CN113538433A - Mechanical casting defect detection method and system based on artificial intelligence - Google Patents

Mechanical casting defect detection method and system based on artificial intelligence Download PDF

Info

Publication number
CN113538433A
CN113538433A CN202111089875.XA CN202111089875A CN113538433A CN 113538433 A CN113538433 A CN 113538433A CN 202111089875 A CN202111089875 A CN 202111089875A CN 113538433 A CN113538433 A CN 113538433A
Authority
CN
China
Prior art keywords
crack defect
image
suspected
defect area
suspected crack
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202111089875.XA
Other languages
Chinese (zh)
Other versions
CN113538433B (en
Inventor
徐玲雅
戴凤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NANTONGYOUYUAN ART DESIGN Co.,Ltd.
Original Assignee
Haimen Chuangrui Machinery Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Haimen Chuangrui Machinery Co ltd filed Critical Haimen Chuangrui Machinery Co ltd
Priority to CN202111089875.XA priority Critical patent/CN113538433B/en
Publication of CN113538433A publication Critical patent/CN113538433A/en
Application granted granted Critical
Publication of CN113538433B publication Critical patent/CN113538433B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-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
    • G06T7/12—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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/30116—Casting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to the technical field of mechanical part detection, in particular to a method and a system for detecting defects of a mechanical casting based on artificial intelligence. The method comprises the following steps: acquiring at least two mechanical casting initial images; carrying out image segmentation on the image to obtain an image of a suspected crack defect area; acquiring the principal component direction of the suspected crack defect area image, and determining different convolution kernel sizes and sliding step lengths of the suspected crack defect area images based on different principal component directions; inputting each suspected crack defect area image into a convolutional neural network for image feature extraction, and finally outputting the probability that each pixel point in the image belongs to a crack defect; and detecting the defects of the mechanical casting based on the trained convolutional neural network. The speed of network training is greatly improved, the detection accuracy is improved while the detail of the characteristic comparison in the network training process is improved, and the defect identification efficiency is improved.

