Apparatus and method for detecting surface tear and crush defects in honeycomb material

Abstract

The application relates to the technical field of surface defect detection, and discloses a detection device and a detection method for surface tearing and crushing defects of honeycomb materials, the method comprising the following steps: extracting a morphological line diagram of a honeycomb material to be detected according to a to-be-detected image; splitting the morphological line diagram into a plurality of to-be-detected subgraphs with the same shape; extracting a to-be-detected feature vector of the to-be-detected subgraph through a trained feature extraction network; inputting the to-be-detected feature vector into a Gaussian model to obtain a Gaussian probability of the to-be-detected subgraph, wherein the Gaussian model is obtained by fitting feature vectors of a plurality of defect-free training subgraphs; restoring a plurality of to-be-detected subgraphs into the to-be-detected image, and marking and outputting a detection result of the restored to-be-detected image based on the Gaussian probability. The embodiment of the application realizes automatic detection of whether tearing and crushing defects exist on the surface of the honeycomb material to be detected, and improves the detection efficiency.

Description

Technical Field

[0001] This invention relates to the field of surface defect detection technology, specifically to a detection device and method for tearing and crushing defects on the surface of honeycomb material processing. Background Technology

[0002] Honeycomb material is a structural material creatively invented by humans through studying the characteristics of natural honeycomb structures. Its structure is periodic, consisting of thin-walled, regular polygonal frames arranged periodically. Unlike traditional materials, honeycomb material has a mesh-like surface and a hollow, discontinuous material body, giving it unique characteristics. The primary application of honeycomb material is in bonding it to upper and lower skins using adhesives to form a honeycomb sandwich structure. This structure works similarly to an I-beam; the honeycomb material, as the core, primarily bears normal stress, while the upper and lower skins primarily bear shear stress. This allows for high bending strength requirements with relatively low weight, significantly reducing the weight of structural components.

[0003] Currently, aircraft manufacturers have increased the use of honeycomb materials. Because honeycomb materials primarily serve as filler and need to bond with the upper and lower skins, their surface quality is crucial. Poor surface quality in the honeycomb material processing leads to a weak bond between the honeycomb material and the upper and lower skins, affecting not only the mechanical properties of the structural component but also significantly shortening its service life. Research institutions have studied the honeycomb manufacturing process and found that defects such as surface tearing and crushing can occur on the honeycomb surface during processing. These defects directly affect the bonding strength between the honeycomb material and the skin, reducing the service life of the honeycomb sandwich structure.

[0004] For defects such as surface tearing and crushing on honeycomb surfaces, current methods mainly rely on optical microscopy for imaging, followed by human observation and empirical judgment. This method depends heavily on manual observation, is highly subjective, and raises questions about repeatability. Furthermore, it involves significant manual labor, making it unsuitable for industrial production and resulting in low detection efficiency. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a detection device and method for detecting tearing and crushing defects on the surface of honeycomb material processing, so as to solve the technical problem of low efficiency of existing tearing and crushing defect detection methods.

[0006] In a first aspect, the present invention provides a method for detecting tearing and crushing defects on the surface of a honeycomb material, comprising: extracting a morphological line map of the honeycomb material to be detected from an image to be detected; dividing the morphological line map into several sub-images to be detected; extracting the feature vectors to be detected of the sub-images to be detected using a trained feature extraction network; inputting the feature vectors to be detected into a Gaussian model to obtain the Gaussian probability of the sub-images to be detected, wherein the Gaussian model is obtained by fitting the feature vectors of several defect-free training sub-images; restoring the several sub-images to be detected into an image to be detected, and marking and outputting the detection results of the restored image to be detected based on the Gaussian probability.

[0007] The method for detecting surface tearing and crushing defects in honeycomb materials according to embodiments of the present invention involves extracting a morphological line map of the honeycomb material to be detected from an image to be detected, dividing the morphological line map into several sub-images to be detected, extracting the feature vectors of the sub-images to be detected using a trained feature extraction network, and inputting the feature vectors to be detected into a Gaussian model to obtain the Gaussian probability of the sub-images to be detected. The Gaussian model is obtained by fitting the feature vectors of several defect-free training sub-images; therefore, the Gaussian probability represents the probability that surface tearing and crushing defects are not present. Furthermore, the several sub-images to be detected are restored to the image to be detected, and the detection result is marked and output based on the Gaussian probability. This achieves automatic detection of whether tearing and crushing defects exist on the surface of the honeycomb material to be detected, improving detection efficiency.

