WO2021227285A1 - 蜂窝结构几何规整度图像识别方法及系统 - Google Patents

蜂窝结构几何规整度图像识别方法及系统 Download PDF

Info

Publication number
WO2021227285A1
WO2021227285A1 PCT/CN2020/109726 CN2020109726W WO2021227285A1 WO 2021227285 A1 WO2021227285 A1 WO 2021227285A1 CN 2020109726 W CN2020109726 W CN 2020109726W WO 2021227285 A1 WO2021227285 A1 WO 2021227285A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel
value
image
window
cell
Prior art date
Application number
PCT/CN2020/109726
Other languages
English (en)
French (fr)
Inventor
王中钢
梁习峰
施冲
周伟
崔灿
熊伟
王鑫鑫
Original Assignee
中南大学
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
Priority claimed from CN202010388489.XA external-priority patent/CN111583242B/zh
Priority claimed from CN202010388353.9A external-priority patent/CN111583239B/zh
Application filed by 中南大学 filed Critical 中南大学
Priority to US17/422,720 priority Critical patent/US11893768B2/en
Publication of WO2021227285A1 publication Critical patent/WO2021227285A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/698—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03B—APPARATUS OR ARRANGEMENTS FOR TAKING PHOTOGRAPHS OR FOR PROJECTING OR VIEWING THEM; APPARATUS OR ARRANGEMENTS EMPLOYING ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ACCESSORIES THEREFOR
    • G03B17/00—Details of cameras or camera bodies; Accessories therefor
    • G03B17/56—Accessories
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/001—Industrial image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—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
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/457—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by analysing connectivity, e.g. edge linking, connected component analysis or slices
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60—Control of cameras or camera modules
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60—Control of cameras or camera modules
    • H04N23/695—Control of camera direction for changing a field of view, e.g. pan, tilt or based on tracking of objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20036—Morphological image processing
    • 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
    • G06T2207/20036—Morphological image processing
    • G06T2207/20044—Skeletonization; Medial axis transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component