Description

Mechanical casting defect detection method and system based on artificial intelligence
Technical Field
The invention relates to the technical field of mechanical part detection, in particular to a method and a system for detecting defects of a mechanical casting based on artificial intelligence.
Background
The manufacturing industry has very strict requirements on the process quality of mechanical castings, a plurality of methods for detecting the defects of the mechanical castings are provided, most of the detection of the general castings still stay in a manual detection mode, the manual naked eye detection consumes manpower and material resources, and the problems of low detection efficiency and low accuracy exist.
The method is characterized in that the defect of the mechanical casting is detected on line by training a neural network, but the recognition process of the neural network is to recognize the whole image, the whole training process is too slow, network parameters are more, and the problem of low efficiency still exists although the detection accuracy is improved.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a mechanical casting defect detection method and system based on artificial intelligence, and the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides an artificial intelligence-based method for detecting defects of a mechanical casting, the method including the following steps:
acquiring at least two mechanical casting initial images;
carrying out image segmentation on the initial image of the mechanical casting to obtain an image of a suspected crack defect area;
performing principal component analysis on the suspected crack defect area image to obtain a principal component direction of the suspected crack defect area image, and determining different convolution kernel sizes and sliding step lengths of the suspected crack defect area image based on different principal component directions of the suspected crack defect area image;
inputting each suspected crack defect area image into a convolutional neural network, extracting image characteristics based on the corresponding convolutional kernel size and sliding step length, and finally outputting the probability that each pixel point in the suspected crack defect area image belongs to a crack defect;
detecting the defects of the mechanical casting based on the trained convolutional neural network;
the step of determining different convolution kernel sizes and sliding step lengths of the images of the suspected crack defect areas based on different principal component directions of the images of the suspected crack defect areas comprises the following steps:
setting three angle ranges, wherein the target directions corresponding to the three angle ranges are respectively a horizontal direction, a vertical direction and a target inclination direction, and the angle corresponding to the target inclination direction is 45oOr 135oDirection;
acquiring an angle range of the main component direction of each suspected crack defect area image based on the different main component directions of each suspected crack defect area image, and determining a target direction corresponding to each suspected crack defect area image according to the angle range;
determining the convolution kernel size and the sliding step length of each suspected crack defect area image according to the target direction corresponding to each suspected crack defect area image, wherein when the principal component direction is the horizontal direction, the convolution kernel size is
Figure DEST_PATH_IMAGE001
The sliding step length in the horizontal direction is 1, and the sliding step length in the vertical direction is 2; when the principal component direction is the vertical direction, the convolution kernel size is
Figure 802907DEST_PATH_IMAGE002
The vertical sliding step length is 1, and the horizontal sliding step length is 2; when the direction of the main component is 45oOr 135oIn direction, the convolution kernel size is
Figure DEST_PATH_IMAGE003
The horizontal direction sliding step length is 1, and the vertical direction sliding step length is 1.
Preferably, the step of performing image segmentation on the initial image of the mechanical casting to obtain an image of a suspected crack defect region includes:
acquiring the gray level of each pixel point in the initial image, counting the gray level of each pixel point, and forming a gray distribution curve by taking the gray level as an abscissa and taking the number of gray statistics as an ordinate;
selecting a gray number threshold straight line to divide the gray distribution curve into an upper part and a lower part, obtaining gray levels corresponding to the statistical times of each gray level below the threshold straight line which can be read, taking the positions of pixel points corresponding to the obtained gray levels as suspected crack defect points, and obtaining images of the suspected crack defect areas according to the suspected crack defect points.
Preferably, the step of performing principal component analysis on the suspected crack defect region image to obtain a principal component direction of the suspected crack defect region image includes:
performing principal component analysis on each pixel point of the suspected crack defect area to obtain a characteristic value of each pixel point;
sorting the characteristic values of each pixel point from big to small, acquiring characteristic values of a preset number, and further acquiring characteristic vectors corresponding to the characteristic values of the preset number to obtain the characteristic vectors of the suspected crack defect area;
and superposing the characteristic vectors to obtain a superposed vector direction, wherein the superposed vector direction is used as a main component direction of the suspected crack defect area image.
Preferably, before the step of detecting the defect of the mechanical casting based on the trained convolutional neural network, the method further comprises the following steps:
based on the probability that each pixel point in the suspected crack defect area image belongs to a crack defect, acquiring pixel points of which the probability of the crack defect is greater than or equal to a preset probability threshold from pixel points belonging to a normal area to obtain initial pixel points, wherein the preset probability threshold is less than 0.5, and the error value between the preset probability threshold and 0.5 is less than a preset error value;
constructing a contrast loss function based on the probability that the initial pixel point belongs to the crack defect and the probability that the pixel point adjacent to the initial pixel point in the suspected crack defect area image belongs to the crack defect;
acquiring the saturation of the initial pixel point and the saturation mean value of the defect area in the suspected crack defect area image, and further constructing a supervision network loss function;
constructing a target loss function according to the contrast loss function and the supervision network loss function;
training a defect growth prediction network according to the target loss function;
correspondingly, after the step of detecting the defect of the mechanical casting based on the trained convolutional neural network, the method further comprises the following steps:
and based on the trained defect growth prediction network, performing defect growth prediction on the defect region of the mechanical casting.
Preferably, the contrast loss function is specifically:
Figure DEST_PATH_IMAGE005
wherein the content of the first and second substances,
Figure 640545DEST_PATH_IMAGE006
representing the crack defect probability of the defect image edge pixel points;
Figure DEST_PATH_IMAGE007
and representing the crack defect probability of the initial pixel points of the edges of the adjacent defect images.
Preferably, the supervision network loss function is specifically:
Figure DEST_PATH_IMAGE009
wherein the content of the first and second substances,
Figure 655862DEST_PATH_IMAGE010
representing the saturation of the adjacent initial pixel points of the defect image;
Figure DEST_PATH_IMAGE011
representing the mean saturation of the defective area.
Preferably, the target loss function is specifically:
Figure DEST_PATH_IMAGE013
wherein the content of the first and second substances,
Figure 172425DEST_PATH_IMAGE006
representing the probability of crack defects of pixel points adjacent to the initial pixel points in the suspected crack defect area image;
Figure 124201DEST_PATH_IMAGE007
representing the crack defect probability of the initial pixel point;
Figure 754027DEST_PATH_IMAGE010
representing the saturation of the initial pixel point;
Figure 398635DEST_PATH_IMAGE011