[0008] In one optional implementation, the morphology line diagram is divided into several sub-images of the same shape to be detected, including: obtaining the corner points of the honeycomb material to be detected in the morphology line diagram; and dividing the morphology line diagram into several sub-images of the same shape to be detected according to the position of the corner points.

[0009] The method for detecting tearing and crushing defects on the surface of honeycomb material processing according to the embodiments of the present invention divides the morphological line map into several sub-images of the same shape according to the position of the corner points. The sub-images of the ...

[0010] In one optional implementation, obtaining the corner points of the cellular material to be detected in the morphology line diagram includes: extracting intersection points in the morphology line diagram; constructing feature vectors of the intersection points; performing K-means++ clustering on the feature vectors of the intersection points to obtain accurate corner points; and predicting the remaining corner points in the morphology line diagram based on the accurate corner points, wherein the corner points of the cellular material to be detected in the morphology line diagram include accurate corner points and predicted corner points.

[0011] The method for detecting tearing and crushing defects on the surface of honeycomb material processing according to the present invention obtains accurate corner points through K-means++ clustering, and then predicts the remaining corner points in the morphological line diagram based on the accurate corner points. This avoids directly identifying inaccurate corner points due to inaccurate images obtained from the shooting environment, thereby ensuring the accuracy of the corner points.

[0012] In one optional implementation, constructing the feature vector of the intersection point includes: for each intersection point, obtaining the three nearest neighboring intersection points that are closest to the current intersection point; and constructing the feature vector of the current intersection point based on the variance of the distance from the current intersection point to the three nearest neighboring intersection points and the variance of the distance between each pair of the three nearest neighboring intersection points.

[0013] The spaces on the surface of the honeycomb material are regular polygons. The variance of the distance from each intersection point to the three adjacent intersection points, and the variance of the distance between each pair of the three adjacent intersection points, are normally 0. The feature vector constructed by this method can quickly identify normal intersection points.

[0014] In one optional implementation, extracting the morphological line map of the cellular material to be detected from the image to be detected includes: binarizing the image to be detected to obtain a binary image; extracting the skeleton from the binary image to obtain the morphological line map of the cellular material to be detected; and / or, marking and outputting the detection result of the restored image to be detected based on the Gaussian probability, including: converting the restored image to be detected into a heat map and outputting it based on the Gaussian probability.

[0015] Binarization removes noise and enhances contrast in the image to be detected. Skeleton extraction yields a morphological line map, facilitating subsequent recognition steps. The reconstructed image is then converted into a heatmap for visualization and easy observation.

[0016] Secondly, the present invention provides a detection device for surface tearing and crushing defects in honeycomb material processing, comprising: a three-coordinate displacement platform; a workpiece clamping part disposed in an area corresponding to the moving range of the execution end of the three-coordinate displacement platform, for mounting the honeycomb material to be detected; an image acquisition device disposed at the execution end of the three-coordinate displacement platform and moved along the three-dimensional coordinate direction by the three-coordinate displacement platform, for acquiring an image of the honeycomb material to be detected; and a control terminal connected to the three-coordinate displacement platform and the image acquisition device respectively, for controlling the three-coordinate displacement platform and the image acquisition device to acquire the image to be detected and execute the detection method for surface tearing and crushing defects in honeycomb material processing according to the first aspect or any corresponding embodiment described above.

[0017] The detection device for tearing and crushing defects on the surface of honeycomb material processing according to the present invention controls the three-coordinate displacement platform to drive the image acquisition device to move, thereby acquiring the image of the honeycomb material to be detected, and automatically identifying tearing and crushing defects on the surface of honeycomb material processing through the detection method, thereby improving the detection efficiency.

[0018] In one optional embodiment, the detection device for surface tearing and crushing defects in honeycomb material processing further includes: a camera clamping part, the camera clamping part including an installation part and a connecting part connected to each other, the installation part being disposed on the execution end of the three-coordinate displacement platform, and the connecting part being used to install the image acquisition device and adjust the tilt angle of the image acquisition device.

[0019] The image acquisition device is mounted by the camera clamp and the tilt angle of the image acquisition device can be adjusted to keep the tilt angle of the image acquisition device in the most suitable position for detecting the honeycomb material. This can eliminate shadows and make the acquired image easy to identify.

[0020] In one optional embodiment, the image acquisition device employs a structured light camera; and / or, a stage is further provided in the area corresponding to the movement range of the execution end of the three-coordinate displacement platform, the workpiece clamping part includes a bracket, a pressure block, and a base plate, the bracket is disposed above the stage, the base plate is disposed above the bracket, the pressure block is used to fix the bracket and the stage, the base plate is made of a light-absorbing material, and the base plate is used to place the honeycomb material to be tested.

