8 and 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311027608.9A CN116758067B (en) | 2023-08-16 | 2023-08-16 | Metal structural member detection method based on feature matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311027608.9A CN116758067B (en) | 2023-08-16 | 2023-08-16 | Metal structural member detection method based on feature matching |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116758067A CN116758067A (en) | 2023-09-15 |
| CN116758067B true CN116758067B (en) | 2023-12-01 |
Family
ID=87961223
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311027608.9A Active CN116758067B (en) | 2023-08-16 | 2023-08-16 | Metal structural member detection method based on feature matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116758067B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117197129B (en) * | 2023-11-03 | 2024-02-13 | 浙江鑫柔科技有限公司 | Blackening degree detection method and device and computer equipment |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1491879A2 (en) * | 2003-06-10 | 2004-12-29 | hema electronic GmbH | Method for the adaptive recognition of a defect on a structured surface |
| JP2009217799A (en) * | 2008-02-14 | 2009-09-24 | Seiko Epson Corp | Contour detection method, contour detection device, and contour detection program |
| EP2109079A1 (en) * | 2008-04-09 | 2009-10-14 | Peugeot Citroën Automobiles Société Anonyme | Method and system of characterisation of a surface defect of a workpiece |
| JP2011196982A (en) * | 2010-03-17 | 2011-10-06 | Cognex Kk | Defect detection method, defect detection device, and program |
| CN105678733A (en) * | 2014-11-21 | 2016-06-15 | 中国科学院沈阳自动化研究所 | Infrared and visible-light different-source image matching method based on context of line segments |
| CN110717896A (en) * | 2019-09-24 | 2020-01-21 | 东北大学 | Plate strip steel surface defect detection method based on saliency label information propagation model |
| CN111179230A (en) * | 2019-12-18 | 2020-05-19 | 星际空间（天津）科技发展有限公司 | Remote sensing image contrast change detection method and device, storage medium and electronic equipment |
| CN115272336A (en) * | 2022-09-29 | 2022-11-01 | 加力卡塞阀门(江苏)有限公司 | Metal part defect accurate detection method based on gradient vector |
| CN115311275A (en) * | 2022-10-11 | 2022-11-08 | 南通百仕灵新能源科技有限公司 | Method and system for detecting surface defects of electronic component |
| CN115330791A (en) * | 2022-10-13 | 2022-11-11 | 江苏东晨机械科技有限公司 | Part burr detection method |
| CN116091790A (en) * | 2023-03-21 | 2023-05-09 | 苏州奥沃汽车配件有限公司 | Online detection system for quality of driven disc of electrically-controlled silicone oil clutch |
| CN116228747A (en) * | 2023-05-04 | 2023-06-06 | 青岛穗禾信达金属制品有限公司 | Metal cabinet processing quality monitoring method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8917940B2 (en) * | 2013-04-26 | 2014-12-23 | Mitutoyo Corporation | Edge measurement video tool with robust edge discrimination margin |
| WO2015132817A1 (en) * | 2014-03-05 | 2015-09-11 | 三菱電機株式会社 | Edge detection device, edge detection method, and program |
| CN106950283B (en) * | 2017-02-14 | 2019-07-26 | 清华大学 | Magnetosonic array structure method of adjustment and device is imaged in metal plate defect profile guided wave |
-
2023
- 2023-08-16 CN CN202311027608.9A patent/CN116758067B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1491879A2 (en) * | 2003-06-10 | 2004-12-29 | hema electronic GmbH | Method for the adaptive recognition of a defect on a structured surface |
| JP2009217799A (en) * | 2008-02-14 | 2009-09-24 | Seiko Epson Corp | Contour detection method, contour detection device, and contour detection program |
| EP2109079A1 (en) * | 2008-04-09 | 2009-10-14 | Peugeot Citroën Automobiles Société Anonyme | Method and system of characterisation of a surface defect of a workpiece |
| JP2011196982A (en) * | 2010-03-17 | 2011-10-06 | Cognex Kk | Defect detection method, defect detection device, and program |
| CN105678733A (en) * | 2014-11-21 | 2016-06-15 | 中国科学院沈阳自动化研究所 | Infrared and visible-light different-source image matching method based on context of line segments |