and representing the mean saturation value of the defect area in the suspected crack defect area image.
In a second aspect, another embodiment of the present invention provides an artificial intelligence based mechanical casting defect detection system, comprising: memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the steps of the above method are implemented when the processor executes the computer program.
The invention has the following beneficial effects: the method comprises the steps of obtaining a suspected crack defect area image by carrying out image segmentation on an initial image of a mechanical casting, then carrying out principal component analysis on the suspected crack defect area image to obtain different principal component directions, formulating different convolution kernel sizes and sliding step lengths according to the obtained different principal component directions, being capable of better extracting the contrast relation among defect characteristics in the principal component directions, accurately distinguishing defect characteristic details, improving the accuracy of convolution kernel extraction on characteristics beneficial to defect identification, effectively improving the convolution speed of the image by adopting convolution kernels with various sizes and sliding step lengths, extracting the defect characteristics in the image more quickly and completing defect detection more accurately; inputting each suspected crack defect area image into a convolutional neural network, and performing feature extraction on the image based on the corresponding convolutional kernel size so as to train the convolutional neural network; and finally, detecting the defects of the mechanical casting based on the trained convolutional neural network, and greatly improving the speed of network training due to less parameters of the network, improving the detection accuracy and the network training efficiency and further improving the efficiency of defect identification.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a mechanical casting defect detection method based on artificial intelligence according to an embodiment of the present invention.
Detailed Description
In order to further explain the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description, the structure, the features and the effects of the method and the system for detecting defects of mechanical castings based on artificial intelligence according to the present invention are provided with the accompanying drawings and the preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
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 describes a specific scheme of a mechanical casting defect detection method and system based on artificial intelligence in detail with reference to the accompanying drawings.
Referring to fig. 1, a flow chart of a method for detecting defects of a mechanical casting based on artificial intelligence according to an embodiment of the present invention is shown, wherein the method comprises the following steps:
step S100, at least two mechanical casting initial images are obtained.
Specifically, a camera is installed above the mechanical casting quality inspection detection platform and used for collecting images, in order to completely collect a complete image of the surface of the casting, the camera shoots a high overlooking visual angle, and the collected image is used as a first image.
It should be noted that, in the process of image acquisition, the position of the camera is fixed, and the position of the acquired mechanical casting is fixed, so that the size of the first image acquired each time is the same, and the number of pixel points included in the image is the same.
In order to make the acquired first image clearer, in the embodiment of the invention, the first image is subjected to noise filtering processing by adopting Gaussian denoising, and a Gaussian kernel adopts a template with the size of 3 × 3, so that an initial image of a mechanical casting is finally obtained. The obtained initial images of the mechanical casting are used for subsequent network training, the number of the initial images of the mechanical casting is at least two, and the specific number of the initial images of the mechanical casting is set by an implementer according to the actual situation.
And S200, carrying out image segmentation on the initial image of the mechanical casting to obtain an image of a suspected crack defect area.
The range of the initial image of the mechanical casting is large, and besides a part of defect areas, more normal areas exist, so that the defect areas are preliminarily divided and identified, the defect areas are processed in a targeted manner, and the subsequent calculation amount is reduced. The image segmentation can be performed on the initial image of the mechanical casting by adopting the existing image segmentation algorithm, and the image segmentation process in the embodiment of the invention specifically comprises the following steps:
acquiring a gray distribution curve of pixel points in the initial image, wherein in the gray distribution curve, the abscissa represents gray level, and the ordinate represents gray statistics times; and selecting a gray frequency threshold straight line to divide a gray distribution curve into an upper part and a lower part, taking the position of a pixel point corresponding to a gray image below the gray frequency threshold straight line as a suspected crack defect point, and acquiring an image of the suspected crack defect area according to the suspected crack defect point.
Specifically, the mechanical casting initial image obtained in step S100 is subjected to graying processing to obtain a grayscale image, a grayscale value of each pixel point in the image is obtained, the grayscale value of each pixel point in the grayscale image is counted, and a grayscale distribution curve is finally obtained according to the number of times each grayscale level appears. And removing the background and the grid of the obtained gray distribution curve to obtain a gray distribution curve graph of each pixel point in the initial image, wherein the abscissa of the gray distribution curve is gray level, and the ordinate is gray statistics times.
Then, a gray number threshold straight line is selected to segment the obtained gray distribution curve, the gray number threshold straight line can be understood as a horizontal plane straight line in a watershed algorithm, the gray distribution curve is divided into an upper part and a lower part through the gray number threshold straight line, and pixel points which are below the horizontal plane straight line and represent less gray levels and have less occurrence times are called low-frequency pixel points; the pixels above the horizontal line that represent more occurrences of gray scale are called high-frequency pixels.
For a normal initial image of a mechanical casting, the gray value of the surface of the image is relatively smooth, the gray value difference is relatively small, when the mechanical casting has defects such as cracks, pixel points with large texture information and gray value changes can appear, but the number of the pixel points is relatively larger than that of the whole initial image, so that the low-frequency pixel points with small number in a gray distribution curve are defect pixel points, the low-frequency pixel points can be segmented from the initial image of the mechanical casting through a gray number threshold straight line to obtain a plurality of suspected crack defect pixel points, and further to obtain suspected crack defect areas, the number of the suspected crack defect areas is indefinite, only one suspected crack defect area or a plurality of suspected crack defect areas are possible, and the irregular edges of the suspected crack defect areas are supplemented into regular edges for facilitating subsequent network training and convolution kernel feature extraction, the method includes the steps that a part of non-defect areas are selected as suspected defect areas to fill edges, multiple images of the suspected crack defect areas in the rectangular regular shape are finally obtained, edge filling can be conducted by adopting an existing edge filling algorithm, and the minimum circumscribed rectangle of the suspected crack defect areas can be obtained by adopting a minimum circumscribed rectangle algorithm.
And step S300, performing principal component analysis on the suspected crack defect area image to obtain the principal component direction of the suspected crack defect area image, and determining different convolution kernel sizes and sliding step lengths of the suspected crack defect area image based on different principal component directions of the suspected crack defect area image.