[0021] In addition to acquiring two-dimensional image information, structured light cameras can also obtain effective height-related information, providing richer quantitative analysis capabilities for detecting tearing and crushing defects on processed surfaces. The stage and workpiece clamping part can fix the honeycomb material to be inspected, and the good light absorption characteristics of the base plate can avoid various random optical interferences throughout the optical inspection process.

[0022] Thirdly, the present invention provides a computer device, comprising:

[0023] The memory and the processor are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the detection method of the first aspect or any corresponding embodiment described above.

[0024] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions, the computer instructions being used to cause a computer to perform the detection method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of the structure of the detection device according to an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the structure of an image acquisition device according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the structure of the camera clamping part according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the workpiece clamping part according to an embodiment of the present invention;

[0030] Figure 5 This is a flowchart of a detection method according to an embodiment of the present invention;

[0031] Figure 6 This is a flowchart of another detection method according to an embodiment of the present invention;

[0032] Figure 7 This is a schematic diagram of the feature extraction network structure according to an embodiment of the present invention;

[0033] Figure 8 This is a diagram illustrating the detection results of an embodiment of the present invention;

[0034] Figure 9 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention;

[0035] Explanation of reference numerals in the attached figures:

[0036] 1-Second Y-axis motion module; 2-First Y-axis motion module; 3-X-axis motion module; 4-Z-axis motion module; 5-Stage; 6-Image acquisition device; 7-Camera clamping part; 71-Connecting part; 72-Mounting part; 8-Bracket; 9-Pressure block; 10-Base plate; 11-Honeycomb material to be tested; 901-Processor; 902-Memory; 903-Input device; 904-Output device. Detailed Implementation

[0037] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0039] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0040] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0041] Honeycomb materials primarily function as filler and need to be bonded to the upper and lower skins, making their surface quality crucial. Related research has revealed significant defects in honeycomb materials during the adhesive coating, molding, and curing stages. In the adhesive coating stage, misalignment defects appear on the honeycomb surface; in the molding stage, understretching defects occur; and in the curing stage, adhesive blockage defects appear. During the processing stage, defects such as tearing, burrs, cell deformation, and crushing can occur on the honeycomb surface. This application proposes a detection device and method for detecting tearing and crushing defects on the processed surface of honeycomb materials, based on the detection of these defects during the processing stage.

[0042] This invention provides a detection device for tearing and crushing defects on the surface of honeycomb material processing, such as... Figures 1 to 4As shown, it includes: a three-coordinate displacement platform; a workpiece clamping part, which is set in the area corresponding to the moving range of the execution end of the three-coordinate displacement platform, for mounting the honeycomb material 11 to be tested; an image acquisition device 6, which is set in the execution end of the three-coordinate displacement platform and is driven by the three-coordinate displacement platform to move along the three-dimensional coordinate direction, for acquiring the image to be tested of the honeycomb material 11; and a control terminal (not shown in the figure), which is connected to the three-coordinate displacement platform and the image acquisition device 6 respectively, for controlling the three-coordinate displacement platform and the image acquisition device 6 to acquire the image to be tested and to execute the detection method for tearing and crushing defects on the processing surface of honeycomb material in the embodiment of the present invention.

[0043] Specifically, the three-coordinate displacement platform can achieve independent movement in the X, Y, and Z directions within the workspace, meaning the end effector can reach any point in the workspace. The movement accuracy in the X, Y, and Z directions is 10 μm, and the minimum movement speed in the X, Y, and Z directions is 1 mm / s, with a maximum movement speed of 30 mm / s. For example, the three-coordinate displacement platform includes a first Y-axis motion module 2, a second Y-axis motion module 1, an X-axis motion module 3, and a Z-axis motion module 4. The first Y-axis motion module 2 and the second Y-axis motion module 1 are arranged parallel to each other. The two ends of the X-axis motion module 3 are mounted on the sliders of the first Y-axis motion module 2 and the second Y-axis motion module 1. The Z-axis motion module 4 is mounted on the slider of the X-axis motion module 3. All three modules—first Y-axis motion module 2, second Y-axis motion module 1, X-axis motion module 3, and Z-axis motion module 4—are driven by motors to rotate precision-ground ball screws. The precision-ground ball screws drive the sliders to move along their corresponding directions. The forward and reverse rotation of the motors is controlled by a control terminal.