Definitions

  • the invention relates to the fields of design, manufacture and application of lightweight structural products for transportation, machinery, aerospace, ship and other equipment, and in particular to an image recognition method and system for the geometric regularity of a honeycomb structure.
  • Lightweight honeycomb products are widely used because of their excellent carrying capacity and good energy absorption characteristics.
  • light honeycomb products are widely used in high-speed trains.
  • honeycomb products are prone to cell deformation during production, transportation and use, and this deformation will have an important impact on the performance of the honeycomb product. Therefore, it is necessary to evaluate the cell deformation of the honeycomb product, that is, the geometric regularity, so as to make a judgment on the quality of the honeycomb product.
  • the Chinese patent application number is 201910503383.7 and its filing date is September 13, 2019. It discloses a method for identifying the edges of honeycombs from the surface measurement data of honeycomb cores. It includes the following steps: transform the collected three-dimensional data on the surface of the honeycomb core into two-dimensional coordinates, and identify the corner points in the two-dimensional plane projection image of the honeycomb core through the corner detection algorithm.
  • the corner points include human-shaped vertices, Y-shaped vertices, Pseudo vertices, including unrecognized missing corner points; based on the proposed corner point type judgment algorithm, the two endpoints of the honeycomb edge are identified in sequence.
  • One of the endpoints is determined when the adjacent edges are identified, and the other endpoint is passed.
  • the local analysis of the recognition vertex is determined; during the realization of the honeycomb edge, the false vertices can be effectively eliminated, and the missing vertices can be supplemented at the same time, so as to realize stable and high-precision honeycomb edge recognition.
  • the geometric regularity of the honeycomb product can be evaluated, and the quality of the honeycomb product can be judged accordingly. This method has the advantages of high accuracy and good robustness, but it needs to scan the surface of the honeycomb core point by point, which is time-consuming and long and the steps are cumbersome.
  • the purpose of the present invention is to overcome the shortcomings of the prior art, and provide a method and system for recognizing the geometric regularity of the honeycomb structure with high work efficiency and high analysis accuracy.
  • An image recognition method for geometric regularity of a honeycomb structure includes the following steps: image acquisition, image processing, vertex extraction, cell reconstruction, and quality evaluation; the step “acquiring an image” includes capturing an image and reading an image by a computer The step “extracting vertices” is to find and record the vertices of the cells on the basis of “image processing”; the step “reconstructing the cells” is to connect the extracted vertices according to the mapping relationship between the cells and the vertices, Obtain a cell reconstruction map;
  • the step “binarization” is set between the step “image processing” and the step “vertex extraction”.
  • the step “binarization” is to set the pixel value of the background in the image to 0 and set the pixels of the honeycomb skeleton in the image Set the value to 1 to form a binarized graph;
  • the step "quality evaluation” is based on the cell reconstruction graph, calculates the angular deviation value and its average value, line deviation value and its average value of all cells, and compares it with the set tolerance zone to determine whether it is qualified .
  • step “vertex extraction” is based on the “binarized graph", and the “binarized graph” is subjected to closed operation processing to obtain a smooth honeycomb vertex image;
  • the processing of taking the largest circle center of the honeycomb wall is performed on the image of the intersection of only the honeycomb walls to obtain the apex of the honeycomb cell.
  • step “vertex extraction” is based on the “binarized graph", which sequentially executes the determination of wall thickness, pixel assignment, determination of cell side length and pixel annihilation;
  • the pixel point is determined as the edge vertex and recorded; if the pixel value The number of changes is 6, and it has been shown that there are three straight lines at the pixel point, then the pixel point is determined to be the middle vertex and recorded.
  • the Harris algorithm is used to calculate the corner response function value R corresponding to the center point of the window, and then one hundredth of the maximum value of the corner response function value R is the limit value, and all corner points in the same window
  • the first step of the "reconstructing cell” is to arbitrarily select x vertices at the center of the "binarized graph” as reference points, find the closest vertex at each reference point as the neighboring point, and calculate Find the distance between the reference point and the adjacent point, keep the three records with the shortest distance, and then find the average of the distances of the 3x records as the cell side length A;
  • the second step is to divide the obtained vertices into active connection points and passive There are two types of connection points; namely: first divide the skeleton graph into two areas, the area around the skeleton graph with a width of 1A ⁇ 2A is the edge area, the area surrounded by the edge area is the central area; the vertex located in the central area is Active connection points, the vertices located in the edge area are passive connection points;
  • the third step is to connect all active connection points with the three closest vertices with line segments, that is: each active connection point must and can only retain the three shortest ones Line segment, after the connection is completed, the
  • a system suitable for an image recognition method of geometric regularity of a honeycomb structure including a detection station, a digital camera, and a computer; the digital camera and the computer are electrically connected;
  • the detection platform is a movable working platform, including a storage table, a lifting device, and a clamp.
  • the lifting device is installed at the bottom of the storage table; the tested honeycomb is placed on the storage table; the lifting device includes a vertical guide rail and an electric push rod Or electro-hydraulic push rod, which can push the table to move up and down along the vertical guide rail, adjust the height of the honeycomb under test to ensure that the upper end surface of the honeycomb under test is flush with the upper end surface of the fixture; the control part of the lifting device is connected to the computer electronics connect;
  • the clamp is composed of four flat plates and a driving device, which can be moved closer to the honeycomb part to be tested under the action of the driving device, and is locked after being close to the honeycomb part to be tested, and is used for positioning and fixing the honeycomb part to be tested.
  • the system adds a walking gantry, a sliding rail, and a moving device;
  • the digital camera is installed on the beam of the walking gantry, which can move laterally along the beam under the drive of the mobile device;
  • the walking gantry can move longitudinally along the slide rail under the drive of the mobile device, and the movement of the digital camera and the walking gantry are controlled by the computer.
  • the computer includes:
  • the vertex extraction module is used to find and record the vertices of the cell
  • the cell reconstruction module is used to connect the extracted vertices according to the mapping relationship between the cells and the vertices to obtain the cell reconstruction graph;
  • the quality evaluation module is used to calculate the angular deviation value and average value of all cells and the linear deviation value and average value based on the cell reconstruction map, and then compare it with the set tolerance zone to determine whether it is qualified.
  • the computer also includes:
  • the binarization module is used to set the pixel value of the background in the image to 0, and set the pixel value of the honeycomb skeleton in the image to 1, to form a binarized map.
  • the vertex extraction module includes:
  • the cell side length unit which is used to find the pixel point with the largest value after completing the "pixel assignment", and create a square area with this point as the center.
  • the initial side length E wall thickness L, and four areas are calculated
  • the side length A of the honeycomb cell element can be obtained by solving the two coordinate values;
  • the pixel annihilation unit is used to find the pixel with the maximum value, and determine the pixel as the vertex and record it, and then use the pixel as the center to establish a square annihilation window with the side length A of the honeycomb cell as the side length.
  • the pixel with the largest assignment is found in the remaining assignments, determined as the vertex and recorded, and the operation of the annihilation window is repeated until the pixel
  • the reconstructed cell module is specifically used to arbitrarily select x vertices as reference points in the central position of the "binarized graph", find the closest vertex at each reference point as the adjacent point, and calculate the reference point The distance to the adjacent point, keep the three records with the shortest distance, and then calculate the average of the distance of 3x records as the cell side length A; then divide the obtained vertices into two categories: active connection points and passive connection points; That is: first divide the skeleton graph into two areas, the area with a width of 1A ⁇ 2A around the skeleton graph is the edge area, and the area surrounded by the edge area is the central area; the vertex located in the central area is the active connection point, located at the edge The vertices of the zone are passive connection points; finally, all active connection points and the closest three vertices are connected by line segments, that is: each active connection point must and can only retain the three shortest line segments, and the cell is formed after the connection is completed Refactor the diagram.
  • the reconstructed cell module includes:
  • the edge expansion unit is used on the basis of the morphological image. On the outermost edges of the four sides of the morphological image, all extend outward by at least 1 pixel to form an extended area. The pixel values of all pixels in the extended area are all set to 1. Get extended image;
  • the vertex connection unit is used to take the extended image as the object, and traverse the extended image in the order from left to right and top to bottom.
  • An electronic device including a memory, a display, a processor, and a computer program stored on the memory and capable of running on the processor.
  • the processor implements any one of the above technical solutions when the program is executed. The steps of the method.
  • the method and system of the present invention are scientific and reasonable. Simple and easy to implement, high precision, high work efficiency and other advantages.
  • Figure 1 is a flow chart of the method of the present invention
  • FIG. 2 is a schematic diagram of vertex extraction and cell reconstruction according to an embodiment of the method of the present invention
  • Figure 3 is a schematic diagram of the device configuration of an embodiment of the system of the present invention.
  • Fig. 4 is a top view of Fig. 3.
  • An image recognition method for geometric regularity of a honeycomb structure includes the following steps: image acquisition, image processing, vertex extraction, cell reconstruction, and quality evaluation; the step “acquiring an image” includes capturing an image and reading an image by a computer The step “extracting vertices” is to find and record the vertices of the cells on the basis of “image processing”; the step “reconstructing the cells” is to connect the extracted vertices according to the mapping relationship between the cells and the vertices, Obtain a cell reconstruction map;
  • the step “binarization” is set between the step “image processing” and the step “vertex extraction”.
  • the step “binarization” is to set the pixel value of the background in the image to 0 and set the pixels of the honeycomb skeleton in the image Set the value to 1 to form a binarized graph;
  • the step "quality evaluation” is based on the cell reconstruction graph, calculates the angular deviation value and its average value, line deviation value and its average value of all cells, and compares it with the set tolerance zone to determine whether it is qualified .
  • step “vertex extraction” is based on the “binarized graph", and the “binarized graph” is subjected to closed operation processing to obtain a smooth honeycomb vertex image;
  • the processing of taking the largest circle center of the honeycomb wall is performed on the image of the intersection of only the honeycomb walls to obtain the apex of the honeycomb cell.
  • step “vertex extraction” is based on the “binarized graph", which sequentially executes the determination of wall thickness, pixel assignment, determination of cell side length and pixel annihilation;
  • the pixel point is determined as the edge vertex and recorded; if the pixel value The number of changes is 6, and it has been shown that there are three straight lines at the pixel point, then the pixel point is determined to be the middle vertex and recorded.