Performing principal component analysis on each pixel point of the suspected crack defect area to obtain a characteristic value of each pixel point; sorting the characteristic values of each pixel point from big to small, acquiring the characteristic values of the front preset number, and further acquiring the characteristic vectors corresponding to the characteristic values of the front preset number to obtain the characteristic vectors of the suspected crack defect area; and superposing the characteristic vectors to obtain a superposed vector direction, wherein the superposed vector direction is used as a main component direction of the suspected crack defect area image.
Since the processing procedure for each suspected crack defect area image is the same, a detailed description will be given by taking any one suspected crack defect image as an example. Specifically, principal component analysis is performed on each pixel point of the suspected crack defect region image obtained in step S200 to obtain a characteristic value corresponding to each pixel point, and Principal Component Analysis (PCA) is a known algorithm and is not described in detail here. And sequencing all the obtained characteristic values from large to small. In the embodiment of the present invention, the first 10 eigenvalues are selected, then the 10 eigenvalues are the largest 10 eigenvalues of all eigenvalues, each eigenvalue corresponds to one eigenvector, then 10 eigenvectors are obtained correspondingly, 10 eigenvectors corresponding to the eigenvectors are extracted, then 10 eigenvectors are superimposed to obtain a new vector and a direction thereof, the direction of the superimposed vector is used as the principal component direction of the suspected crack defect area image, and the principal component direction is the principal grayscale direction of the suspected crack defect area image.
And setting the sizes and the sliding step lengths of different convolution kernels in the convolution neural network according to the main component direction of the suspected crack defect area image.
Specifically, three angle ranges are set, the target directions corresponding to the three angle ranges are respectively the horizontal direction, the vertical direction and the target inclination direction, and the angle corresponding to the target inclination direction is 45 degreesoOr 135oAnd (4) direction.
It should be noted that, in the embodiment of the present invention, three angular ranges are formed
Figure 120604DEST_PATH_IMAGE014
Angular ranges, and there is no overlapping angular range or value between the three angular ranges, nor is there any value for the angle
Figure 236589DEST_PATH_IMAGE014
The angular range or value of the hollow, for example: the angle range corresponding to the horizontal direction is
Figure DEST_PATH_IMAGE015
And
Figure 980642DEST_PATH_IMAGE016
the target inclination direction corresponds to an angle range of
Figure DEST_PATH_IMAGE017
And
Figure 186365DEST_PATH_IMAGE018
the angle range corresponding to the vertical direction is
Figure DEST_PATH_IMAGE019
。
Based on different principal component directions of the images of the suspected crack defect regions, acquiring an angle range of the principal component direction of each image of the suspected crack defect regions, and determining a target direction corresponding to each image of the suspected crack defect regions according to the angle range, wherein when the angle of the principal component direction is 53 degrees, the angle range is
Figure 818466DEST_PATH_IMAGE017
Then the target direction may be determined to be the target pitch direction.
Determining the convolution kernel size and the sliding step length of each suspected crack defect area image according to the target direction corresponding to each suspected crack defect area image; the convolution kernel with multiple sizes can improve the convolution speed of the image, quickly extract defect characteristics in the image and more accurately finish defect detection and defect positioning.
Specifically, when the principal component direction is the horizontal direction, the convolution kernel size is
Figure 643202DEST_PATH_IMAGE001
The horizontal direction sliding step is 1, and the vertical direction sliding step is 2.
It should be noted that the convolution kernel size is set to
Figure 122987DEST_PATH_IMAGE001
The method is used for better extracting the contrast relation among defect features in the horizontal direction, distinguishing the details of the defect features and improving the capability of extracting the features which are beneficial to identifying the defects by a convolution kernel; the sliding step length in the horizontal direction is set to be 1 so that the detailed feature comparison in the image principal component direction can be obtained through each sliding, a more accurate convolution feature graph is obtained, the sliding step length in the vertical direction is set to be 2, the feature extraction efficiency can be improved, the convolution speed in the non-principal component direction is improved, the detailed feature extraction between pixel points in the non-principal component direction is omitted, and the network is enabled to be rapidly converged.
When the principal component direction is the vertical direction, in order to better extract the contrast relation between defect characteristics in the vertical direction, the size of the convolution kernel is set to be
Figure 234032DEST_PATH_IMAGE002
The vertical direction sliding step is 1, and the horizontal direction sliding step is 2.
It should be noted that the convolution kernel is set to
Figure 930592DEST_PATH_IMAGE002
The defect feature details in the vertical direction can be better extracted, the details in the image principal component direction can be obtained at each sliding with the sliding step size of 1 in the vertical direction, and the sliding step size of 2 in the horizontal direction is used for improving the convolution speed in the non-principal component direction.
When the direction of the main component is 45oOr 135oIn the direction of the defect, the defect direction is in the diagonal direction relative to the image, so in order to extract complete defect features, the feature extraction needs to be completed by using the size of a regular convolution kernel, so that the interior of the convolution kernel can contain more defect images each time, the contrast relation between the defect features is better extracted, and the size of the convolution kernel is set to be equal to the size of the defect image
Figure 794905DEST_PATH_IMAGE003
The horizontal direction sliding step is 1, and the vertical direction sliding step is 1.
And S400, inputting each suspected crack defect area image into a convolutional neural network, extracting image characteristics based on the corresponding convolutional kernel size and sliding step length, and finally outputting the probability that each pixel point in the suspected crack defect area image belongs to a crack defect.
Specifically, the training process of the convolutional neural network is as follows:
(1) the input of the convolutional neural network is a plurality of acquired images of suspected crack defect areas, and the structure of the convolutional neural network is an encoder-full connection network layer;
(2) the loss function in the training of the convolutional neural network is a cross entropy loss function;
(3) and for any suspected crack defect area image, extracting the characteristics of the encoder according to the determined convolution kernel size and sliding step length of the suspected crack defect area image, and outputting the characteristic image as a characteristic image of the suspected crack defect area image.
And passing the obtained characteristic diagram of the suspected crack defect area image through a full-connection network layer, outputting the crack defect probability P of each pixel point through a SoftMax activation function, and obtaining real crack defect pixel points in all the pixel points of the suspected crack defect area image according to the crack defect probability of each pixel point, wherein the real crack defect pixel points form a defect area.
It should be noted that, in the embodiment of the present invention, when the defect probability P is greater than or equal to 0.5, the defect point is set; accordingly, less than 0.5 is a normal point.
In the defect detection method, the point with the crack defect probability P being more than or equal to 0.5 is used as the real crack defect pixel point, so that when the crack defect probability P is 0.4, the crack defect pixel point is possibly a defect pixel point or a pixel point with a defect possibly having a growth trend, and therefore the region where the defect of the mechanical casting grows is predicted in the embodiment of the invention.