[0044] The end effector of a coordinate measuring machine (CMM) platform refers to the end tool or device used to perform tasks on the CMM platform. It can be replaced or adjusted as needed to complete different tasks. Typically, these end effectors can be robotic arms, grippers, sensors, cameras, etc., and their function depends on the platform's design and application. For example, in manufacturing, the end effector can be a robotic arm used for assembly, processing, and packaging tasks on automated production lines. In this embodiment of the invention, the end effector is a tool holder on a machine tool processing the honeycomb material 11 to be inspected. When inspection is not required, the CMM platform moves the tool holder to various positions on the honeycomb material to process it. After processing, an image acquisition device 6 is installed on the tool holder. The CMM platform moves the image acquisition device 6 to various positions on the honeycomb material to acquire images of each area to be inspected. The image acquisition device 6, installed on the machine tool's tool holder, has the potential to develop corresponding in-machine inspection applications and is more practical than traditional off-site sample image acquisition.

[0045] The control terminal can be a computer, mobile phone, or other terminal. The control terminal controls the movement of the three-coordinate displacement platform and acquires the image to be detected through the image acquisition device 6. Then, the image to be detected is detected by the detection method for tearing and crushing defects on the surface of honeycomb material processing in the above method embodiment to obtain the detection results of tearing and crushing defects.

[0046] The detection device for tearing and crushing defects on the surface of honeycomb material processing according to the present invention controls the three-coordinate displacement platform to drive the image acquisition device 6 to move, thereby acquiring the image of the honeycomb material 11 to be detected, and automatically identifying tearing and crushing defects on the surface of honeycomb material processing through the detection method of tearing and crushing defects on the surface of honeycomb material processing, thereby improving the detection efficiency.

[0047] In an optional embodiment, the detection device for tearing and crushing defects on the surface of honeycomb material processing further includes: a camera clamping part 7, which includes an installation part 72 and a connecting part 71 connected to each other. The installation part 72 is disposed on the execution end of the three-coordinate displacement platform, and the connecting part 71 is used to install the image acquisition device 6 and adjust the tilt angle of the image acquisition device 6.

[0048] Specifically, the angles of the connecting part 71 and the mounting part 72 are adjustable. By adjusting the angle of the connecting part 71, the tilt angle of the image acquisition device 6 can be adjusted. For example, the connecting part 71 has a magnet inside, and the mounting part 72 is made of magnetic material, which allows for easy adjustment of the relative angle between the connecting part 71 and the mounting part 72. The image acquisition device 6 can be mounted on the connecting part 71 using bolts or other connecting parts. When the connecting part 71 rotates, the tilt angle of the image acquisition device 6 also rotates accordingly. In this embodiment, the tilt angle of the image acquisition device 6 is 12°. At this angle, the tilt angle of the image acquisition device 6 is parallel to the axis of the honeycomb grid, which can eliminate shadows, and the image captured at this angle is optimal.

[0049] The image acquisition device 6 is mounted by the camera clamping part 7, and the tilt angle of the image acquisition device 6 can be adjusted to keep the tilt angle of the image acquisition device 6 in the most suitable position for detecting the honeycomb material. The tilt angle of the image acquisition device 6 is parallel to the axis of the honeycomb pores, which can eliminate shadows and make the acquired image to be detected easier to identify.

[0050] In one alternative implementation, the image acquisition device 6 employs a structured light camera.

[0051] Specifically, the structured light camera can capture high-resolution grayscale images of the surface of the honeycomb material. The output of a single acquisition is a grayscale image of a certain field of view at the current position of the structured light camera. Preferably, the working distance of the structured light camera is 130mm, the single acquisition time is 200ms, the camera field of view is 30mm×16mm, and the image resolution is 2160×4096.

[0052] Using a structured light camera, in addition to acquiring two-dimensional image information, effective height-direction information can be obtained, eliminating some outlier data points. The image information is closer to the real material, resulting in better training of the feature extraction network and subsequent detection performance. This provides richer quantitative analysis capabilities for detecting tearing and crushing defects on the processed surface. Existing technologies using optical microscopes can only study certain locations on the processed surface of honeycomb materials, and the images captured by optical microscopes are only two-dimensional information. However, the embodiments of this invention, by using a three-coordinate displacement platform to move the image acquisition device 6, can complete the characterization of any location on a larger honeycomb material surface, making the measurement of large honeycomb parts more effective.

[0053] In one optional embodiment, a stage 5 is also provided in the area corresponding to the movement range of the execution end of the three-coordinate displacement platform. The workpiece clamping part includes a bracket 8, a pressure block 9 and a base plate 10. The bracket 8 is located above the stage 5, and the base plate 10 is located above the bracket 8. The pressure block 9 is used to fix the bracket 8 and the stage 5. The base plate 10 is made of light-absorbing material and is used to place the honeycomb material 11 to be tested.