  • the Harris algorithm is used to calculate the corner response function value R corresponding to the center point of the window, and then one hundredth of the maximum value of the corner response function value R is the limit value, and all corner points in the same window
  • the first step of the "reconstructing cell” is to arbitrarily select x vertices at the center of the "binarized graph” as reference points, find the closest vertex at each reference point as the neighboring point, and calculate Find the distance between the reference point and the adjacent point, keep the three records with the shortest distance, and then find the average of the distances of the 3x records as the cell side length A;
  • the second step is to divide the obtained vertices into active connection points and passive There are two types of connection points; namely: first divide the skeleton graph into two areas, the area around the skeleton graph with a width of 1A ⁇ 2A is the edge area, the area surrounded by the edge area is the central area; the vertex located in the central area is Active connection points, the vertices located in the edge area are passive connection points;
  • the third step is to connect all active connection points with the three closest vertices with line segments, that is: each active connection point must and can only retain the three shortest ones Line segment, after the connection is completed, the
  • a system suitable for an image recognition method of geometric regularity of a honeycomb structure comprising a detection station 1, a digital camera 2 and a computer 3; the digital camera 2 and the computer 3 are electrically connected;
  • the detection platform 1 is a movable working platform, including a storage table 4, a lifting device 5, and a clamp 6.
  • the lifting device 5 is installed at the bottom of the storage table 4; the honeycomb to be tested is placed on the storage table 4; the lifting device 5 Including vertical rails, electric push rods or electro-hydraulic push rods, which can push the table 4 to move up and down along the vertical rails and adjust the height of the honeycomb under test to ensure that the upper end surface of the honeycomb under test is flat with the upper end surface of the fixture 6 Qi; the control part of the lifting device 5 is electrically connected to the computer 3;
  • the clamp 6 is composed of four flat plates and a driving device, which can move closer to the honeycomb part to be tested under the action of the driving device, and lock the honeycomb part to be tested after being close to the honeycomb part to be tested for positioning and fixing the honeycomb part to be tested.
  • the system adds a walking gantry 7, a sliding rail 8, and a mobile device 9;
  • the digital camera 2 is installed on the beam of the walking gantry 7, and can move laterally along the beam under the drive of the moving device 9;
  • the walking gantry 7 can move longitudinally along the slide rail 8 under the drive of the moving device 9, and the movement of the digital camera 2 and the walking gantry 7 are controlled by the computer 3.
  • the computer 3 includes:
  • the vertex extraction module is used to find and record the vertices of the cell
  • the cell reconstruction module is used to connect the extracted vertices according to the mapping relationship between the cells and the vertices to obtain the cell reconstruction graph;
  • the quality evaluation module is used to calculate the angular deviation value and average value of all cells and the linear deviation value and average value based on the cell reconstruction map, and then compare it with the set tolerance zone to determine whether it is qualified.
  • the computer 3 also includes:
  • the binarization module is used to set the pixel value of the background in the image to 0, and set the pixel value of the honeycomb skeleton in the image to 1, to form a binarized map.
  • the vertex extraction module includes:
  • the cell side length unit which is used to find the pixel point with the largest value after completing the "pixel assignment", and create a square area with this point as the center.
  • the initial side length E wall thickness L, and four areas are calculated
  • the side length A of the honeycomb cell element can be obtained by solving the two coordinate values;
  • the pixel annihilation unit is used to find the pixel with the maximum value, and determine the pixel as the vertex and record it, and then use the pixel as the center to establish a square annihilation window with the side length A of the honeycomb cell as the side length.
  • the pixel with the largest assignment is found in the remaining assignments, determined as the vertex and recorded, and the operation of the annihilation window is repeated until the pixel
  • the reconstructed cell module is specifically used to arbitrarily select x vertices as reference points in the central position of the "binarized graph", find the closest vertex at each reference point as the adjacent point, and calculate the reference point The distance to the adjacent point, keep the three records with the shortest distance, and then calculate the average of the distance of 3x records as the cell side length A; then divide the obtained vertices into two categories: active connection points and passive connection points; That is: first divide the skeleton graph into two areas, the area with a width of 1A ⁇ 2A around the skeleton graph is the edge area, and the area surrounded by the edge area is the central area; the vertex located in the central area is the active connection point, located at the edge The vertices of the zone are passive connection points; finally, all active connection points and the closest three vertices are connected by line segments, that is: each active connection point must and can only retain the three shortest line segments, and the cell is formed after the connection is completed Refactor the diagram.
  • the reconstructed cell module includes:
  • the edge expansion unit is used on the basis of the morphological image. On the outermost edges of the four sides of the morphological image, all extend outward by at least 1 pixel to form an extended area. The pixel values of all pixels in the extended area are all set to 1. Get extended image;
  • the vertex connection unit is used to take the extended image as the object, and traverse the extended image in the order from left to right and top to bottom.
  • a storage medium in which a computer program is stored, and when the program is executed by a processor, the steps of the method described in any one of the above embodiments are implemented.
  • An electronic device including a memory, a display, a processor, and a computer program stored on the memory and capable of running on the processor.
  • the processor implements any of the above-mentioned embodiments when the program is executed. The steps of the method.
  • An image recognition method for geometric regularity of a honeycomb structure includes the following steps: image acquisition, image processing, vertex extraction, cell reconstruction, and quality evaluation; the step “acquiring an image” includes capturing an image and reading an image by a computer The step “extracting vertices” is to find and record the vertices of the cells on the basis of “image processing”; the step “reconstructing the cells” is to connect the extracted vertices according to the mapping relationship between the cells and the vertices, Obtain a cell reconstruction map;
  • the step “binarization” is set between the step “image processing” and the step “vertex extraction”.
  • the step “binarization” is to set the pixel value of the background in the image to 0 and set the pixels of the honeycomb skeleton in the image Set the value to 1 to form a binarized graph;
  • the step "quality evaluation” is based on the cell reconstruction graph, calculates the angular deviation value and its average value, line deviation value and its average value of all cells, and compares it with the set tolerance zone to determine whether it is qualified .
  • Step “Vertex Extraction” is based on the “binarized graph”, which sequentially executes the determination of wall thickness, pixel assignment, determination of cell side length and pixel annihilation ;
  • the first step of the "vertex extraction” is based on the morphological image, and the line with the pixel value of 1 is drawn with the line segment with the line width of 1 pixel.
  • the eight neighborhoods of the pixel are searched for a week to get a pixel value change times; if the pixel value change times is 4, it shows that there are two straight lines at the pixel point, and when the two straight lines have a reasonable angle through the coordinate calculation ,
  • the pixel is determined to be an edge vertex and recorded; if the number of pixel value changes is 6, and three straight lines have been displayed for the pixel, then the pixel is determined to be an intermediate vertex and recorded.
  • the first step of the "vertex extraction” is based on the "binarized graph", and the line with the pixel value of 1 is used as a line width.
  • One part is the limit value.
  • set the corner response function value R less than the limit value to zero, and repeat the above operation to traverse the entire skeleton diagram; in the next step, use the pixel value 1 and the pixel point with the corner response function value R greater than zero is the center point, and a square window with a size of 3 ⁇ 3 pixels is established. If the R value at the center point of the window is the maximum value in this window, record the point For the vertex, repeat the above operation to traverse the entire skeleton graph.
  • the first step of "reconstructing cells” is to arbitrarily select x vertices as reference points in the center of the "binarized graph", Find the closest vertex at each reference point as the adjacent point, calculate the distance between the reference point and the adjacent point, keep the three records with the shortest distance, and then find the average of the distances of 3x records as the cell edge Length A;
  • the second step is to divide the obtained vertices into active connection points and passive connection points; that is, first divide the skeleton graph into two areas, and use the area around the skeleton graph with a width of 1A ⁇ 2A as the edge Area, the area surrounded by the edge area is the central area; the vertex located in the central area is the active connection point, and the vertex located in the edge area is the passive connection point;
  • the third step is to use line segments between all active connection points and the three closest vertices Connected, that is: each active connection point must and can only retain the three
  • the step “reconstructing cells” includes edge expansion and vertex connection;
  • the step “edge expansion” is based on the morphological image.
  • the outermost edges of the four sides of the morphological image all extend outward by at least 1 pixel to form an extended area.
  • the pixel values of all pixels in the extended area are all set to 1, to obtain an extended image;
  • the step "vertex connection” It takes the extended image as the object and traverses the extended image in the order from left to right and top to bottom.
  • the Moore neighborhood tracking algorithm is used to find and record the same cell.
  • the vertices and their connection sequence are recorded under the name of the cell.
  • each cell retains a maximum of six vertices.
  • a system suitable for an image recognition method of geometric regularity of a honeycomb structure comprising a detection station 1, a digital camera 2 and a computer 3; the digital camera 2 and the computer 3 are electrically connected;
  • the inspection platform 1 is a movable working platform, including a storage table 4, a lifting device 5, and a clamp 6.
  • the lifting device 5 is installed at the bottom of the storage platform 4;
  • the honeycomb part to be tested is placed on the table 4;
  • the lifting device 5 includes a vertical guide rail, an electric push rod or an electro-hydraulic push rod, which can push the table 4 to move up and down along the vertical guide rail to adjust the height of the honeycomb part to be tested.
  • the control part of the lifting device 5 is electrically connected to the computer 3;
  • the clamp 6 is composed of four flat plates and a driving device, which can move closer to the honeycomb part to be tested under the action of the driving device, and lock the honeycomb part to be tested after being close to the honeycomb part to be tested for positioning and fixing the honeycomb part to be tested.
  • System Embodiments 1, 2 They are basically the same as “System Embodiments 1, 2", except that: when the digital camera 2 is installed in a mobile type, the system adds a walking gantry 7, a slide rail 8, and a mobile device 9;
  • the digital camera 2 is installed on the beam of the walking gantry 7, and can move laterally along the beam under the drive of the moving device 9;
  • the walking gantry 7 can move longitudinally along the slide rail 8 under the drive of the moving device 9, and the movement of the digital camera 2 and the walking gantry 7 are controlled by the computer 3.
  • the vertex extraction and cell reconstruction results of an embodiment of the method of the present invention are shown in Figure 2.
  • the calculation results are: the maximum value of the internal angle deviation is 14.42, the average value of the internal angle deviation is 2.62, and the standard deviation of the internal angle deviation is 2.42 , Are within the set value range, and the product quality is judged to be qualified.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

