WO2020228111A1 - 一种基于x光图像的钢丝绳芯输送带接头抽动检测方法 - Google Patents

一种基于x光图像的钢丝绳芯输送带接头抽动检测方法 Download PDF

Info

Publication number
WO2020228111A1
WO2020228111A1 PCT/CN2019/094535 CN2019094535W WO2020228111A1 WO 2020228111 A1 WO2020228111 A1 WO 2020228111A1 CN 2019094535 W CN2019094535 W CN 2019094535W WO 2020228111 A1 WO2020228111 A1 WO 2020228111A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
joint
value
area
conveyor belt
Prior art date
Application number
PCT/CN2019/094535
Other languages
English (en)
French (fr)
Inventor
孙正
黄元麒
孙佳胜
卢纪丽
张晓光
徐桂云
蔺康
张春梅
李辉
Original Assignee
枣庄学院
徐州大恒测控技术有限公司
中国矿业大学
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 枣庄学院, 徐州大恒测控技术有限公司, 中国矿业大学 filed Critical 枣庄学院
Publication of WO2020228111A1 publication Critical patent/WO2020228111A1/zh

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
    • 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
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • 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
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • 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

Definitions

  • the invention belongs to the field of non-destructive testing, and in particular relates to a method for detecting twitching of a steel cord conveyor belt joint based on X-ray images.
  • Belt conveyor transportation is the main mode of transportation in production fields such as coal mines.
  • the conveyor belts used are mainly steel cord conveyor belts.
  • Steel cord conveyor belts are rubber conveyor belts with steel cords as skeleton materials, which are widely used in coal, mines, ports and other fields.
  • a complete steel cord conveyor belt is formed by overlapping several steel cord conveyor belts.
  • the joint area is the lowest and weakest part of the entire steel wire rope core conveyor belt. Wire rope twitching failures often occur, leading to major safety accidents such as belt breakage, which seriously affects safety production. Therefore, it is very important to timely and accurately detect the joint status of the steel cord conveyor belt.
  • the present invention proposes an X-ray image-based wire rope core conveyor belt joint twitch detection method.
  • the purpose is to accurately calculate the end point of each wire rope joint of the wire rope core conveyor belt through the X-ray image of the wire rope core conveyor belt joint Twitching distance improves the operability and accuracy of fault detection of steel cord conveyor belt joints.
  • the technical solution adopted by the present invention is: a method for detecting ticking of a steel cord conveyor belt joint based on X-ray images, including the following steps:
  • Step 1 Take the X-ray image of the steel cord conveyor belt with complete joints collected at time T1, that is, the original image at time T1 as the reference image;
  • Step 2 Perform horizontal mean filtering and global mean filtering on the original image, using the image after global mean filtering as a reference image, and perform dynamic threshold processing on the image after horizontal mean filtering to obtain image A;
  • Step 3 Perform morphological closing operation, hole filling and opening operations on image A to obtain image B;
  • Step 4 Extract the connected components of image B, filter out the joint layered areas from each connected component according to the area feature, and take the union of each layered area to obtain image C;
  • Step 5 Extract the connected components of image A and image C respectively, and calculate the intersection of the connected components in the two images to obtain image D;
  • Step 6 Use the minimum circumscribed rectangle operator to obtain the minimum circumscribed rectangle of each joint layered area in image D, and obtain the direction value of each rectangle, that is, the slope of the upper and lower sides of the rectangle, to obtain image E;
  • Step 7 Use the direction value of each minimum circumscribed rectangle, that is, the slope, and the coordinate points corresponding to the maximum and minimum ordinates in the rectangular area to calculate the fitting straight line of the lower and upper joints of each joint layered area;
  • Step 8 Perform global mean filtering on the original image, use the mean filtered image as a reference image, perform dynamic threshold processing on the original image, and perform an opening operation on the thresholded image to obtain image F;
  • Step 9 Perform morphological corrosion and open operations on image F to obtain image G;
  • Step 10 Perform the intersection operation on each minimum circumscribed rectangle in image E and image G respectively, and extract connected components of the image after the intersection, and then filter out the joint gap area from each connected component according to the area feature to obtain image H ;
  • Step 11 Extract the center of gravity of the joint neutral area of each layer obtained in step 10, calculate the distance from each center of gravity to the fitting straight line of the corresponding upper and lower joints, and further calculate the state value of each upper and lower joint point of each layer of joints , Until the last minimum bounding rectangle calculation is completed;
  • Step 12 Take the X-ray image of the steel cord conveyor belt with complete joints collected at time T2, that is, the original image at time T2 as the image to be inspected, and process the upper and lower parts of each layer of the joint to be inspected according to steps 2-11 Connection point status value;
  • Step 13 After the image to be inspected and the reference image are processed in steps 2-11, the state values of the upper and lower joint points of each layer of the image to be inspected and the reference image are respectively obtained, and the corresponding values in the image to be inspected and the reference image are calculated The difference of the state value of each joint point of, the difference is regarded as the twitch distance of the joint point;
  • Step 14 Calculate the mean value of the joint point state value of the reference image, use the t multiple of the mean value as the threshold, and determine whether the twitch distance value of each joint point of the image to be inspected is greater than this threshold. If it is greater than this threshold, it is determined that the joint point has occurred Twitch failure.
  • step 2 horizontal average filtering and global average filtering are performed on the original image respectively, and the image after global average filtering is used as a reference image, and dynamic threshold processing is performed on the image after horizontal average filtering.
  • the steps are as follows:
  • Step 2.1 Select the average template of a*1 size to perform horizontal average filtering on the original image; because the gray value difference between the joint area and the non-joint area in the image, the joint empty area and the surrounding area is large, the horizontal
  • the filtering process is beneficial to highlight the joint neutral area and the edges of the upper and lower ends of the overall joint and the non-joint area;
  • Step 2.2 Select a mean template of size a*b to perform global mean filtering on the original image
  • Step 2.3 Take the global mean filtered image as the reference image, and perform dynamic threshold processing on the horizontal mean filtered image; compare the horizontal mean filtered image with any point in the reference image for gray value, if the horizontal mean filtered after If the gray value of the pixel in the image is higher than the threshold l, the gray value of the pixel is set to 256, otherwise, the gray value of the pixel is 0.
  • step 3 the morphological closing operation is performed on image A, and the steps of hole filling and opening operations are as follows:
  • Step 3.1 Select a rectangular structure element with a size of c*d to perform morphological closing operation on the image after dynamic threshold segmentation, that is, first expand and then corrode;
  • Step 3.2 Fill holes in each area extracted during the morphological closing operation
  • Step 3.3 Select a rectangular structure element with a size of a*1 to open the hole-filled image obtained in Step 3.2 to remove the interference in the image.
  • step 4 the steps of performing connected component extraction, connected component area feature screening, and connected component union for image B are as follows:
  • Step 4.1 Perform connected component extraction on the image obtained in step 3;
  • Step 4.2 According to the connected component characteristics, filter out the joint layered area from the connected components extracted in step 4.1; the connected component feature is: whether the area of each connected component is within the interval range [S 1 , S 2 ], S 1 and S 2 respectively represent the lower limit and upper limit of the preset area value. If it is within this interval, the connected component is determined to be a joint layered area, otherwise, it is not a joint layered area;
  • Step 4.3 Take the union of each joint layered area obtained in Step 4.2.
  • step 5 since the layered area of the joint extracted after the processing of steps 2 to 4 may be in a distorted state, that is, the obtained area is not the size of the real joint layered area; therefore, the connected components of image A need to be extracted As well as the connected components of image C, the intersection of the two is performed to obtain an accurate joint layered area, that is, a more accurate joint layered area image.
  • the minimum bounding rectangle of each joint layered area can be extracted from the original image, and the direction value of each rectangle, that is, the slope of the upper and lower sides of the rectangle can be obtained.
  • step 7 using the slopes of the upper and lower sides of each minimum bounding rectangle, the coordinate points corresponding to the maximum and minimum ordinates in the rectangular area are used to calculate the lower and upper joints of each joint layered area. Close the line; the method is as follows:
  • Step 7.2 Set the direction of the wire rope in the original image as the Y-axis direction, the direction of the conveyor belt is the positive direction of the Y-axis, and the direction perpendicular to the wire rope is the X-axis direction; first calculate the coordinates of the center of gravity of each minimum bounding rectangle, according to each coordinate Sort the joints in layers according to the Y value size;