[0054] Specifically, the area corresponding to the movement range of the execution end of the three-coordinate displacement platform is located below the three-coordinate displacement platform, that is, the stage 5 is set below the three-coordinate displacement platform, the bracket 8 and the base plate 10 are set on the stage 5 in sequence, and the honeycomb material 11 to be tested is fixed by the stage 5 and the workpiece clamping part. When detecting tear and crush defects on the processing surface of the honeycomb material 11 to be tested, the honeycomb material 11 to be tested is placed on the base plate 10. The good light absorption characteristics of the base plate 10 can avoid various random optical interferences in the entire optical detection process.

[0055] The detection process of the detection equipment for surface tearing and crushing defects in honeycomb materials according to an embodiment of the present invention is as follows:

[0056] First, the honeycomb material 11 to be tested is placed on the base plate 10. The good light absorption characteristics of the base plate 10 can avoid various random optical interferences during the entire optical testing process.

[0057] Second, the Z-axis motion module 4 of the three-coordinate displacement platform carries a structured light camera for movement. By controlling the Z-axis motion module 4, the height of the structured light camera can be adjusted, thereby achieving focusing.

[0058] Third, once the structured light camera has focused, its field of view (length and width) is determined. Based on the surface area (length and width) of the honeycomb material 11 to be tested and its placement, the number of measurements required by the structured light camera and the coordinates of the position to be measured for each test can be calculated.

[0059] Fourth, the structured light camera needs to be moved to the position coordinates calculated in the previous step for each detection. The control terminal controls the first Y-axis motion module 2, the second Y-axis motion module 1, and the X-axis motion module 3 to move the structured light camera to the designated position.

[0060] Fifth, the structured light camera performs an inspection to acquire an image of the processed surface of the honeycomb material 11 under the current field of view. This image is a grayscale image. This grayscale image may contain tearing and crushing defects. The acquired image is saved in the control terminal.

[0061] Sixth, repeat steps four and five until all locations to be measured are completed. After the control terminal receives the image to be detected, it will execute the following steps.

[0062] Seventh, for a certain image to be detected, perform binarization processing, extract the skeleton from the binary image, obtain the morphological line map, and extract all accurate corner points from the acquired grayscale image information.

[0063] Eighth, based on the periodicity of the cellular material, all corner points are predicted accurately.

[0064] Ninth, based on the coordinates of the corner points, the morphological line graph is divided into several sub-graphs of the same shape to be detected.

[0065] Tenth, collect some defect-free sub-images to be inspected in advance. Randomly destroy a portion of these defect-free sub-images, then combine them with the undestroyed sub-images to be inspected. Figure 1 A training set is formed, a feature extraction network is trained, and a Gaussian model is fitted through the defect-free sub-image to be detected. The sub-image to be detected split in step nine needs to be input into the Gaussian model, and the Gaussian model will output a Gaussian probability. The Gaussian probability is marked on the sub-image to be detected, which represents the probability that the sub-image to be detected is defect-free.

[0066] Eleventh, the sub-image to be detected with Gaussian probability is restored to the original image to be detected according to its position at the time of splitting.

[0067] Finally, the test results are output in the form of a heat map to detect tearing and crushing defects.

[0068] In one embodiment, the pore side length of the honeycomb material 11 to be tested is approximately 2 mm, and the pore wall thickness is approximately 0.05 to 0.1 mm. The working distance of the structured light camera is approximately 130 mm, the exposure time is 40,000 microseconds, and the gain is 12. The negative logarithm of the Gaussian probability is used as the brightness of the output heatmap, and the final output heatmap is as follows. Figure 8 As shown, the location of the defect can be seen intuitively.

[0069] The detection device of this invention has the following advantages:

[0070] First, a structured light camera is used to acquire images of the objects to be inspected. The light source is stable, and the imaging effect is also very stable, achieving good results in detecting tears and defects.

[0071] Second, the motion is precise; the three-coordinate displacement platform can accurately reach various locations within the cell for detection.

[0072] Third, it has high detection efficiency. The detection time for each image captured by the structured light camera is only 200ms, and the detection position is automatically switched by moving the three-coordinate displacement platform, which improves the overall detection efficiency.

[0073] This invention also provides a method for detecting tearing and crushing defects on the surface of honeycomb material processing, such as... Figure 5 and Figure 6 As shown, it includes:

[0074] Step S101: Extract the morphological line diagram of the honeycomb material to be detected from the image to be detected.