一种蜂窝结构几何规整度图像识别方法及系统，所述方法包括获取图像、图像处理、顶点提取、重构胞元、质量评估；步骤"图像处理"和"顶点提取"之间设置步骤"二值化"，"二值化"是将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图；步骤"质量评估"是以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再较判定是否合格。所述系统包括检测台、数码相机和计算机；数码相机和计算机电连接；数码相机至少为一台，其分辨率不低于1080P，配置远心镜头，其安装方式为固定式或/和移动式。所述方法及其系统具有科学合理，简单易行，检测精度高，工作效率高等优点。

Description

蜂窝结构几何规整度图像识别方法及系统 技术领域
本发明涉及交通、机械、航空航天、船舶等装备的轻质结构产品设计、制造及应用等领域，特别是涉及一种蜂窝结构几何规整度图像识别方法及系统。
背景技术
轻质蜂窝产品以其优异的承载能力和良好的吸能特性而被广泛应用，比如，高速列车也广泛采用了轻质蜂窝产品。但蜂窝产品在生产、运输及使用过程中易产生胞孔的变形，而该变形会对蜂窝产品的性能产生重要的影响。因此需要对蜂窝产品的胞孔变形即几何规整度进行评估，从而对蜂窝产品的质量作出判断。
中国专利申请号为201910503383.7，其申请日为2019年9月13日，公开了一种从蜂窝芯表面测量数据中识别蜂窝边的方法。包括如下步骤：将采集到的蜂窝芯表面三维数据进行二维坐标变换，通过角点检测算法识别蜂窝芯的二维平面投影图像中的角点，所述角点包括人形顶点、Y形顶点、伪顶点，还包括未识别出的缺失角点；基于所提出的角点类型判断算法，对蜂窝边的两个端点依次识别，其中一个端点是在相邻边识别时确定，另一个端点是通过对该识别顶点的局部分析确定；在蜂窝边的实现过程中能有效地对伪顶点进行排除，同时补充缺失的顶点，实现稳定高精度的蜂窝边识别。识别蜂窝边后可以对蜂窝产品的几何规整度做出评估，并依此对蜂窝产品的质量做出判断。 该方法具有精度高、鲁棒性好的优点，但该方法需要对蜂窝芯表面做逐点扫描，费时较长，步骤繁琐。
发明内容
本发明的目的在于克服现有技术的不足，提供一种工作效率高，分析精度高的蜂窝结构几何规整度图像识别方法及系统。
本发明的目的通过以下技术方案予以实现：
方法：
一种蜂窝结构几何规整度图像识别方法，所述方法包括以下步骤：获取图像、图像处理、顶点提取、重构胞元、质量评估；所述步骤“获取图像”包括拍摄图像和计算机读取图像；所述步骤“顶点提取”是在“图像处理”的基础上寻找胞元的顶点并记录；所述步骤“重构胞元”是将提取的顶点依据胞元与顶点的映射关系连线，得到胞元重构图；
所述步骤“图像处理”和步骤“顶点提取”之间设置步骤“二值化”，所述步骤“二值化”是将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图；
所述步骤“质量评估”是以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再与设置的公差带相比较判定是否合格。
所述步骤“顶点提取”是以“二值化图”为基础，将所述“二值化图”进行闭运算处理，得到平滑蜂窝顶点图像；
将所述平滑蜂窝顶点图像依次经过膨胀处理和腐蚀处理，得到只有蜂窝壁交汇处图像；
在所述只有蜂窝壁交汇处图像上进行蜂窝壁取最大圆圆心处理，得到蜂窝胞元的顶点。
所述步骤“顶点提取”是以“二值化图”为基础，顺序执行确定壁厚、像素赋值、确定胞元边长和像素湮灭；
所述步骤“确定壁厚”是设置一个边长能够从小变大变化的正方形窗口，当某一边长的窗口遍历形态图像后，如果窗口内的像素值＝0的像素数的最小值为非零时，将该正方形窗口的边长定义为壁厚L；
所述步骤“像素赋值”是设置一个边长为壁厚L的正方形的赋值窗口，用赋值窗口遍历形态图像中像素值＝1的像素点，然后将该窗口内的像素值＝1的像素个数的总和赋值给该赋值窗口的中心点的像素点上；
所述步骤“确定胞元边长”是在完成“像素赋值”之后，找到最大赋值的像素点，以该点为中心建立一个正方形区域，其初始边长E＝壁厚L，计算出该区域四个边界上各像素点的赋值与该中心点的赋值的差值绝对值Z，遍历整个形态图像，记录本轮中最小的差值绝对值Z及其坐标；以E＝E+2个像素建立新的正方形区域，重复上述过程，直到最小的差值绝对值Z有明显的反向增大的趋势为止，以此时取出最小的差值绝对值Z的像素点坐标与对应的区域中心点的坐标，通过两者的坐标值求解得到蜂窝胞元边长A；
所述步骤“像素湮灭”是找到最大赋值的像素点，并确定该像素点为顶点并记录，再以该像素点为中心，建立一个以蜂窝胞元边长A为边长的正方形的湮灭窗口，将湮灭窗口内所有像素值＝1的像素点 上的赋值全部清零，在此基础上，再在剩余的赋值中再找到最大赋值的像素点确定为顶点并记录，重复湮灭窗口的操作，直到像素值＝1的像素点上的赋值小于给定阈值为止，顶点提取完毕。
所述步骤“顶点提取”的第一步是以形态图像为基础，将像素值为1的线条采用线宽为1个像素的线段绘制成骨架图；第二步是以具有k个像素点的骨架图为基础，遍历像素点1至像素点k，每当遇到像素值＝1的像素点时，以顺时针或逆时针方向绕该像素点的八邻域搜寻一周，得到一个像素值变化次数；若像素值变化次数为4，显示过该像素点存在两根直线，通过坐标计算得知这两根直线存在合理的夹角时，则确定该像素点为边缘顶点并记录；若像素值变化次数为6，显示过该像素点存在三根直线，则确定该像素点为中间顶点并记录。
所述步骤“顶点提取”的第一步是以“二值化图”为基础，将像素值为1的线条采用线宽为1个像素的线段绘制骨架图；第二步是在骨架图的基础上，以像素值＝1的像素点处为中心点建立尺寸为5×5个像素的正方形窗口，如果该窗口的部分区域溢出骨架图时，先将溢出区域的像素点的像素值全部赋值为0，然后采用Harris算法计算出该窗口中心点对应的角点响应函数值R，再以角点响应函数值R的最大值的百分之一为界限值，在同一窗口内的全部角点响应函数值R中，将小于界限值的角点响应函数值R置为零，重复以上操作遍历整个骨架图；下一步，以像素值＝1且角点响应函数值R大于零的像素点为中心点，建立尺寸为3×3个像素的正方形窗口，若该窗口中心点的R值为本窗口内的最大值，则记录该点为顶点，重复以上操作遍 历整个骨架图。
所述“重构胞元”的第一步是在“二值化图”的中央位置任意选择x个顶点为基准点，在每个基准点处找到与其最接近的顶点为相邻点，计算出基准点与相邻点的距离，保留其中距离最短的三条记录，然后求3x条记录的距离的平均值为胞元边长A；第二步是将获得的顶点分为主动连接点和被动连接点两大类；即：先将骨架图分为两个区，以骨架图周边宽度为1A～2A的区域是为边缘区，被边缘区包围的区域为中央区；位于中央区的顶点为主动连接点，位于边缘区的顶点为被动连接点；第三步是将所有主动连接点与最接近的三个顶点用线段连接起来，即：每个主动连接点必须且只能保留三条最短的线段，连线完成后形成胞元重构图。
所述步骤“重构胞元”包括边缘扩展和顶点连线；所述步骤“边缘扩展”是以形态图像为基础，在形态图像四个边的最外缘，均向外扩展至少1个像素的宽度形成扩展区，扩展区内所有像素点的像素值全部置为1，得到扩展图像；所述步骤“顶点连线”是以扩展图像为对象，以从左至右、自上而下的顺序遍历该扩展图像，当遇到像素值＝0的像素点时，就采用摩尔邻域跟踪算法寻找并记录同一胞元的顶点及其连接顺序并记录在该胞元的名下，以每个胞元最多保留六个顶点为原则，删除重复的记录，按照保留的记录进行顶点连线，画出完整的胞元；随后将该胞元内所有像素点的像素值全部置为1；在此基础上，寻找下一个像素值＝0的像素点，重复以上操作，遍历完成的同时也完成了胞元重构图。
一种适用于蜂窝结构几何规整度图像识别方法的系统，所述系统包括检测台、数码相机和计算机；所述数码相机和计算机电连接；
所述数码相机至少为一台，其分辨率不低于1080P，配置远心镜头，其安装方式为固定式或/和移动式。
所述检测台为活动式工作平台，包括置物台、升降装置和夹具，升降装置安装在置物台的底部；被测蜂窝件放置在置物台上；所述升降装置包括竖向导轨、电动推杆或电液推杆，能够推动置物台沿竖向导轨上下移动，调节被测蜂窝件的高度，以保证被测蜂窝件的上端面与夹具的上端面平齐；升降装置的控制部分与计算机电连接；
所述夹具由四块平板及驱动装置组成，能够在驱动装置的作用下向被测蜂窝件靠拢，靠紧被测蜂窝件后锁死，用于定位及固定被测蜂窝件。
所述数码相机的安装方式为移动式时，所述系统增加行走式龙门架、滑轨、移动装置；
数码相机安装在行走式龙门架的横梁上，能够在移动装置的驱动下沿横梁横向移动；
行走式龙门架能够在移动装置的驱动下沿滑轨纵向移动，数码相机及行走式龙门架的移动均由计算机控制。
所述计算机包括：
顶点提取模块，用于寻找胞元的顶点并记录；
重构胞元模块，用于将提取的顶点依据胞元与顶点的映射关系连线，得到胞元重构图；
质量评估模块，用于以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再与设置的公差带相比较判定是否合格。
所述计算机还包括：
二值化模块，用于将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图。
所述顶点提取模块包括：
确定壁厚单元，用于设置一个边长能够从小变大变化的正方形窗口，当某一边长的窗口遍历形态图像后，如果窗口内的像素值＝0的像素数的最小值为非零时，将该正方形窗口的边长定义为壁厚L；
像素赋值单元，用于设置一个边长为壁厚L的正方形的赋值窗口，用赋值窗口遍历形态图像中像素值＝1的像素点，然后将该窗口内的像素值＝1的像素个数的总和赋值给该赋值窗口的中心点的像素点上；
确定胞元边长单元，用于在完成“像素赋值”之后，找到最大赋值的像素点，以该点为中心建立一个正方形区域，其初始边长E＝壁厚L，计算出该区域四个边界上各像素点的赋值与该中心点的赋值的差值绝对值Z，遍历整个形态图像，记录本轮中最小的差值绝对值Z及其坐标；以E＝E+2个像素建立新的正方形区域，重复上述过程，直到最小的差值绝对值Z有明显的反向增大的趋势为止，以此时取出最小的差值绝对值Z的像素点坐标与对应的区域中心点的坐标，通过两者的坐标值求解得到蜂窝胞元边长A；
像素湮灭单元，用于找到最大赋值的像素点，并确定该像素点为 顶点并记录，再以该像素点为中心，建立一个以蜂窝胞元边长A为边长的正方形的湮灭窗口，将湮灭窗口内所有像素值＝1的像素点上的赋值全部清零，在此基础上，再在剩余的赋值中再找到最大赋值的像素点确定为顶点并记录，重复湮灭窗口的操作，直到像素值＝1的像素点上的赋值小于给定阈值为止，顶点提取完毕。