  • Step 7.3 Select the pixel coordinates (x m , y max ) corresponding to the maximum Y value in the Y-axis direction and the pixel coordinates (x n , y min );
  • k up k down ; from this, the intercept b n can be calculated, and finally the straight line function under the rectangle is obtained, that is, the joint fitting straight line;
  • Step 7.6 follow steps 7.3-7.5 to obtain the upper and lower straight line functions for each minimum circumscribed rectangle in turn, and finally obtain the layered fitting straight line of each joint in the complete joint; among them, the upper straight line obtained from the first circumscribed rectangle and The bottom straight line obtained by the last circumscribed rectangle is useless when calculating the joint twitching distance, so these two straight lines are eliminated.
  • step 8 global mean filtering is performed on the original image, the image after mean filtering is used as a reference image, dynamic threshold processing is performed on the original image, and the threshold processing image is opened.
  • Step 8.1 Select the mean filter template with the size of e*f to filter the original image, and use the filtered image as the reference image to perform dynamic threshold segmentation on the original image, that is, to compare the gray value of any point in the original image and the reference image , If the gray value of the pixel in the original image is higher than the threshold 1, set the gray value of the pixel to 256, otherwise, the gray value of the pixel is 0;
  • Step 8.2 Select a circular structure element with a radius of r 1 to perform a morphological opening operation on the image after the threshold processing, so as to remove the noise of the image and the burr on the edge of the wire rope.
  • step 9 the steps of performing morphological erosion and opening operations on the image F are as follows:
  • Step 9.1 Select a rectangular structural element with a size of g*1 to corrode the image obtained in step 8, reduce the distance between adjacent steel ropes, and highlight the gap area of the joint;
  • Step 9.2 Use a circular structure element with a radius of r 2 to perform morphological opening processing on the corroded image again to remove the noise in the image processing process.
  • step 10 the intersection of each minimum circumscribed rectangle in the image E and the image G is respectively performed; after the intersection operation, each joint gap in the joint layered area is extracted, and at the same time, the small non-joint gap It is also extracted; then the connected components of the image after the intersection are extracted, and the true connected components of the joint neutral are screened out according to the characteristics of the connected components, that is, the area is greater than S 3 .
  • step 11 the method for extracting the center of gravity of the gap area of each layer of joint is as follows:
  • S11 Set the direction of the wire rope as the Y-axis direction, the direction of the conveyor belt is the positive direction of the Y-axis, and the direction perpendicular to the wire rope is the X-axis direction; first calculate the coordinates (x, y) of the center of gravity of each minimum circumscribed rectangle. The y values of the barycentric coordinates are sorted in layers;
  • S13 Calculate the barycentric coordinates (x, y) of each joint gap in each set, and sort and save the coordinates in ascending order according to the x value of the barycentric coordinate.
  • step 11 the calculation method of the state values of the upper and lower joint points of each layer of joints is as follows:
  • step 13 the difference between the state value of each joint point corresponding to the image to be inspected and the reference image is calculated as the joint point twitch distance; the calculation method is:
  • each difference is the twitch distance of each joint point.
  • the wire rope core conveyor belt joint twitch detection method designed by the present invention adopts the method of extracting the smallest circumscribed rectangle of the joint layered area and the center of gravity of the joint neutral area, avoiding the large error caused by directly extracting the joint points to calculate the twitching distance , It is suitable for low-quality X-ray images of steel cord conveyor belts collected under harsh working conditions.
  • the method proposed in the present invention improves the detection accuracy of wire rope core conveyor belt joint twitching, solves the problem of inaccurate calculation of internal wire rope joint twitch, is beneficial to prevent the occurrence of major belt break accidents, and has greater application value.
  • Figure 1 is a flow chart of the overall plan
  • Figure 2 is a flow chart of extracting the smallest bounding rectangle
  • Figure 3 is the result of the intermediate process of extracting the smallest bounding rectangle
  • Figure 4 is a graph of the extraction result of the fitting straight line of the upper and lower joints
  • Figure 5 is a flow chart of extracting the center of gravity of the joint neutral area
  • Figure 6 is the result of the intermediate process of extracting the center of gravity of the joint neutral area
  • Figure 7 is a schematic diagram of the joint areas of each layer and each circumscribed rectangle.
  • an X-ray image-based wire rope core conveyor belt joint twitch detection method The overall scheme is shown in Figure 1.
  • the X-ray inspection image and the reference image of the wire rope core conveyor belt containing a complete joint are separately connected Extract the minimum bounding rectangle of the layered area and extract the center of gravity of the joint gap area. Based on this, calculate the state value of each joint point in the image, and calculate the difference between the corresponding joint point state value in the image to be inspected and the reference image , To realize the joint twitch detection of the image to be inspected.
  • Figure 2 shows the extraction process of the minimum circumscribed rectangle of the joint layered area in the X-ray image of the steel cord conveyor belt
  • Figure 3 shows the result of the intermediate process of the minimum circumscribed rectangle extraction.
  • the method for detecting twitching of a steel cord conveyor belt joint based on X-ray images of the present invention includes the following steps:
  • Step 1 Take the X-ray image of the steel cord conveyor belt with complete joints collected at time T1, that is, the original image at time T1 as the reference image;
  • Step 2 Perform horizontal mean filtering and global mean filtering on the original image, using the image after global mean filtering as a reference image, and perform dynamic threshold processing on the image after horizontal mean filtering to obtain image A;
  • Step 3 Perform morphological closing operation, hole filling and opening operations on image A to obtain image B;
  • Step 4 Extract the connected components of image B, filter out the joint layered areas from each connected component according to the area feature, and take the union of each layered area to obtain image C;
  • Step 5 Extract the connected components of image A and image C respectively, and calculate the intersection of the connected components in the two images to obtain image D;
  • Step 6 Use the minimum circumscribed rectangle operator to obtain the minimum circumscribed rectangle of each joint layered area in image D, and obtain the direction value of each rectangle, that is, the slope of the upper and lower sides of the rectangle, to obtain image E;
  • Step 7 Use the direction value of each minimum circumscribed rectangle, that is, the slope, and the coordinate points corresponding to the maximum and minimum ordinates in the rectangular area to calculate the fitting straight line of the lower and upper joints of each joint layered area;
  • Step 8 Perform global mean filtering on the original image, use the mean-filtered image as a reference image, perform dynamic threshold processing on the original image, and open the thresholded image to obtain image F;
  • Step 9 Perform morphological corrosion and open operations on image F to obtain image G;
  • Step 10 Perform the intersection operation on each minimum circumscribed rectangle in image E and image G respectively, and extract connected components of the image after the intersection, and then filter out the joint gap area from each connected component according to the area feature to obtain image H ;
  • Step 11 Extract the center of gravity of the joint neutral area of each layer obtained in step 10, calculate the distance from each center of gravity to the fitting straight line of the corresponding upper and lower joints, and further calculate the state value of each upper and lower joint point of each layer of joints , Until the last minimum bounding rectangle calculation is completed;
  • Step 12 Take the X-ray image of the steel cord conveyor belt with complete joints collected at time T2, that is, the original image at time T2 as the image to be inspected, and process the upper and lower parts of each layer of the joint to be inspected according to steps 2-11 Connection point status value;
  • Step 13 After the image to be inspected and the reference image are processed in steps 2-11, the state values of the upper and lower joint points of each layer of the image to be inspected and the reference image are respectively obtained, and the corresponding values in the image to be inspected and the reference image are calculated The difference of the state value of each joint point of, the difference is regarded as the twitch distance of the joint point;
  • Step 14 Calculate the mean value of the joint point state value of the reference image, use the t multiple of the mean value as the threshold, and determine whether the twitch distance value of each joint point of the image to be inspected is greater than this threshold. If it is greater than this threshold, it is determined that the joint point has occurred Twitch failure.