[0075] Specifically, the image to be inspected is obtained by capturing the surface of the honeycomb material under inspection using a structured light camera. The honeycomb image captured by the structured light camera is a high-resolution grayscale image. Using a structured light camera, in addition to acquiring two-dimensional image information, it can also obtain effective height-direction information, which helps to provide richer quantitative analysis capabilities for detecting tearing and crushing defects on the processed surface. The honeycomb material is composed of several regular polygonal structures, such as regular hexagons. In the image to be inspected obtained by capturing the surface of the honeycomb material under inspection using a structured light camera, the surface of the honeycomb under inspection is composed of lines formed by several regular hexagons. The morphological line graph includes the main lines in the image to be inspected. By extracting the morphological line graph, irrelevant information can be removed to obtain the image information of the surface of the honeycomb under inspection.

[0076] Step S102: Divide the morphological line graph into several sub-graphs to be detected.

[0077] For example, the resolution of the morphological line graph is 2160×4096. This 2160×4096 morphological line graph is divided into several sub-images according to a set size or a set shape. The set size can be one-thousandth or one-hundredth of the original image, depending on the actual captured image. The set shape refers to the hexagonal units that make up the morphological line graph, including the number of hexagons and the angle between any two adjacent hexagons. The sub-images to be detected after division have simple features. Sub-images with simple features have better training performance for feature extraction networks, and the extracted feature vectors are more representative, thereby improving recognition accuracy.

[0078] Step S103: Extract the feature vector to be detected from the sub-image to be detected using the trained feature extraction network. For example... Figure 7 As shown, the feature extraction network in this embodiment of the invention includes an input layer, two convolutional pooling layers, a flattening layer, a fully connected layer, a feature vector layer, and an output layer. The input of this network is an image, and the output is a 64-dimensional feature vector of the input image. The feature extraction network is trained using a pre-constructed training set to obtain a trained feature extraction network. The pre-constructed training set includes pre-collected and screened defect-free training sub-images and several defective training sub-images. The defect-free training sub-images and the sub-images to be detected undergo the same processing, both being obtained by splitting morphological line graphs and having the same size. The defective training sub-images are obtained by randomly selecting and destroying some defect-free training sub-images.

[0079] Step S104: Input the feature vector to be detected into the Gaussian model to obtain the Gaussian probability of the sub-image to be detected. The Gaussian model is obtained by fitting the feature vectors of several defect-free training sub-images.

[0080] Specifically, a defect-free training subgraph is input into a trained feature extraction network to obtain a corresponding 64-dimensional feature vector. Then, a Gaussian model is fitted to the 64-dimensional feature vector of the defect-free training subgraph. This Gaussian model reflects the characteristics of the feature vector of the defect-free training subgraph in space and can be used to determine whether the subgraph to be detected is normal, i.e., whether it has defects. The feature vector to be detected of the subgraph to be detected is also a 64-dimensional vector. After the feature vector to be detected is input into the Gaussian model, the Gaussian model judges the similarity between the feature vector to be detected and the feature vector of the defect-free training subgraph to obtain the Gaussian probability. The Gaussian probability represents the probability that the subgraph to be detected does not have defects.

[0081] Step S105: Reconstruct several sub-images to be detected into an image to be detected, and mark and output the detection results of the reconstructed image based on Gaussian probability.

[0082] Specifically, the restored image to be detected consists of several sub-images labeled with Gaussian probabilities. The result label can be the Gaussian probability output by the sub-image to be detected or the label corresponding to the Gaussian probability. For example, a Gaussian probability greater than a set threshold is red, and a Gaussian probability less than a set threshold is green. By outputting the image to be detected with the labeled results, the detection results of the surface of the honeycomb material to be detected in the image to be detected can be obtained, thereby determining whether there are tearing and crushing defects on the surface of the honeycomb material to be detected and the location of the defects.

[0083] The method for detecting surface tearing and crushing defects in honeycomb materials according to embodiments of the present invention involves extracting a morphological line map of the honeycomb material to be detected from the image to be detected, dividing the morphological line map into several sub-images of the same shape to be detected, extracting the feature vectors of the sub-images to be detected using a trained feature extraction network, and inputting the feature vectors to be detected into a Gaussian model to obtain the Gaussian probability of the sub-images to be detected. The Gaussian model is obtained by fitting the feature vectors of several training sub-images without defects, so the Gaussian probability is the probability that there are no surface tearing and crushing defects. Furthermore, the several sub-images to be detected are restored to the image to be detected, and the detection result is marked and output based on the Gaussian probability, thereby realizing the automatic detection of whether there are tearing and crushing defects on the surface of the honeycomb material to be detected, and improving the detection efficiency.

[0084] In some optional implementations, step S102, which involves splitting the morphological line graph into several sub-graphs of the same shape to be detected, includes:

[0085] Step S1021: Obtain the corner points of the honeycomb material to be detected in the morphology line diagram.