所述顶点提取模块具体用于以形态图像为基础，将像素值为1的线条采用线宽为1个像素的线段绘制成骨架图；然后以具有k个像素点的骨架图为基础，遍历像素点1至像素点k，每当遇到像素值＝1的像素点时，以顺时针或逆时针方向绕该像素点的八邻域搜寻一周，得到一个像素值变化次数；若像素值变化次数为4，显示过该像素点存在两根直线，通过坐标计算得知这两根直线存在合理的夹角时，则确定该像素点为边缘顶点并记录；若像素值变化次数为6，显示过该像素点存在三根直线，则确定该像素点为中间顶点并记录。
所述顶点提取模块具体用于以“二值化图”为基础，将像素值为1的线条采用线宽为1个像素的线段绘制骨架图；然后在骨架图的基础上，以像素值＝1的像素点处为中心点建立尺寸为5×5个像素的正方形窗口，如果该窗口的部分区域溢出骨架图时，先将溢出区域的像素点的像素值全部赋值为0，然后采用Harris算法计算出该窗口中心点对应的角点响应函数值R，再以角点响应函数值R的最大值的百分之一为界限值，在同一窗口内的全部角点响应函数值R中，将小于界限值的角点响应函数值R置为零，重复以上操作遍历整个骨架图；最后以像素值＝1且角点响应函数值R大于零的像素点为中心点，建 立尺寸为3×3个像素的正方形窗口，若该窗口中心点的R值为本窗口内的最大值，则记录该点为顶点，重复以上操作遍历整个骨架图。
所述重构胞元模块具体用于在“二值化图”的中央位置任意选择x个顶点为基准点，在每个基准点处找到与其最接近的顶点为相邻点，计算出基准点与相邻点的距离，保留其中距离最短的三条记录，然后求3x条记录的距离的平均值为胞元边长A；然后将获得的顶点分为主动连接点和被动连接点两大类；即：先将骨架图分为两个区，以骨架图周边宽度为1A～2A的区域是为边缘区，被边缘区包围的区域为中央区；位于中央区的顶点为主动连接点，位于边缘区的顶点为被动连接点；最后将所有主动连接点与最接近的三个顶点用线段连接起来，即：每个主动连接点必须且只能保留三条最短的线段，连线完成后形成胞元重构图。
所述重构胞元模块包括：
边缘扩展单元用于以形态图像为基础，在形态图像四个边的最外缘，均向外扩展至少1个像素的宽度形成扩展区，扩展区内所有像素点的像素值全部置为1，得到扩展图像；
顶点连线单元用于以扩展图像为对象，以从左至右、自上而下的顺序遍历该扩展图像，当遇到像素值＝0的像素点时，就采用摩尔邻域跟踪算法寻找并记录同一胞元的顶点及其连接顺序并记录在该胞元的名下，以每个胞元最多保留六个顶点为原则，删除重复的记录，按照保留的记录进行顶点连线，画出完整的胞元；随后将该胞元内所有像素点的像素值全部置为1；在此基础上，寻找下一个像素值＝0 的像素点，重复以上操作，遍历完成的同时也完成了胞元重构图。
一种储存介质，所述存储介质上存储有计算机程序，所述程序被处理器执行时实现上述技术方案中任意一项所述方法的步骤。
一种电子设备，包括存储器、显示器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，所述处理器执行所述程序时实现上述技术方案中任意一项所述方法的步骤。
与现有技术相比较，本发明的方法及其系统具有科学合理。简单易行，精度高，工作效率高等优点。
附图说明
图1为本发明方法的流程框图；
图2为本发明方法的一实施例顶点提取及胞元重构示意图；
图3为本发明系统的一实施例设备配置示意图；
图4为图3的俯视图。
图中：1-检测台，2-数码相机，3-计算机，4-置物台，5-升降装置，6-夹具，7-行走式龙门架，8-滑轨，9-移动装置。
具体实施方式
以下结合附图和实施例，对本发明作进一步的说明：
方法，参考附图1，2：
一种蜂窝结构几何规整度图像识别方法，所述方法包括以下步骤：获取图像、图像处理、顶点提取、重构胞元、质量评估；所述步骤“获取图像”包括拍摄图像和计算机读取图像；所述步骤“顶点提取”是在“图像处理”的基础上寻找胞元的顶点并记录；所述步骤“重构胞 元”是将提取的顶点依据胞元与顶点的映射关系连线，得到胞元重构图；
所述步骤“图像处理”和步骤“顶点提取”之间设置步骤“二值化”，所述步骤“二值化”是将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图；
所述步骤“质量评估”是以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再与设置的公差带相比较判定是否合格。
所述步骤“顶点提取”是以“二值化图”为基础，将所述“二值化图”进行闭运算处理，得到平滑蜂窝顶点图像；
将所述平滑蜂窝顶点图像依次经过膨胀处理和腐蚀处理，得到只有蜂窝壁交汇处图像；
在所述只有蜂窝壁交汇处图像上进行蜂窝壁取最大圆圆心处理，得到蜂窝胞元的顶点。
所述步骤“顶点提取”是以“二值化图”为基础，顺序执行确定壁厚、像素赋值、确定胞元边长和像素湮灭；
所述步骤“确定壁厚”是设置一个边长能够从小变大变化的正方形窗口，当某一边长的窗口遍历形态图像后，如果窗口内的像素值＝0的像素数的最小值为非零时，将该正方形窗口的边长定义为壁厚L；
所述步骤“像素赋值”是设置一个边长为壁厚L的正方形的赋值窗口，用赋值窗口遍历形态图像中像素值＝1的像素点，然后将该窗口内的像素值＝1的像素个数的总和赋值给该赋值窗口的中心点的像素点上；
所述步骤“确定胞元边长”是在完成“像素赋值”之后，找到最大赋值的像素点，以该点为中心建立一个正方形区域，其初始边长E＝壁厚L，计算出该区域四个边界上各像素点的赋值与该中心点的赋值的差值绝对值Z，遍历整个形态图像，记录本轮中最小的差值绝对值Z及其坐标；以E＝E+2个像素建立新的正方形区域，重复上述过程，直到最小的差值绝对值Z有明显的反向增大的趋势为止，以此时取出最小的差值绝对值Z的像素点坐标与对应的区域中心点的坐标，通过两者的坐标值求解得到蜂窝胞元边长A；
所述步骤“像素湮灭”是找到最大赋值的像素点，并确定该像素点为顶点并记录，再以该像素点为中心，建立一个以蜂窝胞元边长A为边长的正方形的湮灭窗口，将湮灭窗口内所有像素值＝1的像素点上的赋值全部清零，在此基础上，再在剩余的赋值中再找到最大赋值的像素点确定为顶点并记录，重复湮灭窗口的操作，直到像素值＝1的像素点上的赋值小于给定阈值为止，顶点提取完毕。
所述步骤“顶点提取”的第一步是以形态图像为基础，将像素值为1的线条采用线宽为1个像素的线段绘制成骨架图；第二步是以具有k个像素点的骨架图为基础，遍历像素点1至像素点k，每当遇到像素值＝1的像素点时，以顺时针或逆时针方向绕该像素点的八邻域搜寻一周，得到一个像素值变化次数；若像素值变化次数为4，显示过该像素点存在两根直线，通过坐标计算得知这两根直线存在合理的夹角时，则确定该像素点为边缘顶点并记录；若像素值变化次数为6，显示过该像素点存在三根直线，则确定该像素点为中间顶点并记录。
所述步骤“顶点提取”的第一步是以“二值化图”为基础，将像素值为1的线条采用线宽为1个像素的线段绘制骨架图；第二步是在骨架图的基础上，以像素值＝1的像素点处为中心点建立尺寸为5×5个像素的正方形窗口，如果该窗口的部分区域溢出骨架图时，先将溢出区域的像素点的像素值全部赋值为0，然后采用Harris算法计算出该窗口中心点对应的角点响应函数值R，再以角点响应函数值R的最大值的百分之一为界限值，在同一窗口内的全部角点响应函数值R中，将小于界限值的角点响应函数值R置为零，重复以上操作遍历整个骨架图；下一步，以像素值＝1且角点响应函数值R大于零的像素点为中心点，建立尺寸为3×3个像素的正方形窗口，若该窗口中心点的R值为本窗口内的最大值，则记录该点为顶点，重复以上操作遍历整个骨架图。
所述“重构胞元”的第一步是在“二值化图”的中央位置任意选择x个顶点为基准点，在每个基准点处找到与其最接近的顶点为相邻点，计算出基准点与相邻点的距离，保留其中距离最短的三条记录，然后求3x条记录的距离的平均值为胞元边长A；第二步是将获得的顶点分为主动连接点和被动连接点两大类；即：先将骨架图分为两个区，以骨架图周边宽度为1A～2A的区域是为边缘区，被边缘区包围的区域为中央区；位于中央区的顶点为主动连接点，位于边缘区的顶点为被动连接点；第三步是将所有主动连接点与最接近的三个顶点用线段连接起来，即：每个主动连接点必须且只能保留三条最短的线段，连线完成后形成胞元重构图。
所述步骤“重构胞元”包括边缘扩展和顶点连线；所述步骤“边缘扩展”是以形态图像为基础，在形态图像四个边的最外缘，均向外扩展至少1个像素的宽度形成扩展区，扩展区内所有像素点的像素值全部置为1，得到扩展图像；所述步骤“顶点连线”是以扩展图像为对象，以从左至右、自上而下的顺序遍历该扩展图像，当遇到像素值＝0的像素点时，就采用摩尔邻域跟踪算法寻找并记录同一胞元的顶点及其连接顺序并记录在该胞元的名下，以每个胞元最多保留六个顶点为原则，删除重复的记录，按照保留的记录进行顶点连线，画出完整的胞元；随后将该胞元内所有像素点的像素值全部置为1；在此基础上，寻找下一个像素值＝0的像素点，重复以上操作，遍历完成的同时也完成了胞元重构图。
系统，参考附图3，4：
一种适用于蜂窝结构几何规整度图像识别方法的系统，所述系统包括检测台1、数码相机2和计算机3；所述数码相机2和计算机3电连接；
所述数码相机2至少为一台，其分辨率不低于1080P，配置远心镜头，其安装方式为固定式或/和移动式。
所述检测台1为活动式工作平台，包括置物台4、升降装置5和夹具6，升降装置5安装在置物台4的底部；被测蜂窝件放置在置物台4上；所述升降装置5包括竖向导轨、电动推杆或电液推杆，能够推动置物台4沿竖向导轨上下移动，调节被测蜂窝件的高度，以保证被测蜂窝件的上端面与夹具6的上端面平齐；升降装置5的控制部分 与计算机3电连接；
所述夹具6由四块平板及驱动装置组成，能够在驱动装置的作用下向被测蜂窝件靠拢，靠紧被测蜂窝件后锁死，用于定位及固定被测蜂窝件。
所述数码相机2的安装方式为移动式时，所述系统增加行走式龙门架7、滑轨8、移动装置9；
数码相机2安装在行走式龙门架7的横梁上，能够在移动装置9的驱动下沿横梁横向移动；
行走式龙门架7能够在移动装置9的驱动下沿滑轨8纵向移动，数码相机2及行走式龙门架7的移动均由计算机3控制。
所述计算机3包括：
顶点提取模块，用于寻找胞元的顶点并记录；
重构胞元模块，用于将提取的顶点依据胞元与顶点的映射关系连线，得到胞元重构图；
质量评估模块，用于以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再与设置的公差带相比较判定是否合格。
所述计算机3还包括：
二值化模块，用于将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图。
所述顶点提取模块包括：
确定壁厚单元，用于设置一个边长能够从小变大变化的正方形窗 口，当某一边长的窗口遍历形态图像后，如果窗口内的像素值＝0的像素数的最小值为非零时，将该正方形窗口的边长定义为壁厚L；
像素赋值单元，用于设置一个边长为壁厚L的正方形的赋值窗口，用赋值窗口遍历形态图像中像素值＝1的像素点，然后将该窗口内的像素值＝1的像素个数的总和赋值给该赋值窗口的中心点的像素点上；