| CN110717896A (en) * | 2019-09-24 | 2020-01-21 | 东北大学 | Plate strip steel surface defect detection method based on saliency label information propagation model |
| CN111179230A (en) * | 2019-12-18 | 2020-05-19 | 星际空间（天津）科技发展有限公司 | Remote sensing image contrast change detection method and device, storage medium and electronic equipment |
| CN115272336A (en) * | 2022-09-29 | 2022-11-01 | 加力卡塞阀门(江苏)有限公司 | Metal part defect accurate detection method based on gradient vector |
| CN115311275A (en) * | 2022-10-11 | 2022-11-08 | 南通百仕灵新能源科技有限公司 | Method and system for detecting surface defects of electronic component |
| CN115330791A (en) * | 2022-10-13 | 2022-11-11 | 江苏东晨机械科技有限公司 | Part burr detection method |
| CN116091790A (en) * | 2023-03-21 | 2023-05-09 | 苏州奥沃汽车配件有限公司 | Online detection system for quality of driven disc of electrically-controlled silicone oil clutch |
| CN116228747A (en) * | 2023-05-04 | 2023-06-06 | 青岛穗禾信达金属制品有限公司 | Metal cabinet processing quality monitoring method |
Non-Patent Citations (3)
| Title |
|---|
| Research on edge detection algorithm of work piece defect in machine vision detection system;Shan Luo等;2022 IEEE 6th Information Technology and Mechatronics Engineering Conference (ITOEC);全文 * |
| 基于图像配准的零件轮廓修正方法;吴孟桦;胡晓兵;李航;江代渝;;计算机应用(第04期);全文 * |
| 高速铁路扣件缺陷三维检测系统集成与识别算法;刘东旭;中国优秀硕士学位论文全文数据库 信息科技辑;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116758067A (en) | 2023-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115294113B (en) | Quality detection method for wood veneer | |
| CN108460327B (en) | Automatic identification method for reading of pointer instrument based on image processing | |
| CN111862037A (en) | Method and system for detecting geometric characteristics of precision hole type part based on machine vision | |
| US10062172B2 (en) | Automated tattoo recognition techniques | |
| CN108918526B (en) | Notch defect detection method for flexible IC packaging substrate circuit | |
| CN109900711A (en) | Workpiece, defect detection method based on machine vision | |
| CN105335973B (en) | Apply to the visual processing method of strip machining production line | |
| CN116758067B (en) | Metal structural member detection method based on feature matching | |
| CN109115800B (en) | Method for rapidly detecting burrs of product and accurately measuring length | |
| CN109658391B (en) | Circle radius measuring method based on contour merging and convex hull fitting | |
| CN115830015B (en) | Hardware stamping accessory quality detection method based on computer vision | |
| CN115187548A (en) | Mechanical part defect detection method based on artificial intelligence | |
| CN115984271B (en) | Metal burr identification method based on corner detection | |
| CN111539938B (en) | Method, system, medium and electronic terminal for detecting curvature of rolled strip steel strip head | |
| CN113012098B (en) | Iron tower angle steel punching defect detection method based on BP neural network | |
| CN111539927B (en) | Detection method of automobile plastic assembly fastening buckle missing detection device | |
| CN114926410A (en) | Method for detecting appearance defects of brake disc | |
| CN115311629B (en) | Abnormal bending precision monitoring system of bending machine | |
| CN115937203B (en) | Visual detection method, device, equipment and medium based on template matching | |
| CN115115603A (en) | Automobile accessory flywheel surface detection method based on artificial intelligence | |
| CN111815575B (en) | Bearing steel ball part detection method based on machine vision | |
| CN115511884B (en) | Punching compound die surface quality detection method based on computer vision | |
| CN115753791B (en) | Defect detection method, device and system based on machine vision | |
| CN116188458A (en) | Intelligent recognition method for abnormal deformation of surface of die-casting die of automobile part | |
| JP6397292B2 (en) | Item identification and image discrimination method by image processing |
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 |