[0086] Specifically, the corner points of the honeycomb material to be tested refer to the corner points of each hexagon that makes up the honeycomb material to be tested. These corner points are obtained by identifying the morphological line diagram.

[0087] Step S1022: Based on the position of the corner points, the morphological line graph is divided into several sub-graphs of the same shape to be detected.

[0088] Specifically, based on the periodicity of the cellular material itself, its morphological line diagram can be composed of several identical sub-units, each including one or more regular hexagons. Therefore, the morphological line diagram can be divided into several sub-images of the same shape for detection based on the position of the corner points. Each sub-image includes one or more regular hexagons, and the connection angles of the regular hexagons in each sub-image are the same, meaning each sub-image has the same shape. For example, each sub-image may consist of a single regular hexagon, or two interconnected regular hexagons. Dividing the sub-images based on the corner point position, compared to directly dividing them by size, results in sub-images with identical lines and shapes, leading to more singular features. Sub-images with singular features result in better training of the feature extraction network, extracting more representative feature vectors that are easier to identify distinguishing points, thus improving recognition accuracy.

[0089] In some optional implementations, step S1021, obtaining the corner points of the honeycomb material to be detected in the morphology diagram, includes:

[0090] Step a1: Extract intersection points from the morphology line image. Specifically, the extracted intersection points include accurate and inaccurate corner points in the honeycomb material. Inaccurate corner points are directly identified because the acquired image is inaccurate due to interference from the shooting environment. The characteristic of intersection points is that within a local neighborhood (3×3), the number of black-and-white changes of 8 pixels in an instantaneous clockwise direction is different from the number of black-and-white transformations of non-intersection points. Intersection points in the morphology line image are extracted based on this characteristic.

[0091] Step a2: Construct the feature vector of the intersection point.

[0092] Specifically, for each intersection point, the three nearest neighboring intersection points are obtained. A feature vector for the current intersection point is constructed based on the variance of the distances from the current intersection point to these three neighboring intersection points, and the variance of the pairwise distances between these three neighboring intersection points. Since the pores on the surface of the honeycomb material are regular polygons, the variances of the distances from each intersection point to its three neighboring intersection points, and the variances of the pairwise distances between these three neighboring intersection points, are normally 0. Therefore, the feature vector constructed using this method can quickly identify normal intersection points.

[0093] Step a3: Perform K-means+ clustering on the feature vectors of the intersection points to obtain the accurate corner points.

[0094] Specifically, the intersection points include accurate corner points and inaccurate corner points, i.e., normal and abnormal corner points. Setting the K value to 2 results in two classes. After clustering using K-means++, the class with smaller variance represents accurate corner points, because an ideal honeycomb grid is a regular hexagon with a variance of 0. The class with larger variance represents inaccurate corner points, which are discarded from the morphological line diagram.

[0095] Step a4: Predict the remaining corner points in the morphology line diagram based on the accurate corner points. The corner points of the cellular material to be detected in the morphology line diagram include both accurate corner points and predicted corner points.

[0096] Because cellular materials themselves have periodicity, all corner points in a morphological diagram can be predicted by accurately identifying the corner points and leveraging this periodicity. For example, the remaining three corner points can be predicted based on the three corner points of a regular hexagon.

[0097] The method for detecting tearing and crushing defects on the surface of honeycomb material processing in this embodiment of the invention obtains accurate corner points through K-means++ clustering, and then predicts the remaining corner points in the morphological line diagram based on the accurate corner points. This avoids directly identifying inaccurate corner points due to inaccurate images obtained from the shooting environment, thereby ensuring the accuracy of the corner points and improving the final recognition accuracy.

[0098] In some optional implementations, step S101, extracting the morphological line drawing of the honeycomb material to be detected from the image to be detected, includes:

[0099] Step S1011: Binarize the image to be detected to obtain a binary image.

[0100] Specifically, binarization converts a grayscale image into a binary image. Pixels with grayscale values ​​greater than a certain threshold are set as grayscale maxima, and pixels with grayscale values ​​less than this threshold are set as grayscale minima, thus obtaining a binary image. Binarization can remove noise from the image being tested and enhance contrast.

[0101] Step S1012: Extract the skeleton from the binary image to obtain the morphological line diagram of the honeycomb material to be detected.

[0102] Specifically, skeleton extraction, similar to image thinning, refers to removing a portion of points from a binary image while retaining the original shape of the remaining points, i.e., the image skeleton. The skeleton extraction algorithm specifically employs the Zhang-Suen algorithm. Before skeleton extraction, the binary image is filtered using several dilation-erosion algorithms to remove noise. After skeleton extraction, the binary image is pruned to retain only complete hexagonal line images, resulting in the morphological line image of the honeycomb material to be detected.