确定胞元边长单元，用于在完成“像素赋值”之后，找到最大赋值的像素点，以该点为中心建立一个正方形区域，其初始边长E＝壁厚L，计算出该区域四个边界上各像素点的赋值与该中心点的赋值的差值绝对值Z，遍历整个形态图像，记录本轮中最小的差值绝对值Z及其坐标；以E＝E+2个像素建立新的正方形区域，重复上述过程，直到最小的差值绝对值Z有明显的反向增大的趋势为止，以此时取出最小的差值绝对值Z的像素点坐标与对应的区域中心点的坐标，通过两者的坐标值求解得到蜂窝胞元边长A；
像素湮灭单元，用于找到最大赋值的像素点，并确定该像素点为顶点并记录，再以该像素点为中心，建立一个以蜂窝胞元边长A为边长的正方形的湮灭窗口，将湮灭窗口内所有像素值＝1的像素点上的赋值全部清零，在此基础上，再在剩余的赋值中再找到最大赋值的像素点确定为顶点并记录，重复湮灭窗口的操作，直到像素值＝1的像素点上的赋值小于给定阈值为止，顶点提取完毕。
所述顶点提取模块具体用于以形态图像为基础，将像素值为1的线条采用线宽为1个像素的线段绘制成骨架图；然后以具有k个像素点的骨架图为基础，遍历像素点1至像素点k，每当遇到像素值＝1的 像素点时，以顺时针或逆时针方向绕该像素点的八邻域搜寻一周，得到一个像素值变化次数；若像素值变化次数为4，显示过该像素点存在两根直线，通过坐标计算得知这两根直线存在合理的夹角时，则确定该像素点为边缘顶点并记录；若像素值变化次数为6，显示过该像素点存在三根直线，则确定该像素点为中间顶点并记录。
所述顶点提取模块具体用于以“二值化图”为基础，将像素值为1的线条采用线宽为1个像素的线段绘制骨架图；然后在骨架图的基础上，以像素值＝1的像素点处为中心点建立尺寸为5×5个像素的正方形窗口，如果该窗口的部分区域溢出骨架图时，先将溢出区域的像素点的像素值全部赋值为0，然后采用Harris算法计算出该窗口中心点对应的角点响应函数值R，再以角点响应函数值R的最大值的百分之一为界限值，在同一窗口内的全部角点响应函数值R中，将小于界限值的角点响应函数值R置为零，重复以上操作遍历整个骨架图；最后以像素值＝1且角点响应函数值R大于零的像素点为中心点，建立尺寸为3×3个像素的正方形窗口，若该窗口中心点的R值为本窗口内的最大值，则记录该点为顶点，重复以上操作遍历整个骨架图。
所述重构胞元模块具体用于在“二值化图”的中央位置任意选择x个顶点为基准点，在每个基准点处找到与其最接近的顶点为相邻点，计算出基准点与相邻点的距离，保留其中距离最短的三条记录，然后求3x条记录的距离的平均值为胞元边长A；然后将获得的顶点分为主动连接点和被动连接点两大类；即：先将骨架图分为两个区，以骨架图周边宽度为1A～2A的区域是为边缘区，被边缘区包围的区域为 中央区；位于中央区的顶点为主动连接点，位于边缘区的顶点为被动连接点；最后将所有主动连接点与最接近的三个顶点用线段连接起来，即：每个主动连接点必须且只能保留三条最短的线段，连线完成后形成胞元重构图。
所述重构胞元模块包括：
边缘扩展单元用于以形态图像为基础，在形态图像四个边的最外缘，均向外扩展至少1个像素的宽度形成扩展区，扩展区内所有像素点的像素值全部置为1，得到扩展图像；
顶点连线单元用于以扩展图像为对象，以从左至右、自上而下的顺序遍历该扩展图像，当遇到像素值＝0的像素点时，就采用摩尔邻域跟踪算法寻找并记录同一胞元的顶点及其连接顺序并记录在该胞元的名下，以每个胞元最多保留六个顶点为原则，删除重复的记录，按照保留的记录进行顶点连线，画出完整的胞元；随后将该胞元内所有像素点的像素值全部置为1；在此基础上，寻找下一个像素值＝0的像素点，重复以上操作，遍历完成的同时也完成了胞元重构图。
一种储存介质，所述存储介质上存储有计算机程序，所述程序被处理器执行时实现上述实施例中任意一项所述方法的步骤。
一种电子设备，包括存储器、显示器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，所述处理器执行所述程序时实现上述实施例中任意一项所述方法的步骤。
方法实施例1：
一种蜂窝结构几何规整度图像识别方法，所述方法包括以下步骤： 获取图像、图像处理、顶点提取、重构胞元、质量评估；所述步骤“获取图像”包括拍摄图像和计算机读取图像；所述步骤“顶点提取”是在“图像处理”的基础上寻找胞元的顶点并记录；所述步骤“重构胞元”是将提取的顶点依据胞元与顶点的映射关系连线，得到胞元重构图；
所述步骤“图像处理”和步骤“顶点提取”之间设置步骤“二值化”，所述步骤“二值化”是将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图；
所述步骤“质量评估”是以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再与设置的公差带相比较判定是否合格。
方法实施例2：
与“方法实施例1”基本上相同，不同的是：所述步骤“顶点提取”是以“二值化图”为基础，顺序执行确定壁厚、像素赋值、确定胞元边长和像素湮灭；
所述步骤“确定壁厚”是设置一个边长能够从小变大变化的正方形窗口，当某一边长的窗口遍历形态图像后，如果窗口内的像素值＝0的像素数的最小值为非零时，将该正方形窗口的边长定义为壁厚L；
所述步骤“像素赋值”是设置一个边长为壁厚L的正方形的赋值窗口，用赋值窗口遍历形态图像中像素值＝1的像素点，然后将该窗口内的像素值＝1的像素个数的总和赋值给该赋值窗口的中心点的像素点上；
所述步骤“确定胞元边长”是在完成“像素赋值”之后，找到最大赋值的像素点，以该点为中心建立一个正方形区域，其初始边长E＝壁厚L，计算出该区域四个边界上各像素点的赋值与该中心点的赋值的差值绝对值Z，遍历整个形态图像，记录本轮中最小的差值绝对值Z及其坐标；以E＝E+2个像素建立新的正方形区域，重复上述过程，直到最小的差值绝对值Z有明显的反向增大的趋势为止，以此时取出最小的差值绝对值Z的像素点坐标与对应的区域中心点的坐标，通过两者的坐标值求解得到蜂窝胞元边长A；
所述步骤“像素湮灭”是找到最大赋值的像素点，并确定该像素点为顶点并记录，再以该像素点为中心，建立一个以蜂窝胞元边长A为边长的正方形的湮灭窗口，将湮灭窗口内所有像素值＝1的像素点上的赋值全部清零，在此基础上，再在剩余的赋值中再找到最大赋值的像素点确定为顶点并记录，重复湮灭窗口的操作，直到像素值＝1的像素点上的赋值小于给定阈值为止，顶点提取完毕。
方法实施例3：
与“方法实施例1”基本上相同，不同的是：所述步骤“顶点提取”的第一步是以形态图像为基础，将像素值为1的线条采用线宽为1个像素的线段绘制成骨架图；第二步是以具有k个像素点的骨架图为基础，遍历像素点1至像素点k，每当遇到像素值＝1的像素点时，以顺时针或逆时针方向绕该像素点的八邻域搜寻一周，得到一个像素值变化次数；若像素值变化次数为4，显示过该像素点存在两根直线，通过坐标计算得知这两根直线存在合理的夹角时，则确定该像素点为 边缘顶点并记录；若像素值变化次数为6，显示过该像素点存在三根直线，则确定该像素点为中间顶点并记录。
方法实施例4：
与“方法实施例1”基本上相同，不同的是：所述步骤“顶点提取”的第一步是以“二值化图”为基础，将像素值为1的线条采用线宽为1个像素的线段绘制骨架图；第二步是在骨架图的基础上，以像素值＝1的像素点处为中心点建立尺寸为5×5个像素的正方形窗口，如果该窗口的部分区域溢出骨架图时，先将溢出区域的像素点的像素值全部赋值为0，然后采用Harris算法计算出该窗口中心点对应的角点响应函数值R，再以角点响应函数值R的最大值的百分之一为界限值，在同一窗口内的全部角点响应函数值R中，将小于界限值的角点响应函数值R置为零，重复以上操作遍历整个骨架图；下一步，以像素值＝1且角点响应函数值R大于零的像素点为中心点，建立尺寸为3×3个像素的正方形窗口，若该窗口中心点的R值为本窗口内的最大值，则记录该点为顶点，重复以上操作遍历整个骨架图。
方法实施例5-8：
分别与“方法实施例1-4”基本上相同，不同的是：所述“重构胞元”的第一步是在“二值化图”的中央位置任意选择x个顶点为基准点，在每个基准点处找到与其最接近的顶点为相邻点，计算出基准点与相邻点的距离，保留其中距离最短的三条记录，然后求3x条记录的距离的平均值为胞元边长A；第二步是将获得的顶点分为主动连接点和被动连接点两大类；即：先将骨架图分为两个区，以骨架图周 边宽度为1A～2A的区域是为边缘区，被边缘区包围的区域为中央区；位于中央区的顶点为主动连接点，位于边缘区的顶点为被动连接点；第三步是将所有主动连接点与最接近的三个顶点用线段连接起来，即：每个主动连接点必须且只能保留三条最短的线段，连线完成后形成胞元重构图。
方法实施例9-12：
分别与“方法实施例1-4”基本上相同，不同的是：所述步骤“重构胞元”包括边缘扩展和顶点连线；所述步骤“边缘扩展”是以形态图像为基础，在形态图像四个边的最外缘，均向外扩展至少1个像素的宽度形成扩展区，扩展区内所有像素点的像素值全部置为1，得到扩展图像；所述步骤“顶点连线”是以扩展图像为对象，以从左至右、自上而下的顺序遍历该扩展图像，当遇到像素值＝0的像素点时，就采用摩尔邻域跟踪算法寻找并记录同一胞元的顶点及其连接顺序并记录在该胞元的名下，以每个胞元最多保留六个顶点为原则，删除重复的记录，按照保留的记录进行顶点连线，画出完整的胞元；随后将该胞元内所有像素点的像素值全部置为1；在此基础上，寻找下一个像素值＝0的像素点，重复以上操作，遍历完成的同时也完成了胞元重构图。
系统实施例1：
一种适用于蜂窝结构几何规整度图像识别方法的系统，所述系统包括检测台1、数码相机2和计算机3；所述数码相机2和计算机3电连接；
所述数码相机2至少为一台，其分辨率不低于1080P，配置远心镜头，其安装方式为固定式或/和移动式。
系统实施例2：
与“系统实施例1”基本上相同，不同的是：所述检测台1为活动式工作平台，包括置物台4、升降装置5和夹具6，升降装置5安装在置物台4的底部；被测蜂窝件放置在置物台4上；所述升降装置5包括竖向导轨、电动推杆或电液推杆，能够推动置物台4沿竖向导轨上下移动，调节被测蜂窝件的高度，以保证被测蜂窝件的上端面与夹具6的上端面平齐；升降装置5的控制部分与计算机3电连接；
所述夹具6由四块平板及驱动装置组成，能够在驱动装置的作用下向被测蜂窝件靠拢，靠紧被测蜂窝件后锁死，用于定位及固定被测蜂窝件。
系统实施例3，4：
分别与“系统实施例1，2”基本上相同，不同的是：所述数码相机2的安装方式为移动式时，所述系统增加行走式龙门架7、滑轨8、移动装置9；
数码相机2安装在行走式龙门架7的横梁上，能够在移动装置9的驱动下沿横梁横向移动；
行走式龙门架7能够在移动装置9的驱动下沿滑轨8纵向移动，数码相机2及行走式龙门架7的移动均由计算机3控制。
本发明方法一实施例的顶点提取及胞元重构结果参见附图2，计算结果为：内角偏差量的最大值为14.42，内角偏差量的平均值为 2.62，内角偏差量的标准差为2.42，均在设定值范围内，判定产品质量为合格。