Based on the probability that each pixel point in the suspected crack defect area image belongs to a crack defect, acquiring pixel points of which the probability of the crack defect is greater than or equal to a preset probability threshold from pixel points belonging to a normal area to obtain initial pixel points, wherein the preset probability threshold is less than 0.5, and the error value between the preset probability threshold and 0.5 is less than a preset error value; therefore, the predetermined probability threshold is a probability value smaller than 0.5 and having a small difference from 0.5, for example, 0.4, because in this range, the pixels in the normal region may be misjudged as normal pixels due to detection errors and the like, and the nature of the predetermined probability threshold may be defective pixels or pixels with defects possibly having an increasing trend.
Constructing a contrast loss function based on the probability that the initial pixel point belongs to the crack defect and the probability that the pixel point adjacent to the initial pixel point in the suspected crack defect area image belongs to the crack defect; acquiring the saturation of the initial pixel point and the saturation mean value of the defect area in the suspected crack defect area image, and further constructing a supervision network loss function; constructing a target loss function according to the contrast loss function and the supervision network loss function;
firstly, based on the probability that the initial pixel point belongs to the crack defect and the probability that the pixel point adjacent to the initial pixel point in the suspected crack defect area image belongs to the crack defect, a contrast loss function is constructed, and the prediction of the pixel point in the defect growth direction is carried out, wherein the contrast loss function specifically comprises the following steps:
Figure 893311DEST_PATH_IMAGE005
wherein the content of the first and second substances,
Figure 175257DEST_PATH_IMAGE006
representing the crack defect probability of pixel points adjacent to the initial pixel points in the suspected crack defect area image;
Figure 854719DEST_PATH_IMAGE007
and expressing the crack defect probability of the initial pixel point.
It should be noted that, when the initial pixel point is different, the adjacent pixel points are different, and accordingly, the crack defect probability of the pixel point adjacent to the initial pixel point in the suspected crack defect region image is also different.
Then, in order to improve the accuracy of defect prediction, the original image is obtained by performing inverse gray scale transformation on the identified defect image, and the saturation change of the adjacent image area of the defect image is obtained by using RGB-HSI image color space transformation. Because the normal surface color of the mechanical casting image is consistent, the image hue is consistent, the color is uniform, and the saturation of each pixel point is basically unchanged, if the mechanical casting image surface has the conditions of uneven color distribution, large color difference and large saturation change, the image area can be considered to have a crack defect. Similarly, the adjacent pixel points of the real crack defect pixel points also have saturation change, and the positions of the pixel points with the saturation change are judged as the positions where the crack defect growth is likely to occur.
Acquiring the saturation of each initial pixel point and the saturation mean value of the pixel points in the defect area in the suspected crack defect area image, constructing a supervision network loss function as branch supervision of a defect growth prediction module, wherein the supervision network loss function specifically comprises the following steps:
Figure 21258DEST_PATH_IMAGE009
wherein the content of the first and second substances,
Figure 974171DEST_PATH_IMAGE010
representing the saturation of the initial pixel point;
Figure 928482DEST_PATH_IMAGE011
representing the mean saturation of the defect area in the image of the suspected crack defect area.
Specifically, a target loss function is constructed according to the contrast loss function and the supervision network loss function, and the target loss function specifically includes:
Figure 334056DEST_PATH_IMAGE013
wherein the content of the first and second substances,
Figure 304286DEST_PATH_IMAGE006
representing the crack defect probability of pixel points adjacent to the initial pixel points in the suspected crack defect area image;
Figure 472224DEST_PATH_IMAGE007
representing the crack defect probability of the initial pixel point;
Figure 236918DEST_PATH_IMAGE010
representing the saturation of the initial pixel point;
Figure 129788DEST_PATH_IMAGE011
indicating a suspected crack defect regionSaturation mean of defect regions in the domain image.
Constructing a defect growth prediction network, and training the defect growth prediction network by using a target loss function, wherein the specific process comprises the following steps:
(1) the input of the network is a suspected crack defect area image, and the suspected crack defect area image comprises a determined defect image;
(2) the structure of the network is an encoder-decoder, the defect probability and the saturation of pixel points in the suspected crack defect area image are extracted through the encoder, the loss function is calculated, the pixel points meeting the convergence of the loss function are extracted, and the defect image and the pixel points which are possibly subjected to defect growth are output through the decoder;
(3) the loss function utilizes the constructed target loss function, the target loss function meets the convergence, the network training is completed, and the output of the network is a defect image and pixel points which are possible to increase defects.
And predicting the subsequent defect growth trend based on the trained defect growth prediction network.
And S500, detecting the defects of the mechanical casting based on the trained convolutional neural network.
When the defect detection is needed, acquiring an image of the mechanical casting, acquiring an image of a suspected crack defect area, inputting the image of the suspected crack defect area into a trained convolutional neural network for detection, and obtaining real crack defect pixel points so as to obtain a defect area.
Inputting the suspected crack defect area image into the defect growth prediction network after the training is finished based on the defect growth prediction network after the training is finished, acquiring pixel points of the defect image which are possible to carry out defect growth, and acquiring the defect growth direction according to the acquired pixel points which are possible to carry out defect growth, thereby acquiring a defect growth trend image. In summary, in the embodiment of the present invention, an initial image of a mechanical casting is subjected to image segmentation to obtain an image of a suspected crack defect area, then, principal component analysis is performed on the image of the suspected crack defect area to obtain different principal component directions, different convolution kernel sizes and sliding step lengths are formulated according to the obtained different principal component directions, each image of the suspected crack defect area is input to a convolution neural network, and feature extraction is performed on the image based on the corresponding convolution kernel size, so as to train the convolution network; and finally, detecting the defects of the mechanical casting based on the trained convolutional neural network, greatly improving the speed of network training due to less parameters of the network, improving the detection accuracy while improving the details of characteristic comparison in the network training process, and further improving the efficiency of defect identification.
Based on the same inventive concept as the method embodiment, the embodiment of the invention also provides a mechanical casting defect detection system based on artificial intelligence, which comprises: a processor, a memory, and a computer program stored in the memory and executable on the processor. The processor, when executing the computer program, implements the steps of one embodiment of the artificial intelligence based mechanical casting defect detection method described above, such as the steps shown in fig. 1. The method for detecting defects of mechanical castings based on artificial intelligence has been described in detail in the above embodiments, and is not repeated.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (8)