  • step 2 horizontal average filtering and global average filtering are performed on the original image, and the image after global average filtering is used as a reference image, and the horizontal average filtering image is subjected to dynamic threshold processing.
  • the steps are as follows:
  • Step 2.1 Select a 90*1 mean value template to perform horizontal mean filtering on the original image.
  • the filtering result is shown in Figure 3(a); because of the difference between the joint area and the non-joint area, the joint empty area and the surrounding area in the image The gray value difference between the two is large, and the use of horizontal filtering is beneficial to highlight the gap area of the joint and the edge of the upper and lower ends of the overall joint and the non-joint area;
  • Step 2.2 Select a mean template with a size of 90*180 to perform global mean filtering on the original image, and the filtering result is shown in Figure 3(b);
  • Step 2.3 Take the global mean filtered image as the reference image, and perform dynamic threshold processing on the horizontal mean filtered image; compare the horizontal mean filtered image with any point in the reference image for gray value, if the horizontal mean filtered after If the gray value of the pixel in the image is higher than 3, the gray value of the pixel is set to 256, otherwise, the gray value of the pixel is 0; the dynamic threshold segmentation result is shown in Figure 3(c) , The gaps of each layer of joints in the original image are connected, that is, the five long rectangular areas arranged up and down in the white area in the figure, these five areas are the joint layered areas.
  • step 3 perform morphological closing operation on image A, and the steps of hole filling and opening operations are as follows:
  • Step 3.1 Select a 10*10 rectangular structure element to perform morphological closing operation on the image after dynamic threshold segmentation, that is, first expand and then corrode.
  • the processing effect is shown in Figure 3(d); after the morphological closing operation, In Figure 3(c), the small disconnected areas in the long rectangular area are connected together. At the same time, there are small holes in the long rectangular area, especially the first and last areas counted from the top;
  • Step 3.2 Fill holes in each area extracted during the morphological closing operation.
  • the filled image is shown in Figure 3(e), and the holes in the long rectangular area are completely filled;
  • Step 3.3 In addition to the area that needs to be extracted, there are small interference areas in the image obtained from step 3.2 after the hole is filled, especially the horizontal strip interference small area; therefore, a rectangular structure element with a size of 90*1 is selected to fill the hole To remove the interference in the image, the image after the open operation is shown in Figure 3(f), in which small interference areas are eliminated.
  • step 4 the steps of performing connected component extraction, connected component area feature screening, and connected component union for image B are as follows:
  • Step 4.1 Perform connected component extraction on the image obtained in step 3;
  • Step 4.2 According to the connected component characteristics, filter out the joint layered area from the connected components extracted in step 4.1; the connected component feature is: whether the area of each connected component is within the interval range [10000,9999999], if in Within this interval, the connected component is judged to be a joint layered area, otherwise, it is not a joint layered area;
  • Step 4.3 Take the union of the layered regions of each joint obtained in step 4.2.
  • the processing result is shown in Figure 3(g).
  • Figure 3(f) the interference regions that originally existed at the left and right ends are eliminated, leaving 5 The area where the joint is layered.
  • step 5 since the 5 joint layered regions extracted after the processing of steps 2 to 4 may be in a distorted state, the obtained region is not the size of the real joint layered area; therefore, the connected components of image A need to be extracted As well as the connected components of image C, the intersection of the two is performed to obtain an accurate joint layered area, that is, a more accurate joint layered area image, and the result is shown in Figure 3(h).
  • step 6 the minimum circumscribed rectangle operator is used to obtain the minimum circumscribed rectangle of each joint layered area in image D.
  • the result is shown in Figure 3(i), and the five joint layered areas in Figure 3(h) are fitted Into a regular long rectangle.
  • the smallest bounding rectangle of each joint layered area can be accurately extracted from the original image, and the direction value of each rectangle, that is, the slope of the upper and lower sides of the rectangle can be obtained.
  • step 7 using the slopes of the upper and lower sides of each minimum circumscribed rectangle, the coordinate points corresponding to the maximum and minimum ordinates in the rectangular area are used to calculate the fitting straight line for the lower and upper joints of each joint layered area ;Methods as below:
  • Step 7.2 Set the direction of the wire rope in the original image as the Y-axis direction, the direction of the conveyor belt is the positive direction of the Y-axis, and the direction perpendicular to the wire rope is the X-axis direction; first calculate the coordinates of the center of gravity of each minimum bounding rectangle, according to each coordinate Sort the joints in layers according to the Y value size;
  • Step 7.3 Select the pixel coordinates (x m , y max ) corresponding to the maximum Y value in the Y-axis direction and the pixel coordinates (x n , y min );
  • k up k down ; from this, the intercept b n can be calculated, and finally the straight line function under the rectangle is obtained, that is, the joint fitting straight line;
  • Step 7.6 follow steps 7.3-7.5 to obtain the upper and lower straight line functions for each minimum circumscribed rectangle in turn, and finally obtain the layered fitting straight line of each joint in the complete joint; among them, the upper straight line obtained from the first circumscribed rectangle and The bottom straight line obtained by the last circumscribed rectangle is useless when calculating the joint twitching distance, so the two straight lines are eliminated, and finally 8 straight lines will be calculated.
  • Figure 4 respectively shows the extraction results of the upper and lower fitting straight lines of the second and third smallest bounding rectangles in the image to be inspected.
  • Figure 5 is a flowchart of the center of gravity extraction in the joint neutral region
  • Figure 6 shows the result of the intermediate process of the center of gravity extraction in the joint neutral region.
  • step 8 global mean filtering is performed on the original image, using the mean filtered image as a reference image, dynamic threshold processing is performed on the original image, and the threshold processing image is opened.
  • the steps are as follows:
  • Step 8.1 Select a 21*21 mean filter template to filter the original image.
  • the result is shown in Figure 6(a).
  • the filtered image is used as a reference image to perform dynamic threshold segmentation on the original image, that is, the original image and The gray value of any point in the reference image is compared. If the gray value of the pixel in the original image is higher than 3, the gray value of the pixel is set to 256, otherwise, the gray value of the pixel is 0;
  • Step 8.2 Select a circular structure element with a radius of 1 to perform the morphological opening operation on the image after the threshold processing, so as to remove the noise of the image and the burr on the edge of the wire rope;
  • Figure 6(b) shows the morphological opening operation The result graph.
  • step 9 the steps of performing morphological erosion and opening operations on image F are as follows:
  • Step 9.1 Select a 3*1 rectangular structure element to corrode the image obtained in step 8, reduce the distance between adjacent steel wires, and highlight the gap area of the joint;
  • Step 9.2 Use a circular structure element with a radius of 1.5 to perform the morphological opening operation on the corroded image again to remove the noise in the image processing;
  • Figure 6(c) is the result of the opening operation after the corrosion.
  • the neutral area of the middle joint is more obvious than before.
  • step 10 the intersection of each minimum circumscribed rectangle in image E and image G is performed respectively; the result of the intersection is shown in Figure 6(d), where each joint gap in the joint layered area is extracted.
  • the small non-joint gaps are also extracted; then the connected components of the image after the intersection are extracted, and according to the characteristics of the connected components, that is, the area is greater than 250, the true joint gaps connected components are filtered out; the screening results are as follows As shown in Figure 6(e), the white area in the figure represents the neutral area of each joint; Figure 6(f) shows the neutral area of the first layer of joints.
  • step 11 the method of extracting the center of gravity of the gap area of each layer is as follows:
  • S11 Set the direction of the wire rope as the Y-axis direction, the direction of the conveyor belt is the positive direction of the Y-axis, and the direction perpendicular to the wire rope is the X-axis direction; first calculate the coordinates (x, y) of the center of gravity of each minimum circumscribed rectangle. The y values of the barycentric coordinates are sorted in layers;
  • S13 Calculate the barycentric coordinates (x, y) of each joint gap in each set, and sort and save the coordinates in ascending order according to the x value of the barycentric coordinate.
  • step 11 the calculation method for the state values of the upper and lower joint points of each layer of joints is as follows:
  • step 13 the difference between the state value of each joint point corresponding to the image to be inspected and the reference image is calculated as the joint point twitch distance; the calculation method is:
  • each difference is the twitch distance of each joint point.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Abstract