Claims (11)

  1. 一种蜂窝结构几何规整度图像识别方法，所述方法包括以下步骤：获取图像、图像处理、顶点提取、重构胞元、质量评估；所述步骤“获取图像”包括拍摄图像和计算机读取图像；所述步骤“顶点提取”是在“图像处理”的基础上寻找胞元的顶点并记录；所述步骤“重构胞元”是将提取的顶点依据胞元与顶点的映射关系连线，得到胞元重构图；其特征在于：
    所述步骤“图像处理”和步骤“顶点提取”之间设置步骤“二值化”，所述步骤“二值化”是将图像中背景的像素值置为0，将图像中蜂窝骨架的像素值置为1，形成二值化图；
    所述步骤“质量评估”是以胞元重构图为基础，计算出所有胞元的角偏差值及其平均值、线偏差值及其平均值，再与设置的公差带相比较判定是否合格。
  2. 根据权利要求1所述的方法，其特征在于：所述步骤“顶点提取”是以“二值化图”为基础，顺序执行确定壁厚、像素赋值、确定胞元边长和像素湮灭；
    所述步骤“确定壁厚”是设置一个边长能够从小变大变化的正方形窗口，当某一边长的窗口遍历形态图像后，如果窗口内的像素值＝0的像素数的最小值为非零时，将该正方形窗口的边长定义为壁厚L；
    所述步骤“像素赋值”是设置一个边长为壁厚L的正方形的赋值窗口，用赋值窗口遍历形态图像中像素值＝1的像素点，然后将该窗口内的像素值＝1的像素个数的总和赋值给该赋值窗口的中心点的像 素点上；
    所述步骤“确定胞元边长”是在完成“像素赋值”之后，找到最大赋值的像素点，以该点为中心建立一个正方形区域，其初始边长E＝壁厚L，计算出该区域四个边界上各像素点的赋值与该中心点的赋值的差值绝对值Z，遍历整个形态图像，记录本轮中最小的差值绝对值Z及其坐标；以E＝E+2个像素建立新的正方形区域，重复上述过程，直到最小的差值绝对值Z有明显的反向增大的趋势为止，以此时取出最小的差值绝对值Z的像素点坐标与对应的区域中心点的坐标，通过两者的坐标值求解得到蜂窝胞元边长A；
    所述步骤“像素湮灭”是找到最大赋值的像素点，并确定该像素点为顶点并记录，再以该像素点为中心，建立一个以蜂窝胞元边长A为边长的正方形的湮灭窗口，将湮灭窗口内所有像素值＝1的像素点上的赋值全部清零，在此基础上，再在剩余的赋值中再找到最大赋值的像素点确定为顶点并记录，重复湮灭窗口的操作，直到像素值＝1的像素点上的赋值小于给定阈值为止，顶点提取完毕。
  3. 根据权利要求1所述的方法，其特征在于：所述步骤“顶点提取”是以“二值化图”为基础，顺序执行：
    将所述二值化图像进行闭运算处理，得到平滑蜂窝顶点图像；
    在所述平滑蜂窝顶点图像上进行蜂窝壁取最大圆圆心处理，得到蜂窝胞元的顶点。
  4. 根据权利要求1所述的方法，其特征在于：所述步骤“顶点提取”的第一步是以“二值化图”为基础，将像素值为1的线条采用线宽为1个像素的线段绘制骨架图；第二步是在骨架图的基础上，以 像素值＝1的像素点处为中心点建立尺寸为5×5个像素的正方形窗口，如果该窗口的部分区域溢出骨架图时，先将溢出区域的像素点的像素值全部赋值为0，然后采用Harris算法计算出该窗口中心点对应的角点响应函数值R，再以角点响应函数值R的最大值的百分之一为界限值，在同一窗口内的全部角点响应函数值R中，将小于界限值的角点响应函数值R置为零，重复以上操作遍历整个骨架图；下一步，以像素值＝1且角点响应函数值R大于零的像素点为中心点，建立尺寸为3×3个像素的正方形窗口，若该窗口中心点的R值为本窗口内的最大值，则记录该点为顶点，重复以上操作遍历整个骨架图。
  5. 根据权利要求1-4任选一项所述的方法，其特征在于：所述“重构胞元”的第一步是在“二值化图”的中央位置任意选择x个顶点为基准点，在每个基准点处找到与其最接近的顶点为相邻点，计算出基准点与相邻点的距离，保留其中距离最短的三条记录，然后求3x条记录的距离的平均值为胞元边长A；第二步是将获得的顶点分为主动连接点和被动连接点两大类；即：先将骨架图分为两个区，以骨架图周边宽度为1A～2A的区域是为边缘区，被边缘区包围的区域为中央区；位于中央区的顶点为主动连接点，位于边缘区的顶点为被动连接点；第三步是将所有主动连接点与最接近的三个顶点用线段连接起来，即：每个主动连接点必须且只能保留三条最短的线段，连线完成后形成胞元重构图。
  6. 根据权利要求1-4任选一所述方法，其特征在于：所述步骤“重构胞元”包括边缘扩展和顶点连线；所述步骤“边缘扩展”是以形态 图像为基础，在形态图像四个边的最外缘，均向外扩展至少1个像素的宽度形成扩展区，扩展区内所有像素点的像素值全部置为1，得到扩展图像；所述步骤“顶点连线”是以扩展图像为对象，以从左至右、自上而下的顺序遍历该扩展图像，当遇到像素值＝0的像素点时，就采用摩尔邻域跟踪算法寻找并记录同一胞元的顶点及其连接顺序并记录在该胞元的名下，以每个胞元最多保留六个顶点为原则，删除重复的记录，按照保留的记录进行顶点连线，画出完整的胞元；随后将该胞元内所有像素点的像素值全部置为1；在此基础上，寻找下一个像素值＝0的像素点，重复以上操作，遍历完成的同时也完成了胞元重构图。
  7. 一种适用于权利要求1-6任选一项所述方法的系统，所述系统包括检测台(1)、数码相机(2)和计算机(3)；所述数码相机(2)和计算机(3)电连接；其特征在于：
    所述数码相机(2)至少为一台，其分辨率不低于1080P，配置远心镜头，其安装方式为固定式或/和移动式。
  8. 根据权利要求7所述系统，其特征在于：所述检测台(1)为活动式工作平台，包括置物台(4)、升降装置(5)和夹具(6)，升降装置(5)安装在置物台(4)的底部；被测蜂窝件放置在置物台(4)上；所述升降装置(5)包括竖向导轨、电动推杆或电液推杆，能够推动置物台(4)沿竖向导轨上下移动，调节被测蜂窝件的高度，以保证被测蜂窝件的上端面与夹具(6)的上端面平齐；升降装置(5)的控制部分与计算机(3)电连接；
    所述夹具(6)由四块平板及驱动装置组成，能够在驱动装置的作用下向被测蜂窝件靠拢，靠紧被测蜂窝件后锁死，用于定位及固定被测蜂窝件。
  9. 根据权利要求7或8所述系统，其特征在于：所述数码相机(2)的安装方式为移动式时，所述系统增加行走式龙门架(7)、滑轨(8)、移动装置(9)；
    数码相机(2)安装在行走式龙门架(7)的横梁上，能够在移动装置(9)的驱动下沿横梁横向移动；
    行走式龙门架(7)能够在移动装置(9)的驱动下沿滑轨(8)纵向移动，数码相机(2)及行走式龙门架(7)的移动均由计算机(3)控制。
  10. 一种储存介质，其特征在于，所述存储介质上存储有计算机程序，所述程序被处理器执行时实现权利要求1-6中任意一项所述方法的步骤。
  11. 一种电子设备，其特征在于，包括存储器、显示器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，所述处理器执行所述程序时实现权利要求1-6中任意一项所述方法的步骤。
PCT/CN2020/109726 2020-05-09 2020-08-18 蜂窝结构几何规整度图像识别方法及系统 WO2021227285A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/422,720 US11893768B2 (en) 2020-05-09 2020-08-18 Method and system for recognizing geometric regularity image of honeycomb structure