1. A mechanical casting defect detection method based on artificial intelligence is characterized by comprising the following steps:
acquiring at least two mechanical casting initial images;
carrying out image segmentation on the initial image of the mechanical casting to obtain an image of a suspected crack defect area;
performing principal component analysis on the suspected crack defect area image to obtain a principal component direction of the suspected crack defect area image, and determining different convolution kernel sizes and sliding step lengths of the suspected crack defect area image based on different principal component directions of the suspected crack defect area image;
inputting each suspected crack defect area image into a convolutional neural network, extracting image characteristics based on the corresponding convolutional kernel size and sliding step length, and finally outputting the probability that each pixel point in the suspected crack defect area image belongs to a crack defect;
detecting the defects of the mechanical casting based on the trained convolutional neural network;
the step of determining different convolution kernel sizes and sliding step lengths of the images of the suspected crack defect areas based on different principal component directions of the images of the suspected crack defect areas comprises the following steps:
setting three angle ranges, wherein the target directions corresponding to the three angle ranges are respectively a horizontal direction, a vertical direction and a target inclination direction, and the angle corresponding to the target inclination direction is 45oOr 135oDirection;
acquiring an angle range of the main component direction of each suspected crack defect area image based on the different main component directions of each suspected crack defect area image, and determining a target direction corresponding to each suspected crack defect area image according to the angle range;
determining the convolution kernel size and the sliding step length of each suspected crack defect area image according to the target direction corresponding to each suspected crack defect area image, wherein when the principal component direction is the horizontal direction, the convolution kernel size is
Figure DEST_PATH_IMAGE002
The sliding step length in the horizontal direction is 1, and the sliding step length in the vertical direction is 2; when the principal component direction is the vertical direction, the convolution kernel size is
Figure DEST_PATH_IMAGE004
The vertical sliding step length is 1, and the horizontal sliding step length is 2; when the direction of the main component is 45oOr 135oIn direction, the convolution kernel size is
Figure DEST_PATH_IMAGE006
The horizontal direction sliding step length is 1, and the vertical direction sliding step length is 1.
2. The method of claim 1, wherein the step of image segmenting the mechanical casting initial image to obtain an image of a suspected crack defect region comprises:
acquiring the gray level of each pixel point in the initial image, counting the gray level of each pixel point, and forming a gray distribution curve by taking the gray level as an abscissa and taking the number of gray statistics as an ordinate;
selecting a gray frequency threshold straight line to divide the gray distribution curve into an upper part and a lower part, obtaining gray levels corresponding to all gray statistics times below the gray frequency threshold straight line, taking the positions of all pixel points corresponding to the obtained gray levels as suspected crack defect points, and obtaining images of the suspected crack defect areas according to the suspected crack defect points.
3. The method according to claim 1, wherein the step of performing principal component analysis on the suspected crack defect region image to obtain a principal component direction of the suspected crack defect region image comprises:
performing principal component analysis on each pixel point of the suspected crack defect area to obtain a characteristic value of each pixel point;
sorting the characteristic values of each pixel point from big to small, acquiring characteristic values of a preset number, and further acquiring characteristic vectors corresponding to the characteristic values of the preset number to obtain the characteristic vectors of the suspected crack defect area;
and superposing the characteristic vectors to obtain a superposed vector direction, wherein the superposed vector direction is used as a main component direction of the suspected crack defect area image.
4. The method of claim 1, wherein prior to the step of detecting the defects in the mechanical casting based on the trained convolutional neural network, the method further comprises the steps of:
based on the probability that each pixel point in the suspected crack defect area image belongs to a crack defect, acquiring pixel points of which the probability of the crack defect is greater than or equal to a preset probability threshold from pixel points belonging to a normal area to obtain initial pixel points, wherein the preset probability threshold is less than 0.5, and the error value between the preset probability threshold and 0.5 is less than a preset error value;
constructing a contrast loss function based on the probability that the initial pixel point belongs to the crack defect and the probability that the pixel point adjacent to the initial pixel point in the suspected crack defect area image belongs to the crack defect;
acquiring the saturation of the initial pixel point and the saturation mean value of the defect area in the suspected crack defect area image, and further constructing a supervision network loss function;
constructing a target loss function according to the contrast loss function and the supervision network loss function;
training a defect growth prediction network according to the target loss function;
correspondingly, after the step of detecting the defect of the mechanical casting based on the trained convolutional neural network, the method further comprises the following steps:
and based on the trained defect growth prediction network, performing defect growth prediction on the defect region of the mechanical casting.
5. The method according to claim 4, characterized in that the contrast loss function is in particular:
Figure DEST_PATH_IMAGE008
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE010
representing the probability of crack defects of pixel points adjacent to the initial pixel points in the suspected crack defect area image;
Figure DEST_PATH_IMAGE012
and expressing the crack defect probability of the initial pixel point.
6. The method according to claim 4, wherein the supervision network loss function is specifically:
Figure DEST_PATH_IMAGE014
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE016
representing the saturation of the initial pixel point;
Figure DEST_PATH_IMAGE018
and representing the mean saturation value of the defect area in the suspected crack defect area image.
7. The method according to claim 4, characterized in that the objective loss function is in particular:
Figure DEST_PATH_IMAGE020
wherein the content of the first and second substances,
Figure 732998DEST_PATH_IMAGE010
representing the probability of crack defects of pixel points adjacent to the initial pixel points in the suspected crack defect area image;
Figure 978035DEST_PATH_IMAGE012
representing the crack defect probability of the initial pixel point;
Figure 85668DEST_PATH_IMAGE016
representing the saturation of the initial pixel point;
Figure 575817DEST_PATH_IMAGE018
and representing the mean saturation value of the defect area in the suspected crack defect area image.
8. An artificial intelligence based mechanical casting defect detection system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the method of claim 1
Figure DEST_PATH_IMAGE022
7 the steps of any one of the methods.
CN202111089875.XA 2021-09-17 2021-09-17 Mechanical casting defect detection method and system based on artificial intelligence Active CN113538433B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111089875.XA CN113538433B (en) 2021-09-17 2021-09-17 Mechanical casting defect detection method and system based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111089875.XA CN113538433B (en) 2021-09-17 2021-09-17 Mechanical casting defect detection method and system based on artificial intelligence