一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，旨在准确高效计算出钢丝绳芯输送带接头抽动距离，提高钢丝绳芯输送带接头抽动检测的可操作性和准确性，以实现对接头状况的判断。检测方法基于钢丝绳芯输送带图像中接头分层区域最小外接矩形以及接头空档区域的提取结果，通过计算上下接头拟合直线、接头空档区域重心，实现对各钢丝绳接头端点的抽动距离检测。检测方法的算法高效且准确率高，能够应用于基于X光的钢丝绳芯输送带无损检测系统，实现X光的钢丝绳芯输送带接头抽动检测，有益于预防重大断带安全事故，具有较大的应用价值。

Description

一种基于X光图像的钢丝绳芯输送带接头抽动检测方法 技术领域
本发明属于无损检测领域，尤其涉及一种基于X光图像的钢丝绳芯输送带接头抽动检测方法。
背景技术
带式输送机运输是煤矿等生产领域中的主要运输方式，其使用的输送带以钢丝绳芯输送带为主。钢丝绳芯输送带是以钢丝绳作为骨架材料的橡胶输送带，广泛运用于煤炭、矿山、港口等领域。在实际应用中，一条完整的钢丝绳芯输送带由若干段钢丝绳输送带搭接而成。接头区域是整条钢丝绳芯输送带拉伸强度最低、最为薄弱的部位，经常会发生钢丝绳抽动故障，导致断带等重大安全事故的发生，严重影响安全生产。因此，及时准确的检测钢丝绳芯输送带接头状态十分重要。
随着X光检测技术的发展，能够通过X光无损检测系统获取包含钢丝绳芯输送带接头信息的X光图像，目前已存在的接头抽动距离计算方法都是先提取出钢丝绳端点信息，然后根据各端点进行直线拟合，通过计算点到拟合直线的距离实现抽动距离的计算。但由于在恶劣工况环境下采集到的X光图像质量较低，通过现有方法提取到的钢丝绳端点信息存在较大误差，导致接头抽动检测方法存在着检测效率低、准确性较差等问题。综上所述，需要一种能够高效可靠判断接头抽动故障的检测方法。
发明内容
发明目的：针对以上问题，本发明提出一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，目的是通过钢丝绳芯输送带接头的X光图像，准确计算钢丝绳芯输送带各钢丝绳接头端点的抽动距离，提高钢丝绳芯输送带接头故障检测的可操作性和准确性。
技术方案：为实现本发明的目的，本发明所采用的技术方案是：一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，包含以下步骤：
步骤1：将T1时刻采集到的含有完整接头的钢丝绳芯输送带X光图像，即T1时刻的原始图像作为基准图像；
步骤2：对原始图像分别进行横向均值滤波和全局均值滤波，以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理，得到图像A；
步骤3：对图像A进行形态学闭操作，孔洞填充、开操作，得到图像B；
步骤4：对图像B进行连通分量提取，并根据面积特征从各连通分量中筛选出接头分层区域，对各分层区域求并集，得到图像C；
步骤5：分别对图像A与图像C进行连通分量提取，并对两图像中的连通分量求交集，得到图像D；
步骤6：利用最小外接矩形算子获取图像D中各接头分层区域的最小外接矩形，并取得每个矩形的方向值，即矩形上下两边的斜率，得到图像E；
步骤7：利用每个最小外接矩形的方向值，即斜率，与此矩形区域中纵坐标最大、最小时相对应的坐标点，计算得到每个接头分层区域的下、上接头拟合直线；
步骤8：对原始图像进行全局均值滤波，以均值滤波后的图像为参考图像，对原始图像进行动态阈值处理，并对阈值处理后的图像进行开操作，得到图像F；
步骤9：对图像F进行形态学腐蚀以及开操作，得到图像G；
步骤10：分别对图像E中每一个最小外接矩形与图像G进行交集运算，并对求交后图像进行连通分量提取，再根据面积特征从各连通分量中筛选出接头空档区域，得到图像H；
步骤11：对步骤10得到的每一层接头空档区域，提取接头空档区域的重心，计算各重心到对应上下接头拟合直线的距离，进一步计算每一层接头的各上下接头点状态值，直至最后一个最小外接矩形运算完成；
步骤12：对T2时刻采集到的含有完整接头的钢丝绳芯输送带X光图像，即T2时刻的原始图像作为待检图像，根据步骤2-11处理得到待检图像的每一层接头的各上下接头点状态值；
步骤13：对待检图像与基准图像经过步骤2-11处理后，分别得到待检图像与基准图像的每一层接头的各上下接头点状态值，计算所述待检图像与基准图像中相对应的每一接头点的状态值的差值，该差值作为接头点抽动距离；
步骤14：计算基准图像接头点状态值的均值，将此均值的t倍数作为阈值，判断待检图像每一接头点抽动距离值是否大于此阈值，若大于此阈值，则判定该接头点发生了抽动故障。
进一步，在步骤2中，对原始图像分别进行横向均值滤波和全局均值滤波，以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理，步骤如下：
步骤2.1：选取a*1大小的均值模板对原始图像进行横向均值滤波；由于图像中接头区域与非接头区域之间、接头空档区域与其周围区域之间的灰度值差异较大，利用横向滤波处理有利于突出接头空档区域以及整体接头上下端与非接头区域的边缘；
步骤2.2：选取a*b大小的均值模板对原始图像进行全局均值滤波；
步骤2.3：以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理；将横向均值滤波后的图像与参考图像中任一点进行灰度值比较，若横向均值滤波后的图像中像素点灰度值高于阈值l，则设定该像素点的灰度值为256，反之，该像素点的灰度值则为0。
进一步，在步骤3中，对图像A进行形态学闭操作，孔洞填充、开操作的步骤如下：
步骤3.1：选取c*d大小的矩形结构元对动态阈值分割后的图像进行形态学闭操作，即先膨胀后腐蚀；
步骤3.2：对形态学闭操作过程中提取出来的各区域进行孔洞填充；
步骤3.3：选取a*1大小的矩形结构元对步骤3.2得到的孔洞填充后的图像进行开操作，以此去除图像中存在的干扰。
进一步，在步骤4中，对图像B进行连通分量提取、连通分量面积特征筛选、以及连通分量求并集的步骤如下：
步骤4.1：对步骤3得到的图像进行连通分量提取；
步骤4.2：根据连通分量特征，从步骤4.1提取出的各连通分量中筛选出接头分层区域；所述连通分量特征为：各连通分量面积是否在区间范围[S 1，S 2]之内，S 1和S 2分别表示预设面积值的下限和上限，若在此区间范围内，则判定此连通分量为接头分层区域，反之，则不为接头分层区域；
步骤4.3：对步骤4.2得到的各接头分层区域求并集。
进一步，在步骤5中，由于经步骤2至4处理后提取出的接头分层区域可能处于失真状态，即得到的区域并非真实接头分层面积大小的区域；因此需要提取出图像A的连通分量以及图像C的连通分量，对两者进行交集运算，以获取准确的接头分层区域，即更精确的接头分层区域图像。
通过以上步骤2至6可以从原始图像中提取出各接头分层区域的最小外接矩形，并取得每个矩形的方向值，即矩形上下两边的斜率。
进一步，在步骤7中，利用每个最小外接矩形的上下两边的斜率，与此矩形区域中纵坐标最大、最小时相对应的坐标点，计算得到每个接头分层区域的下、上接头拟合直线；方法如下：
步骤7.1：设最小外接矩形中上下两边的直线y＝k ix+b i，其中，k i为斜率，b i为截距，i＝0，1，...，2M，M为接头分层数目；
步骤7.2：设定原始图像中钢丝绳方向为Y轴方向，输送带运行的方向为Y轴正方向，垂直于钢丝绳的方向为X轴方向；首先计算出各个最小外接矩形的重心坐标，根据各个坐标的Y值大小进行接头分层排序；
步骤7.3：在拟合得到的第j个最小外接矩形中选取Y轴方向上Y值最大时对应的像素点坐标(x m，y max)以及Y值最小时对应的像素点坐标(x n，y min)；
步骤7.4：将Y值最大时对应的像素点坐标(x m，y max)代入y＝k upx+b m，其中k up代表已知的第j个矩形上边的斜率，由此可以计算得到截距b m，最后求得矩形上边的直线函数，即接头拟合直线；
步骤7.5：将Y值最小时对应的像素点坐标(x n，y min)代入y＝k downx+b n，其中k down代 表已知的第j个矩形下边的斜率，在同一矩形中，k up＝k down；由此可以计算得到截距b n，最后求得矩形下边的直线函数，即接头拟合直线；
步骤7.6：按照步骤7.3-7.5依次对每个最小外接矩形求取上下边直线函数，最后得到完整接头中每一条接头分层拟合直线；其中，从第一个外接矩形求得的上边直线以及最后一个外接矩形求得的下边直线在计算接头抽动距离时是无用的，因此将这两条直线剔除。