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202010388489.XA CN111583242B (zh) 2020-05-09 2020-05-09 蜂窝规整度检测的形态学顶点提取方法及检测装置
CN202010388353.9A CN111583239B (zh) 2020-05-09 2020-05-09 蜂窝结构几何规整度图像识别方法及系统
CN202010388489.X 2020-05-09
CN202010388353.9 2020-05-09

Publications (1)

Publication Number Publication Date
WO2021227285A1 true WO2021227285A1 (zh) 2021-11-18

Family

ID=78526337

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/109726 WO2021227285A1 (zh) 2020-05-09 2020-08-18 蜂窝结构几何规整度图像识别方法及系统

Country Status (2)

Country Link
US (1) US11893768B2 (zh)
WO (1) WO2021227285A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104369188A (zh) * 2014-11-20 2015-02-25 中国计量学院 基于机器视觉和超声波传感器的工件抓取装置与方法
CN110473174A (zh) * 2019-07-26 2019-11-19 东南大学 一种基于图像计算铅笔精确数目的方法
CN111054782A (zh) * 2019-12-31 2020-04-24 太原科技大学 一种宽厚板板形检测装置及方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104369188A (zh) * 2014-11-20 2015-02-25 中国计量学院 基于机器视觉和超声波传感器的工件抓取装置与方法
CN110473174A (zh) * 2019-07-26 2019-11-19 东南大学 一种基于图像计算铅笔精确数目的方法
CN111054782A (zh) * 2019-12-31 2020-04-24 太原科技大学 一种宽厚板板形检测装置及方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CUI CAN; WANG ZHONGGANG; ZHOU WEI; WU YONGJUN; WEI WANHUI: "Branch point algorithm for structural irregularity determination of honeycomb", COMPOSITES PART B, ELSEVIER, AMSTERDAM, NL, vol. 162, 19 October 2018 (2018-10-19), AMSTERDAM, NL, pages 323 - 330, XP085596185, ISSN: 1359-8368, DOI: 10.1016/j.compositesb.2018.10.062 *

Also Published As

Publication number Publication date
US20230214987A1 (en) 2023-07-06
US11893768B2 (en) 2024-02-06

Similar Documents

Publication Publication Date Title
CN102032875B (zh) 一种基于图像处理的电缆护套厚度测量方法
CN104236498B (zh) 一种道岔钢轨件端面尺寸的测量方法
Zong et al. An intelligent and automated 3D surface defect detection system for quantitative 3D estimation and feature classification of material surface defects
WO2021227287A1 (zh) 生产线上蜂窝产品规整度的检测方法及检测装置
CN107392954A (zh) 一种基于序列图像的粗大误差点剔除方法
WO2021227285A1 (zh) 蜂窝结构几何规整度图像识别方法及系统
CN111583237B (zh) 蜂窝规整度检测的邻域窗口递推胞元重构方法
CN111583243B (zh) 蜂窝规整度检测的相邻点胞元重构方法
CN111583235B (zh) 蜂窝规整度检测的分支点识别顶点提取方法及其系统
CN111583239B (zh) 蜂窝结构几何规整度图像识别方法及系统
CN111583241B (zh) 超大面积蜂窝产品规整度的移动式检测方法及装置
WO2021227288A1 (zh) 蜂窝产品的几何形态评估方法和检测及评估系统
Khaloo et al. Extracting structural models through computer vision
CN111524133B (zh) 蜂窝规整度检测的窗口统计顶点提取方法及其系统
WO2021227286A1 (zh) 超大面积蜂窝产品规整度的移动式检测方法及装置
CN111583242B (zh) 蜂窝规整度检测的形态学顶点提取方法及检测装置
Wang et al. Machine vision based roundness and cylindricity detection of bearing rollers
CN111583234B (zh) 蜂窝产品的几何形态评估方法和检测及评估系统
Li et al. Laser Strip Center Extraction Methodology for the Detection of Weld Seam
CN111536890A (zh) 蜂窝规整度检测的Harris顶点提取方法及其系统
Xu et al. On-line visual inspection system for backside weld of tailored blanks laser welding
CN117848238B (zh) 一种基于激光点云的轨道交通站台实际限界测量方法
CN113447419B (zh) 一种多孔介质连通孔隙结构单元的划分系统
CN113418926B (zh) 基于激光条中心线定位提取算法的轮对踏面损伤检测方法
CN115824070B (zh) 一种钢轨件尺寸测量方法及装置

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: 20935302

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: 20935302

Country of ref document: EP

Kind code of ref document: A1