Publications (2)

Publication Number Publication Date
CN113538433A true CN113538433A (en) 2021-10-22
CN113538433B CN113538433B (en) 2021-11-26

Family

ID=78093307

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111089875.XA Active CN113538433B (en) 2021-09-17 2021-09-17 Mechanical casting defect detection method and system based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN113538433B (en)

Cited By (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113724258A (en) * 2021-11-02 2021-11-30 山东中都机器有限公司 Conveyor belt tearing detection method and system based on image processing
CN113793337A (en) * 2021-11-18 2021-12-14 汶上海纬机车配件有限公司 Locomotive accessory surface abnormal degree evaluation method based on artificial intelligence
CN114299066A (en) * 2022-03-03 2022-04-08 清华大学 Defect detection method and device based on salient feature pre-extraction and image segmentation
CN114387438A (en) * 2022-03-23 2022-04-22 武汉锦辉压铸有限公司 Machine vision-based die casting machine parameter regulation and control method
CN114723701A (en) * 2022-03-31 2022-07-08 南通博莹机械铸造有限公司 Gear defect detection method and system based on computer vision
CN114842011A (en) * 2022-07-04 2022-08-02 启东万惠机械制造有限公司 Bearing wear detection method and system based on image processing
CN114894804A (en) * 2022-04-22 2022-08-12 江苏名欧精密机械有限公司 Method for detecting surface cracks of precision standard part
CN114937004A (en) * 2022-04-28 2022-08-23 南通启锦智能科技有限公司 Method for detecting surface pore defects of mechanical parts based on computer vision
CN114937042A (en) * 2022-07-26 2022-08-23 海门市腾飞橡塑厂 Plastic product quality evaluation method based on machine vision
CN114972187A (en) * 2022-04-20 2022-08-30 北京与子成科技有限公司 Crack defect evaluation method based on artificial intelligence
CN114998350A (en) * 2022-08-04 2022-09-02 山东泗水泰和石材有限公司 Stone defect detection method based on image processing
CN115018840A (en) * 2022-08-08 2022-09-06 珠海市南特金属科技股份有限公司 Method, system and device for detecting cracks of precision casting
CN115239709A (en) * 2022-09-21 2022-10-25 南通钇龙玻璃制品有限公司 Quality detection method for glass bottle gold stamping process
CN115496918A (en) * 2022-11-16 2022-12-20 山东高速股份有限公司 Method and system for detecting abnormal highway conditions based on computer vision
CN115512252A (en) * 2022-11-18 2022-12-23 东北电力大学 Unmanned aerial vehicle-based power grid inspection automation method and system
CN115564771A (en) * 2022-11-17 2023-01-03 山东上辰建设集团有限公司 Concrete crack identification method based on building foundation column
CN115861307A (en) * 2023-02-21 2023-03-28 深圳市百昌科技有限公司 Fascia gun power supply drive plate welding fault detection method based on artificial intelligence
CN116309557A (en) * 2023-05-16 2023-06-23 山东聚宁机械有限公司 Method for detecting fracture of track shoe of excavator
CN116645364A (en) * 2023-07-18 2023-08-25 金乡县金沪合金钢有限公司 Alloy steel casting air hole defect detection method based on image data
CN116758061A (en) * 2023-08-11 2023-09-15 山东优奭趸泵业科技有限公司 Casting surface defect detection method based on computer vision
CN117078667A (en) * 2023-10-13 2023-11-17 山东克莱蒙特新材料科技有限公司 Mineral casting detection method based on machine vision
CN117095004A (en) * 2023-10-20 2023-11-21 金成技术股份有限公司 Excavator walking frame main body welding deformation detection method based on computer vision
CN117893532A (en) * 2024-03-14 2024-04-16 山东神力索具有限公司 Die crack defect detection method for die forging rigging based on image processing

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090009755A1 (en) * 2007-07-02 2009-01-08 Masayuki Yamamoto Method for detecting position of defect on semiconductor wafer
US20110193952A1 (en) * 2010-02-09 2011-08-11 Aisin Seiki Kabushiki Kaisha Defect detection apparatus and defect detection method
CN105118044A (en) * 2015-06-16 2015-12-02 华南理工大学 Method for automatically detecting defects of wheel-shaped cast product
CN107085846A (en) * 2017-05-08 2017-08-22 湘潭大学 Surface Flaw image-recognizing method
CN108230317A (en) * 2018-01-09 2018-06-29 北京百度网讯科技有限公司 Steel plate defect detection sorting technique, device, equipment and computer-readable medium
CN109741326A (en) * 2019-01-14 2019-05-10 千顺智能(珠海)有限公司 Based on the determination system and method for the magnetic core surface crack testing of convolutional neural networks
WO2020129799A1 (en) * 2018-12-17 2020-06-25 株式会社モリタ製作所 Identification device, scanner system, identification method, and identification program
CN112232400A (en) * 2020-10-12 2021-01-15 太原科技大学 Stainless steel weld ultrasonic defect detection method based on depth feature fusion

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090009755A1 (en) * 2007-07-02 2009-01-08 Masayuki Yamamoto Method for detecting position of defect on semiconductor wafer
US20110193952A1 (en) * 2010-02-09 2011-08-11 Aisin Seiki Kabushiki Kaisha Defect detection apparatus and defect detection method
CN105118044A (en) * 2015-06-16 2015-12-02 华南理工大学 Method for automatically detecting defects of wheel-shaped cast product
CN107085846A (en) * 2017-05-08 2017-08-22 湘潭大学 Surface Flaw image-recognizing method
CN108230317A (en) * 2018-01-09 2018-06-29 北京百度网讯科技有限公司 Steel plate defect detection sorting technique, device, equipment and computer-readable medium
WO2020129799A1 (en) * 2018-12-17 2020-06-25 株式会社モリタ製作所 Identification device, scanner system, identification method, and identification program
CN109741326A (en) * 2019-01-14 2019-05-10 千顺智能(珠海)有限公司 Based on the determination system and method for the magnetic core surface crack testing of convolutional neural networks
CN112232400A (en) * 2020-10-12 2021-01-15 太原科技大学 Stainless steel weld ultrasonic defect detection method based on depth feature fusion