进一步，在步骤8中，对原始图像进行全局均值滤波，以均值滤波后的图像为参考图像，对原始图像进行动态阈值处理，并对阈值处理后的图像进行开操作，步骤如下：
步骤8.1：选取e*f大小的均值滤波模板对原始图像进行滤波处理，以滤波后的图像作为参考图像，对原始图像进行动态阈值分割，即原始图像与参考图像中任一点进行灰度值比较，若原始图像中像素点灰度值高于阈值l，则设定该像素点的灰度值为256，反之，该像素点的灰度值则为0；
步骤8.2：选取半径为r 1的圆形结构元对阈值处理后的图像进行形态学开操作，以此去除图像的噪点以及钢丝绳边缘的毛刺。
进一步，在步骤9中，对图像F进行形态学腐蚀以及开操作的步骤如下：
步骤9.1：选取g*1大小的矩形结构元对步骤8得到的图像进行腐蚀，缩小相邻钢丝绳之间的间距，突出接头空档区域；
步骤9.2：采用半径为r 2的圆形结构元再次对腐蚀后的图像进行形态学开操作处理，去除图像在处理过程中存在的噪点。
进一步，在步骤10中，分别对图像E中每一个最小外接矩形与图像G进行交集运算；求交运算后接头分层区域中各接头空档被提取了出来，同时，细小的非接头空档也被提取了出来；再对求交后图像进行连通分量提取，并根据连通分量的特征，即面积大于S 3，筛选出真正的接头空档连通分量。
进一步，在步骤11中，各层接头空档区域重心提取的方法如下：
S11：设定钢丝绳方向为Y轴方向，输送带运行的方向为Y轴正方向，垂直于钢丝绳的方向为X轴方向；首先计算出各个最小外接矩形的重心坐标(x，y)，根据各重心坐标的y值大小进行接头分层排序；
S12：将排好序的最小外接矩形逐个与图像E和图像G求交后的图像求交集，得到各行接头空档集合E i，i＝1，2，...，m，m为矩形个数；
S13：计算每个集合中各接头空档的重心坐标(x，y)，按照重心坐标的x值大小将各坐标进行升序排序并保存。
进一步，在步骤11中，每一层接头的各上下接头点状态值的计算方法如下：
S21：从第一个外接矩形开始，计算每个接头空档区域重心到上拟合直线的距离， 各距离记为J i，i＝1，2，...，n，其中n为接头空档区域重心个数；计算距离的均值，表示为：
Figure PCTCN2019094535-appb-000001
S22：计算接头空档区域重心到上拟合直线的距离J i与均值L up的差值P i＝J i-L up，i＝1，2，...，n；
S23：计算每个接头空档区域重心到下拟合直线的距离，各距离记为K i，i＝1，2，...，n，其中n为接头空档区域重心个数；计算距离的均值，表示为：
Figure PCTCN2019094535-appb-000002
S24：计算接头空档区域重心到下拟合直线距离K i与均值L down的差值Q i＝K i-L down，i＝1，2，...，n；
S25：根据步骤S21到S24的方法分别计算出五个外接矩形各对应的P i与Q i；
S26：剔除第一个外接矩形的P i与第五个外接矩形的Q i；将第一个外接矩形的Q i作为第一层接头的各上接头点的状态值，第二个外接矩形的P i作为第一层接头的各下接头点的状态值；将第二个外接矩形的Q i作为第二层接头的各上接头点的状态值，第三个外接矩形的P i作为第二层接头的各下接头点的状态值；将第三个外接矩形的Q i作为第三层接头的各上接头点的状态值，第四个外接矩形的P i作为第三层接头的各下接头点的状态值；将第四个外接矩形的Q i作为第四层接头的各上接头点的状态值，第五个外接矩形的P i作为第四层接头的各下接头点的状态值。
进一步，在步骤13中，计算所述待检图像与基准图像中相对应的每一接头点的状态值的差值，以此作为接头点抽动距离；其计算方法为：
从待检图像中第一层接头的第一个上接头点开始，计算此接头点状态值与相匹配的基准图像接头点状态值的差值，直至最后一层接头的下接头点中最后一个接头点状态值差值；各差值则为各接头点的抽动距离。
有益效果：与现有技术相比，本发明的技术方案具有以下有益的技术效果：
(1)本发明设计的钢丝绳芯输送带接头抽动检测方法，采用了提取接头分层区域最小外接矩形以及接头空档区域重心的方式，避免了直接提取接头点以计算抽动距离产生的较大误差，适应于在恶劣工况环境下采集到的低质量的钢丝绳芯输送带X光图像。
(2)本发明提出的方法提高了钢丝绳芯输送带接头抽动检测精度，解决了内部钢丝绳接头抽动计算不准确的问题，有益于预防重大断带事故的发生，具有较大的应用价值。
附图说明
图1是整体方案流程图；
图2是最小外接矩形提取流程图；
图3是最小外接矩形提取中间过程结果图；
图4是上下接头拟合直线提取结果图；
图5是接头空档区域重心提取流程图；
图6是接头空档区域重心提取中间过程结果图；
图7是各层接头区域和各个外接矩形示意图。
具体实施方式
下面结合附图和实施例对本发明的技术方案作进一步的说明。
本发明所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，整体方案流程如图1所示，对含有完整接头的钢丝绳芯输送带X光待检图像与基准图像分别进行接头分层区域的最小外接矩形提取以及接头空档区域重心提取，以此为基础，计算得到图像中各接头点状态值，通过计算待检图像与基准图像中相对应的接头点状态值的差值，实现待检图像的接头抽动检测。
图2展示了钢丝绳芯输送带X光图像中接头分层区域的最小外接矩形提取过程，图3展示了最小外接矩形提取中间过程结果图。
本发明所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，包括以下步骤：
步骤1：将T1时刻采集到的含有完整接头的钢丝绳芯输送带X光图像，即T1时刻的原始图像作为基准图像；
步骤2：对原始图像分别进行横向均值滤波和全局均值滤波，以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理，得到图像A；
步骤3：对图像A进行形态学闭操作，孔洞填充、开操作，得到图像B；
步骤4：对图像B进行连通分量提取，并根据面积特征从各连通分量中筛选出接头分层区域，对各分层区域求并集，得到图像C；
步骤5：分别对图像A与图像C进行连通分量提取，并对两图像中的连通分量求交集，得到图像D；
步骤6：利用最小外接矩形算子获取图像D中各接头分层区域的最小外接矩形，并取得每个矩形的方向值，即矩形上下两边的斜率，得到图像E；
步骤7：利用每个最小外接矩形的方向值，即斜率，与此矩形区域中纵坐标最大、最小时相对应的坐标点，计算得到每个接头分层区域的下、上接头拟合直线；
步骤8：对原始图像进行全局均值滤波，以均值滤波后的图像为参考图像，对原始 图像进行动态阈值处理，并对阈值处理后的图像进行开操作，得到图像F；
步骤9：对图像F进行形态学腐蚀以及开操作，得到图像G；
步骤10：分别对图像E中每一个最小外接矩形与图像G进行交集运算，并对求交后图像进行连通分量提取，再根据面积特征从各连通分量中筛选出接头空档区域，得到图像H；
步骤11：对步骤10得到的每一层接头空档区域，提取接头空档区域的重心，计算各重心到对应上下接头拟合直线的距离，进一步计算每一层接头的各上下接头点状态值，直至最后一个最小外接矩形运算完成；
步骤12：对T2时刻采集到的含有完整接头的钢丝绳芯输送带X光图像，即T2时刻的原始图像作为待检图像，根据步骤2-11处理得到待检图像的每一层接头的各上下接头点状态值；
步骤13：对待检图像与基准图像经过步骤2-11处理后，分别得到待检图像与基准图像的每一层接头的各上下接头点状态值，计算所述待检图像与基准图像中相对应的每一接头点的状态值的差值，该差值作为接头点抽动距离；
步骤14：计算基准图像接头点状态值的均值，将此均值的t倍数作为阈值，判断待检图像每一接头点抽动距离值是否大于此阈值，若大于此阈值，则判定该接头点发生了抽动故障。
在步骤2中，对原始图像分别进行横向均值滤波和全局均值滤波，以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理，步骤如下：
步骤2.1：选取90*1大小的均值模板对原始图像进行横向均值滤波，滤波结果如图3(a)所示；由于图像中接头区域与非接头区域之间、接头空档区域与其周围区域之间的灰度值差异较大，利用横向滤波处理有利于突出接头空档区域以及整体接头上下端与非接头区域的边缘；
步骤2.2：选取90*180大小的均值模板对原始图像进行全局均值滤波，滤波结果如图3(b)所示；
步骤2.3：以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理；将横向均值滤波后的图像与参考图像中任一点进行灰度值比较，若横向均值滤波后的图像中像素点灰度值高于3，则设定该像素点的灰度值为256，反之，该像素点的灰度值则为0；动态阈值分割结果如图3(c)所示，原始图像中每一层接头空档之间被连接，即图中白色区域中5个上下排列的长矩形的区域，这5个区域为接头分层区域。
在步骤3中，对图像A进行形态学闭操作，孔洞填充、开操作的步骤如下：
步骤3.1：选取10*10大小的矩形结构元对动态阈值分割后的图像进行形态学闭操 作，即先膨胀后腐蚀，其处理效果如图3(d)所示；经形态学闭操作之后，图3(c)中长矩形区域内断开的小区域被连接在一起，同时，长矩形的区域之中存在细小的孔洞，尤其是从上往下数第一个与最后一个区域；
步骤3.2：对形态学闭操作过程中提取出来的各区域进行孔洞填充，填充后的图像如图3(e)所示，长矩形区域中的孔洞被完全填充；
步骤3.3：由步骤3.2得到孔洞填充后的图像中除了需要提取的区域之外，存在干扰小区域，尤其是横向条状干扰小区域；因此，选取90*1大小的矩形结构元对孔洞填充后的图像进行开操作，以此去除图像中存在的干扰；开操作后的图像如图3(f)所示，图中干扰小区域被剔除。
在步骤4中，对图像B进行连通分量提取、连通分量面积特征筛选、以及连通分量求并集的步骤如下：
步骤4.1：对步骤3得到的图像进行连通分量提取；
步骤4.2：根据连通分量特征，从步骤4.1提取出的各连通分量中筛选出接头分层区域；所述连通分量特征为：各连通分量面积是否在区间范围[10000，9999999]之内，若在此区间范围内，则判定此连通分量为接头分层区域，反之，则不为接头分层区域；
步骤4.3：对步骤4.2得到的各接头分层区域求并集，处理结果如图3(g)所示，图3(f)中左右两端原本存在的干扰区域被剔除，留下了5个接头分层处的区域。
在步骤5中，由于经步骤2至4处理后提取出的5个接头分层区域可能处于失真状态，即得到的区域并非真实接头分层面积大小的区域；因此需要提取出图像A的连通分量以及图像C的连通分量，对两者进行交集运算，以获取准确的接头分层区域，即更精确的接头分层区域图像，结果如图3(h)所示。
在步骤6中，利用最小外接矩形算子获取图像D中各接头分层区域的最小外接矩形，结果如图3(i)所示，图3(h)中5个接头分层区域被拟合成规则的长矩形。
通过以上步骤2至6可以精确地从原始图像中提取出各接头分层区域的最小外接矩形，并取得每个矩形的方向值，即矩形上下两边的斜率。
在步骤7中，利用每个最小外接矩形的上下两边的斜率，与此矩形区域中纵坐标最大、最小时相对应的坐标点，计算得到每个接头分层区域的下、上接头拟合直线；方法如下：
步骤7.1：设最小外接矩形中上下两边的直线y＝k ix+b i，其中，k i为斜率，b i为截距，i＝0，1，...，2M，M为接头分层数目；
步骤7.2：设定原始图像中钢丝绳方向为Y轴方向，输送带运行的方向为Y轴正方向，垂直于钢丝绳的方向为X轴方向；首先计算出各个最小外接矩形的重心坐标，根据各个坐标的Y值大小进行接头分层排序；
步骤7.3：在拟合得到的第j个最小外接矩形中选取Y轴方向上Y值最大时对应的像素点坐标(x m，y max)以及Y值最小时对应的像素点坐标(x n，y min)；
步骤7.4：将Y值最大时对应的像素点坐标(x m，y max)代入y＝k upx+b m，其中k up代表已知的第j个矩形上边的斜率，由此可以计算得到截距b m，最后求得矩形上边的直线函数，即接头拟合直线；
步骤7.5：将Y值最小时对应的像素点坐标(x n，y min)代入y＝k downx+b n，其中k down代表已知的第j个矩形下边的斜率，在同一矩形中，k up＝k down；由此可以计算得到截距b n，最后求得矩形下边的直线函数，即接头拟合直线；
步骤7.6：按照步骤7.3-7.5依次对每个最小外接矩形求取上下边直线函数，最后得到完整接头中每一条接头分层拟合直线；其中，从第一个外接矩形求得的上边直线以及最后一个外接矩形求得的下边直线在计算接头抽动距离时是无用的，因此将这两条直线剔除，最后将计算得到8条直线。
图4分别展示了待检图像中第二、三个最小外接矩形的上下拟合直线提取结果。
图5为接头空档区域重心提取流程图，图6展示了接头空档区域重心提取中间过程结果图。
在步骤8中，对原始图像进行全局均值滤波，以均值滤波后的图像为参考图像，对原始图像进行动态阈值处理，并对阈值处理后的图像进行开操作，步骤如下：
步骤8.1：选取21*21大小的均值滤波模板对原始图像进行滤波处理，结果如图6(a)所示，以滤波后的图像作为参考图像，对原始图像进行动态阈值分割，即原始图像与参考图像中任一点进行灰度值比较，若原始图像中像素点灰度值高于3，则设定该像素点的灰度值为256，反之，该像素点的灰度值则为0；
步骤8.2：选取半径为1的圆形结构元对阈值处理后的图像进行形态学开操作，以此去除图像的噪点以及钢丝绳边缘的毛刺；如图6(b)所示为形态学开操作后结果图。
在步骤9中，对图像F进行形态学腐蚀以及开操作的步骤如下：
步骤9.1：选取3*1大小的矩形结构元对步骤8得到的图像进行腐蚀，缩小相邻钢丝绳之间的间距，突出接头空档区域；
步骤9.2：采用半径为1.5的圆形结构元再次对腐蚀后的图像进行形态学开操作处理，去除图像在处理过程中存在的噪点；图6(c)为腐蚀后开操作的结果图，图中接头空档区域相较于处理之前更明显。
在步骤10中，分别对图像E中每一个最小外接矩形与图像G进行交集运算；求交结果如图6(d)所示，图中接头分层区域中各接头空档被提取了出来，同时，细小的非接头空档也被提取了出来；再对求交后图像进行连通分量提取，并根据连通分量的特征，即面积大于250，筛选出真正的接头空档连通分量；筛选结果如图6(e)所示，图中白色 区域代表各接头空档区域；图6(f)展示了第一层接头空档区域。
在步骤11中，各层接头空档区域重心提取的方法如下：
S11：设定钢丝绳方向为Y轴方向，输送带运行的方向为Y轴正方向，垂直于钢丝绳的方向为X轴方向；首先计算出各个最小外接矩形的重心坐标(x，y)，根据各重心坐标的y值大小进行接头分层排序；
S12：将排好序的最小外接矩形逐个与图像E和图像G求交后的图像求交集，得到各行接头空档集合E i，i＝1，2，...，m，m为矩形个数；
S13：计算每个集合中各接头空档的重心坐标(x，y)，按照重心坐标的x值大小将各坐标进行升序排序并保存。
在步骤11中，每一层接头的各上下接头点状态值的计算方法如下：
S21：从第一个外接矩形开始，计算每个接头空档区域重心到上拟合直线的距离，各距离记为J i，i＝1，2，...，n，其中n为接头空档区域重心个数；计算距离的均值，表示为：
Figure PCTCN2019094535-appb-000003
S22：计算接头空档区域重心到上拟合直线的距离J i与均值L up的差值P i＝J i-L up，i＝1，2，...，n；
S23：计算每个接头空档区域重心到下拟合直线的距离，各距离记为K i，i＝1，2，...，n，其中n为接头空档区域重心个数；计算距离的均值，表示为：
Figure PCTCN2019094535-appb-000004
S24：计算接头空档区域重心到下拟合直线距离K i与均值L down的差值Q i＝K i-L down，i＝1，2，...，n；
S25：根据步骤S21到S24的方法分别计算出五个外接矩形各对应的P i与Q i；
S26：剔除第一个外接矩形的P i与第五个外接矩形的Q i；将第一个外接矩形的Q i作为第一层接头的各上接头点的状态值，第二个外接矩形的P i作为第一层接头的各下接头点的状态值；将第二个外接矩形的Q i作为第二层接头的各上接头点的状态值，第三个外接矩形的P i作为第二层接头的各下接头点的状态值；将第三个外接矩形的Q i作为第三层接头的各上接头点的状态值，第四个外接矩形的P i作为第三层接头的各下接头点的状态值；将第四个外接矩形的Q i作为第四层接头的各上接头点的状态值，第五个外接矩形的P i作为第四层接头的各下接头点的状态值。
在步骤13中，计算所述待检图像与基准图像中相对应的每一接头点的状态值的差 值，以此作为接头点抽动距离；其计算方法为：
从待检图像中接头第一层的第一个上接头点开始，计算此接头点状态值与相匹配的基准图像接头点状态值的差值，直至第四层接头的下接头点中最后一个接头点状态值差值；各差值则为各接头点的抽动距离。