Cited By (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113724258A (en) * 2021-11-02 2021-11-30 山东中都机器有限公司 Conveyor belt tearing detection method and system based on image processing
CN113793337A (en) * 2021-11-18 2021-12-14 汶上海纬机车配件有限公司 Locomotive accessory surface abnormal degree evaluation method based on artificial intelligence
CN114299066A (en) * 2022-03-03 2022-04-08 清华大学 Defect detection method and device based on salient feature pre-extraction and image segmentation
CN114299066B (en) * 2022-03-03 2022-05-31 清华大学 Defect detection method and device based on salient feature pre-extraction and image segmentation
CN114387438A (en) * 2022-03-23 2022-04-22 武汉锦辉压铸有限公司 Machine vision-based die casting machine parameter regulation and control method
CN114723701A (en) * 2022-03-31 2022-07-08 南通博莹机械铸造有限公司 Gear defect detection method and system based on computer vision
CN114972187A (en) * 2022-04-20 2022-08-30 北京与子成科技有限公司 Crack defect evaluation method based on artificial intelligence
CN114972187B (en) * 2022-04-20 2024-01-02 烟台大视工业智能科技有限公司 Crack defect evaluation method based on artificial intelligence
CN114894804A (en) * 2022-04-22 2022-08-12 江苏名欧精密机械有限公司 Method for detecting surface cracks of precision standard part
CN114894804B (en) * 2022-04-22 2024-06-21 辽宁中科力勒检测技术服务有限公司 Surface crack detection method for precision standard part
CN114937004A (en) * 2022-04-28 2022-08-23 南通启锦智能科技有限公司 Method for detecting surface pore defects of mechanical parts based on computer vision
CN114937004B (en) * 2022-04-28 2024-01-02 烟台大视工业智能科技有限公司 Method for detecting surface air hole defects of mechanical part based on computer vision
CN114842011A (en) * 2022-07-04 2022-08-02 启东万惠机械制造有限公司 Bearing wear detection method and system based on image processing
CN114937042B (en) * 2022-07-26 2022-10-04 海门市腾飞橡塑厂 Plastic product quality evaluation method based on machine vision
CN114937042A (en) * 2022-07-26 2022-08-23 海门市腾飞橡塑厂 Plastic product quality evaluation method based on machine vision
CN114998350B (en) * 2022-08-04 2022-10-25 山东泗水泰和石材有限公司 Stone defect detection method based on image processing
CN114998350A (en) * 2022-08-04 2022-09-02 山东泗水泰和石材有限公司 Stone defect detection method based on image processing
CN115018840B (en) * 2022-08-08 2022-11-18 珠海市南特金属科技股份有限公司 Method, system and device for detecting cracks of precision casting
CN115018840A (en) * 2022-08-08 2022-09-06 珠海市南特金属科技股份有限公司 Method, system and device for detecting cracks of precision casting
CN115239709A (en) * 2022-09-21 2022-10-25 南通钇龙玻璃制品有限公司 Quality detection method for glass bottle gold stamping process
CN115496918A (en) * 2022-11-16 2022-12-20 山东高速股份有限公司 Method and system for detecting abnormal highway conditions based on computer vision
CN115564771A (en) * 2022-11-17 2023-01-03 山东上辰建设集团有限公司 Concrete crack identification method based on building foundation column
CN115512252A (en) * 2022-11-18 2022-12-23 东北电力大学 Unmanned aerial vehicle-based power grid inspection automation method and system
CN115861307A (en) * 2023-02-21 2023-03-28 深圳市百昌科技有限公司 Fascia gun power supply drive plate welding fault detection method based on artificial intelligence
CN115861307B (en) * 2023-02-21 2023-04-28 深圳市百昌科技有限公司 Fascia gun power supply driving plate welding fault detection method based on artificial intelligence
CN116309557A (en) * 2023-05-16 2023-06-23 山东聚宁机械有限公司 Method for detecting fracture of track shoe of excavator
CN116645364B (en) * 2023-07-18 2023-10-27 金乡县金沪合金钢有限公司 Alloy steel casting air hole defect detection method based on image data
CN116645364A (en) * 2023-07-18 2023-08-25 金乡县金沪合金钢有限公司 Alloy steel casting air hole defect detection method based on image data
CN116758061B (en) * 2023-08-11 2023-10-31 山东优奭趸泵业科技有限公司 Casting surface defect detection method based on computer vision
CN116758061A (en) * 2023-08-11 2023-09-15 山东优奭趸泵业科技有限公司 Casting surface defect detection method based on computer vision
CN117078667A (en) * 2023-10-13 2023-11-17 山东克莱蒙特新材料科技有限公司 Mineral casting detection method based on machine vision
CN117078667B (en) * 2023-10-13 2024-01-09 山东克莱蒙特新材料科技有限公司 Mineral casting detection method based on machine vision
CN117095004A (en) * 2023-10-20 2023-11-21 金成技术股份有限公司 Excavator walking frame main body welding deformation detection method based on computer vision
CN117095004B (en) * 2023-10-20 2024-01-12 金成技术股份有限公司 Excavator walking frame main body welding deformation detection method based on computer vision
CN117893532A (en) * 2024-03-14 2024-04-16 山东神力索具有限公司 Die crack defect detection method for die forging rigging based on image processing
CN117893532B (en) * 2024-03-14 2024-05-24 山东神力索具有限公司 Die crack defect detection method for die forging rigging based on image processing

Also Published As

Publication number Publication date
CN113538433B (en) 2021-11-26

Similar Documents

Publication Publication Date Title
CN113538433B (en) Mechanical casting defect detection method and system based on artificial intelligence
CN109816652B (en) Complex casting defect identification method based on gray level significance
CN116664559B (en) Machine vision-based memory bank damage rapid detection method
CN109523529B (en) Power transmission line defect identification method based on SURF algorithm
CN109978848B (en) Method for detecting hard exudation in fundus image based on multi-light-source color constancy model
CN116168026A (en) Water quality detection method and system based on computer vision
CN115063430B (en) Electric pipeline crack detection method based on image processing
CN114782329A (en) Bearing defect damage degree evaluation method and system based on image processing
CN116468728B (en) Production quality detection method for hardware stamping part die
CN116883408B (en) Integrating instrument shell defect detection method based on artificial intelligence
CN116664565A (en) Hidden crack detection method and system for photovoltaic solar cell
CN111814825B (en) Apple detection grading method and system based on genetic algorithm optimization support vector machine
CN115311289A (en) Method for detecting oil stain defects of plain-color cloth
CN117635615B (en) Defect detection method and system for realizing punching die based on deep learning
CN114926387A (en) Weld defect detection method and device based on background estimation and edge gradient suppression
CN115272350A (en) Method for detecting production quality of computer PCB mainboard
CN116823725A (en) Aeroengine blade surface defect detection method based on support vector machine
CN117746077A (en) Chip defect detection method, device, equipment and storage medium
CN115797314B (en) Method, system, equipment and storage medium for detecting surface defects of parts
CN117036314A (en) Method for detecting oxidation area of high-density flexible IC substrate
CN116433978A (en) Automatic generation and automatic labeling method and device for high-quality flaw image
CN116486273A (en) Method for extracting water body information of small sample remote sensing image
CN115082449B (en) Electronic component defect detection method
CN113643290B (en) Straw counting method and device based on image processing and storage medium
CN116563659A (en) Optical smoke detection method combining priori knowledge and feature classification

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20220314

Address after: 226100 Fuxing Road, Yuet Lai Town, Haimen, Nantong, Jiangsu, 20

Patentee after: NANTONGYOUYUAN ART DESIGN Co.,Ltd.

Address before: 226000 group 10, Yuelai village, Yuelai Town, Haimen City, Nantong City, Jiangsu Province

Patentee before: Haimen chuangrui Machinery Co.,Ltd.