Claims (10)

  1. 一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：该方法包括如下步骤：
    步骤1：将T1时刻采集到的含有完整接头的钢丝绳芯输送带X光图像，即T1时刻的原始图像作为基准图像；
    步骤2：对原始图像分别进行横向均值滤波和全局均值滤波，以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理，得到图像A；
    步骤3：对图像A进行形态学闭操作，孔洞填充、开操作，得到图像B；
    步骤4：对图像B进行连通分量提取，并根据面积特征从各连通分量中筛选出接头分层区域，对各分层区域求并集，得到图像C；
    步骤5：分别对图像A与图像C进行连通分量提取，并对两图像中的连通分量求交集，得到图像D；
    步骤6：利用最小外接矩形算子获取图像D中各接头分层区域的最小外接矩形，并取得每个矩形的方向值，即矩形上下两边的斜率，得到图像E；
    步骤7：利用每个最小外接矩形的方向值，即斜率，与此矩形区域中纵坐标最大、最小时相对应的坐标点，计算得到每个接头分层区域的下、上接头拟合直线；
    步骤8：对原始图像进行全局均值滤波，以均值滤波后的图像为参考图像，对原始图像进行动态阈值处理，并对阈值处理后的图像进行开操作，得到图像F；
    步骤9：对图像F进行形态学腐蚀以及开操作，得到图像G；
    步骤10：分别对图像E中每一个最小外接矩形与图像G进行交集运算，并对求交后图像进行连通分量提取，再根据面积特征从各连通分量中筛选出接头空档区域，得到图像H；
    步骤11：对步骤10得到的每一层接头空档区域，提取接头空档区域的重心，计算各重心到对应上下接头拟合直线的距离，进一步计算每一层接头的各上下接头点状态值，直至最后一个最小外接矩形运算完成；
    步骤12：对T2时刻采集到的含有完整接头的钢丝绳芯输送带X光图像，即T2时刻的原始图像作为待检图像，根据步骤2-11处理得到待检图像的每一层接头的各上下接头点状态值；
    步骤13：对待检图像与基准图像经过步骤2-11处理后，分别得到待检图像与基准图像的每一层接头的各上下接头点状态值，计算所述待检图像与基准图像中相对应的每一接头点的状态值的差值，该差值作为接头点抽动距离；
    步骤14：计算基准图像接头点状态值的均值，将此均值的t倍数作为阈值，判断待检图像每一接头点抽动距离值是否大于此阈值，若大于此阈值，则判定该接头点发生了抽动故障。
  2. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤2中，对原始图像分别进行横向均值滤波和全局均值滤波，以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理，步骤如下：
    步骤2.1：选取a*1大小的均值模板对原始图像进行横向均值滤波；
    步骤2.2：选取a*b大小的均值模板对原始图像进行全局均值滤波；
    步骤2.3：以全局均值滤波后的图像为参考图像，对横向均值滤波后的图像进行动态阈值处理；将横向均值滤波后的图像与参考图像中任一点进行灰度值比较，若横向均值滤波后的图像中像素点灰度值高于阈值l，则设定该像素点的灰度值为256，反之，该像素点的灰度值则为0。
  3. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤3中，对图像A进行形态学闭操作，孔洞填充、开操作，步骤如下：
    步骤3.1：选取c*d大小的矩形结构元对动态阈值分割后的图像进行形态学闭操作，即先膨胀后腐蚀；
    步骤3.2：对形态学闭操作过程中提取出来的各区域进行孔洞填充；
    步骤3.3：选取a*1大小的矩形结构元对步骤3.2得到的孔洞填充后的图像进行开操作，以此去除图像中存在的干扰。
  4. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤4中，对图像B进行连通分量提取、连通分量面积特征筛选、以及连通分量求并集，步骤如下：
    步骤4.1：对步骤3得到的图像进行连通分量提取；
    步骤4.2：根据连通分量特征，从步骤4.1提取出的各连通分量中筛选出接头分层区域；所述连通分量特征为：各连通分量面积是否在区间范围[S 1，S 2]之内，S 1和S 2分别表示预设面积值的下限和上限，若在此区间范围内，则判定此连通分量为接头分层区域，反之，则不为接头分层区域；
    步骤4.3：对步骤4.2得到的各接头分层区域求并集。
  5. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤7中，利用每个最小外接矩形的上下两边的斜率，与此矩形区域中纵坐标最大、最小时相对应的坐标点，计算得到每个接头分层区域的下、上接头拟合直线；方法如下：
    步骤7.1：设最小外接矩形中上下两边的直线y＝k ix+b i，其中，k i为斜率，b i为截距，i＝0，1，...，2M，M为接头分层数目；
    步骤7.2：设定原始图像中钢丝绳方向为Y轴方向，输送带运行的方向为Y轴正方向，垂直于钢丝绳的方向为X轴方向；首先计算出各个最小外接矩形的重心坐标，根据 各个坐标的Y值大小进行接头分层排序；
    步骤7.3：在拟合得到的第j个最小外接矩形中选取Y轴方向上Y值最大时对应的像素点坐标(x m，y max)以及Y值最小时对应的像素点坐标(x n，y min)；
    步骤7.4：将Y值最大时对应的像素点坐标(x m，y max)代入y＝k upx+b m，其中k up代表已知的第j个矩形上边的斜率，由此可以计算得到截距b m，最后求得矩形上边的直线函数，即接头拟合直线；
    步骤7.5：将Y值最小时对应的像素点坐标(x n，y min)代入y＝k downx+b n，其中k down代表已知的第j个矩形下边的斜率，在同一矩形中，k up＝k down；由此可以计算得到截距b n，最后求得矩形下边的直线函数，即接头拟合直线；
    步骤7.6：按照步骤7.3-7.5依次对每个最小外接矩形求取上下边直线函数，其中，从第一个外接矩形求得的上边直线以及最后一个外接矩形求得的下边直线在计算接头抽动距离时是无用的，将这两条直线剔除，最后得到完整接头中每一条接头分层拟合直线。
  6. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤8中，对原始图像进行全局均值滤波，以均值滤波后的图像为参考图像，对原始图像进行动态阈值处理，并对阈值处理后的图像进行开操作，步骤如下：
    步骤8.1：选取e*f大小的均值滤波模板对原始图像进行滤波处理，以滤波后的图像作为参考图像，对原始图像进行动态阈值分割，即原始图像与参考图像中任一点进行灰度值比较，若原始图像中像素点灰度值高于阈值l，则设定该像素点的灰度值为256，反之，该像素点的灰度值则为0；
    步骤8.2：选取半径为r 1的圆形结构元对阈值处理后的图像进行形态学开操作，以此去除图像的噪点以及钢丝绳边缘的毛刺。
  7. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤9中，对图像F进行形态学腐蚀以及开操作的步骤如下：
    步骤9.1：选取g*1大小的矩形结构元对步骤8得到的图像进行腐蚀，缩小相邻钢丝绳之间的间距，突出接头空档区域；
    步骤9.2：采用半径为r 2的圆形结构元再次对腐蚀后的图像进行形态学开操作处理，去除图像在处理过程中存在的噪点。
  8. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤11中，各层接头空档区域重心提取的方法如下：
    S11：设定钢丝绳方向为Y轴方向，输送带运行的方向为Y轴正方向，垂直于钢丝绳的方向为X轴方向；首先计算出各个最小外接矩形的重心坐标(x，y)，根据各重心坐标的y值大小进行接头分层排序；
    S12：将排好序的最小外接矩形逐个与图像E和图像G求交后的图像求交集，得到各行接头空档集合E i，i＝1，2，...，m，m为矩形个数；
    S13：计算每个集合中各接头空档的重心坐标(x，y)，按照重心坐标的x值大小将各坐标进行升序排序并保存。
  9. 根据权利要求1所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤11中，每一层接头的各上下接头点状态值的计算方法如下：
    S21：从第一个外接矩形开始，计算每个接头空档区域重心到上拟合直线的距离，各距离记为J i，i＝1，2，...，n，其中n为接头空档区域重心个数；计算距离的均值，表示为：
    Figure PCTCN2019094535-appb-100001
    S22：计算接头空档区域重心到上拟合直线的距离J i与均值L up的差值P i＝J i-L up，i＝1，2，...，n；
    S23：计算每个接头空档区域重心到下拟合直线的距离，各距离记为K i，i＝1，2，...，n，其中n为接头空档区域重心个数；计算距离的均值，表示为：
    Figure PCTCN2019094535-appb-100002
    S24：计算接头空档区域重心到下拟合直线距离K i与均值L down的差值Q i＝K i-L down，i＝1，2，...，n；
    S25：根据步骤S21到S24的方法分别计算出五个外接矩形各对应的P i与Q i；
    S26：剔除第一个外接矩形的P i与第五个外接矩形的Q i；将第一个外接矩形的Q i作为第一层接头的各上接头点的状态值，第二个外接矩形的P i作为第一层接头的各下接头点的状态值；将第二个外接矩形的Q i作为第二层接头的各上接头点的状态值，第三个外接矩形的P i作为第二层接头的各下接头点的状态值；将第三个外接矩形的Q i作为第三层接头的各上接头点的状态值，第四个外接矩形的P i作为第三层接头的各下接头点的状态值；将第四个外接矩形的Q i作为第四层接头的各上接头点的状态值，第五个外接矩形的P i作为第四层接头的各下接头点的状态值。
  10. 根据权利要求1-9任一所述的一种基于X光图像的钢丝绳芯输送带接头抽动检测方法，其特征在于：在步骤13中，计算所述待检图像与基准图像中相对应的每一接头点的状态值的差值，以此作为接头点抽动距离；其计算方法为：
    从待检图像中第一层接头的第一个上接头点开始，计算此接头点状态值与相匹配的基准图像接头点状态值的差值，直至最后一层接头的下接头点中最后一个接头点状态值 差值；各差值则为各接头点的抽动距离。
PCT/CN2019/094535 2019-05-16 2019-07-03 一种基于x光图像的钢丝绳芯输送带接头抽动检测方法 WO2020228111A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910408860.1 2019-05-16
CN201910408860.1A CN110288562B (zh) 2019-05-16 2019-05-16 一种基于x光图像的钢丝绳芯输送带接头抽动检测方法

Publications (1)

Publication Number Publication Date
WO2020228111A1 true WO2020228111A1 (zh) 2020-11-19

Family

ID=68002203

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/094535 WO2020228111A1 (zh) 2019-05-16 2019-07-03 一种基于x光图像的钢丝绳芯输送带接头抽动检测方法

Country Status (2)

Country Link
CN (1) CN110288562B (zh)
WO (1) WO2020228111A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113553926A (zh) * 2021-07-08 2021-10-26 天津工业大学 一种基于图像匹配的钢丝绳芯输送带接头抽动自动检测方法
CN113570551A (zh) * 2021-07-05 2021-10-29 天津市三特电子有限公司 球团链篦机篦条状态监测方法及其监测系统
CN113989643A (zh) * 2021-10-26 2022-01-28 萱闱(北京)生物科技有限公司 管道状态检测方法、装置、介质及计算设备
CN113983966A (zh) * 2021-10-27 2022-01-28 北京半导体专用设备研究所（中国电子科技集团公司第四十五研究所） 一种芯片平面度测量方法、装置、电子设备及存储介质
CN116416251A (zh) * 2023-06-12 2023-07-11 梁山水泊胶带股份有限公司 基于图像处理的整芯阻燃输送带质量检测方法及相关装置
CN116563223A (zh) * 2023-04-11 2023-08-08 新创碳谷集团有限公司 一种玻纤纱线绕罗拉检测方法、设备及存储介质
CN117110330A (zh) * 2023-10-25 2023-11-24 山西慧达澳星科技有限公司 一种输送带探伤检测方法、装置、设备及存储介质
CN117218549A (zh) * 2023-11-07 2023-12-12 山东力加力钢结构有限公司 一种公路桥梁的状态评估方法及系统

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004028891A (ja) * 2002-06-27 2004-01-29 Anritsu Sanki System Co Ltd Ｘ線画像処理フィルタ自動設定方法，ｘ線異物検出方法及びｘ線異物検出装置
CN102590247A (zh) * 2012-03-05 2012-07-18 上海海事大学 一种基于x光图像处理技术的钢丝绳损伤检测方法
CN103995011A (zh) * 2014-05-27 2014-08-20 天津工业大学 一种提高钢丝绳芯输送带x光图像质量的方法
CN104730083A (zh) * 2015-04-09 2015-06-24 中煤科工集团重庆研究院有限公司 钢丝绳芯输送带接头抽动自动检测方法
CN105976368A (zh) * 2016-04-28 2016-09-28 华北电力大学（保定） 一种绝缘子定位方法
CN106296700A (zh) * 2016-08-15 2017-01-04 南京工程学院 一种钢丝绳输送带接头抽动检测方法
CN108169253A (zh) * 2018-02-12 2018-06-15 无锡百年通工业输送有限公司 一种x射线探测芳纶输送带指型接头抽动的方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2013099520A (ja) * 2011-10-14 2013-05-23 Toshiba Corp Ｘ線コンピュータ断層撮像装置、医用画像処理装置、及び医用画像処理方法
CN104698944B (zh) * 2015-03-19 2017-06-27 天津工业大学 一种基于fpga的钢丝绳芯输送带x光探测器采集控制与信号处理装置
CN108169332A (zh) * 2017-12-14 2018-06-15 太原理工大学 基于热声成像的钢丝绳芯检测方法和装置
CN108550135B (zh) * 2018-03-07 2021-07-30 天津工业大学 一种基于x光图像的钢丝绳芯输送带接头伸长自动检测方法
CN108470333B (zh) * 2018-03-07 2021-10-08 天津工业大学 一种x光钢丝绳芯输送带接头图像筛选与拼接的方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004028891A (ja) * 2002-06-27 2004-01-29 Anritsu Sanki System Co Ltd Ｘ線画像処理フィルタ自動設定方法，ｘ線異物検出方法及びｘ線異物検出装置
CN102590247A (zh) * 2012-03-05 2012-07-18 上海海事大学 一种基于x光图像处理技术的钢丝绳损伤检测方法
CN103995011A (zh) * 2014-05-27 2014-08-20 天津工业大学 一种提高钢丝绳芯输送带x光图像质量的方法
CN104730083A (zh) * 2015-04-09 2015-06-24 中煤科工集团重庆研究院有限公司 钢丝绳芯输送带接头抽动自动检测方法
CN105976368A (zh) * 2016-04-28 2016-09-28 华北电力大学（保定） 一种绝缘子定位方法
CN106296700A (zh) * 2016-08-15 2017-01-04 南京工程学院 一种钢丝绳输送带接头抽动检测方法
CN108169253A (zh) * 2018-02-12 2018-06-15 无锡百年通工业输送有限公司 一种x射线探测芳纶输送带指型接头抽动的方法

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113570551B (zh) * 2021-07-05 2023-07-07 天津市三特电子有限公司 球团链篦机篦条状态监测方法及其监测系统
CN113570551A (zh) * 2021-07-05 2021-10-29 天津市三特电子有限公司 球团链篦机篦条状态监测方法及其监测系统
CN113553926B (zh) * 2021-07-08 2023-08-11 天津工业大学 一种基于图像匹配的钢丝绳芯输送带接头抽动自动检测方法
CN113553926A (zh) * 2021-07-08 2021-10-26 天津工业大学 一种基于图像匹配的钢丝绳芯输送带接头抽动自动检测方法
CN113989643B (zh) * 2021-10-26 2023-09-01 萱闱(北京)生物科技有限公司 管道状态检测方法、装置、介质及计算设备
CN113989643A (zh) * 2021-10-26 2022-01-28 萱闱(北京)生物科技有限公司 管道状态检测方法、装置、介质及计算设备
CN113983966B (zh) * 2021-10-27 2023-07-21 北京半导体专用设备研究所（中国电子科技集团公司第四十五研究所） 一种芯片平面度测量方法、装置、电子设备及存储介质
CN113983966A (zh) * 2021-10-27 2022-01-28 北京半导体专用设备研究所（中国电子科技集团公司第四十五研究所） 一种芯片平面度测量方法、装置、电子设备及存储介质
CN116563223B (zh) * 2023-04-11 2023-09-26 新创碳谷集团有限公司 一种玻纤纱线绕罗拉检测方法、设备及存储介质
CN116563223A (zh) * 2023-04-11 2023-08-08 新创碳谷集团有限公司 一种玻纤纱线绕罗拉检测方法、设备及存储介质
CN116416251B (zh) * 2023-06-12 2023-08-25 梁山水泊胶带股份有限公司 基于图像处理的整芯阻燃输送带质量检测方法及相关装置
CN116416251A (zh) * 2023-06-12 2023-07-11 梁山水泊胶带股份有限公司 基于图像处理的整芯阻燃输送带质量检测方法及相关装置
CN117110330A (zh) * 2023-10-25 2023-11-24 山西慧达澳星科技有限公司 一种输送带探伤检测方法、装置、设备及存储介质
CN117110330B (zh) * 2023-10-25 2024-01-30 山西慧达澳星科技有限公司 一种输送带探伤检测方法、装置、设备及存储介质
CN117218549A (zh) * 2023-11-07 2023-12-12 山东力加力钢结构有限公司 一种公路桥梁的状态评估方法及系统
CN117218549B (zh) * 2023-11-07 2024-03-05 山东力加力钢结构有限公司 一种公路桥梁的状态评估方法及系统

Also Published As

Publication number Publication date
CN110288562B (zh) 2023-01-17
CN110288562A (zh) 2019-09-27

Similar Documents

Publication Publication Date Title
WO2020228111A1 (zh) 一种基于x光图像的钢丝绳芯输送带接头抽动检测方法
CN102175700B (zh) 数字x射线图像焊缝分割和缺陷检测方法
DE102012107287B4 (de) Computergestütztes Verfahren und Vorrichtung zum Identifizieren von Fehlern in einem Material
CN106290388B (zh) 一种绝缘子故障自动检测方法
WO2017063355A1 (zh) 超声相控阵检测聚乙烯电熔接头缺陷类型自动识别方法
US20010028732A1 (en) Method and system for detecting defects on a printed circuit board
CN113409245B (zh) 一种电子元器件x射线检查缺陷自动识别方法
US9996766B2 (en) Imaging-based methods for detecting and measuring defects in extruded cellular ceramic articles
CN109934808B (zh) 一种基于图像多形状法向梯度差值pcb板缺陷分类方法
CN108550135B (zh) 一种基于x光图像的钢丝绳芯输送带接头伸长自动检测方法
CN111932490A (zh) 一种用于工业机器人的视觉系统抓取信息提取方法
CN109829910A (zh) 一种基于邻域查找的pcb缺陷检测方法
CN114897908A (zh) 基于机器视觉的激光选区铺粉烧结面缺陷分析方法及系统
CN114663672A (zh) 一种输电线路塔钢构件腐蚀检测方法及系统
CN107220946B (zh) 一种岩石运输带上不良块度图像的实时剔除方法
CN116258682A (zh) 基于PSPNet和改进YOLOv7的PCB锡膏缺陷检测方法
CN108492306A (zh) 一种基于图像轮廓的x型角点提取方法
CN109767426B (zh) 一种基于图像特征识别的盾构隧道渗漏水检测方法
CN116993804B (zh) 一种基于lsm算法的箍筋尺寸检测方法及系统
CN113793384A (zh) 一种基于图像处理的石油钻杆变径定位方法及系统
TWI519802B (zh) 一種線路資訊擷取方法、非線路區資訊擷取方法及運用前述方法的線路缺陷檢測方法
Chen et al. Surface Defect Detection of Cable Based on Threshold Image Difference
CN115953387A (zh) 一种基于深度学习的射线图像焊缝缺陷检测方法
KR102585146B1 (ko) 딥러닝 기반의 도로손상 탐지 장치 및 방법
CN110363783B (zh) 基于Canny算子的岩体结构面迹线半自动检测方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19928700

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19928700

Country of ref document: EP

Kind code of ref document: A1