[0103] Binarization can remove noise from the image to be detected and enhance contrast. Skeleton extraction can obtain a morphological line map, which facilitates subsequent recognition steps.

[0104] In some optional implementations, step S105, which involves marking and outputting the detection result of the restored image to be detected based on Gaussian probability, includes: converting the restored image to be detected into a heatmap based on Gaussian probability and outputting it.

[0105] Specifically, the Gaussian probability of the sub-image to be detected is obtained in the previous detection process. This Gaussian probability is displayed as an integer between 0 and 255. The negative logarithm of the Gaussian probability is taken as the output brightness of the sub-image to be detected. For example, if the Gaussian probability is N, the brightness of the heatmap is -logN. This converts the reconstructed image to be detected into a heatmap for output. The final output heatmap is shown below. Figure 8 As shown, by converting the restored image to be detected into a heatmap and outputting it, a visual output is achieved for easy observation.

[0106] This invention also provides a computer device, such as... Figure 9As shown, the computer device includes one or more processors 901, memory 902, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 9 Take the 901 processor as an example.

[0107] Processor 901 may be a central processing unit, a network processor, or a combination thereof. Processor 901 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0108] The memory 902 stores instructions executable by at least one processor 901 to cause at least one processor 901 to perform the method shown in the above embodiments.

[0109] The memory 902 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 902 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 902 may optionally include memory remotely located relative to the processor 901, and these remote memories can be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0110] The memory 902 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 902 may also include a combination of the above types of memory.

[0111] The computer device also includes an input device 903 and an output device 904. The processor 901, memory 902, input device 903, and output device 902 can be connected via a bus or other means. Figure 9 Taking the example of a connection between China and Israel via a bus.

[0112] Input device 903 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 904 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0113] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0114] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for detecting tearing and crushing defects on the surface of honeycomb material processing, characterized in that, include: Extract the morphological line diagram of the honeycomb material to be detected from the image to be detected; The morphological line graph is divided into several sub-graphs to be detected; The detection feature vector of the sub-graph to be detected is extracted by the trained feature extraction network; The feature vector to be detected is input into a Gaussian model to obtain the Gaussian probability of the sub-image to be detected, wherein the Gaussian model is obtained by fitting the feature vectors of several defect-free training sub-images; The sub-images to be detected are restored to the image to be detected, and the detection results are marked and output based on the Gaussian probability of the restored image to be detected. The process of extracting the morphological line map of the honeycomb material to be detected from the image to be detected includes: The image to be detected is binarized to obtain a binary image; The skeleton is extracted from the binary image to obtain the morphological line diagram of the honeycomb material to be detected. Specifically, the morphological line graph is divided into several sub-graphs to be detected, including: Obtain the corner points of the honeycomb material to be detected in the morphological line diagram; Based on the position of the corner points, the morphological line graph is divided into several sub-graphs of the same shape to be detected; The process of obtaining the corner points of the honeycomb material to be detected in the morphological line diagram includes: Extract the intersection points from the morphological line graph; Construct the feature vector of the intersection point; K-means clustering is performed on the feature vectors of the intersection points to obtain the accurate corner points; The remaining corner points in the morphology line diagram are predicted based on the accurate corner points, wherein the corner points of the cellular material to be detected in the morphology line diagram include accurate corner points and predicted corner points; The construction of the feature vector of the intersection point includes: For each intersection point, obtain the three nearest neighboring intersection points that are closest to the current intersection point; The feature vector of the current intersection point is constructed based on the variance of the distances from the current intersection point to the three neighboring intersection points, and the variance of the distances between each pair of the three neighboring intersection points.

2. The detection method according to claim 1, characterized in that, The detection results are labeled and output based on the Gaussian probability of the restored image to be detected, including: The restored image to be detected is converted into a heatmap and output according to the Gaussian probability.

3. A device for detecting tearing and crushing defects on the surface of honeycomb material processing, characterized in that, include: Three-coordinate displacement platform; The workpiece clamping part is located in the area corresponding to the movement range of the execution end of the three-coordinate displacement platform, and is used to install the honeycomb material to be tested; An image acquisition device is installed at the execution end of the three-coordinate displacement platform and is driven by the three-coordinate displacement platform to move along the three-dimensional coordinate direction, for acquiring the image of the honeycomb material to be detected; The control terminal is connected to the three-coordinate displacement platform and the image acquisition device, respectively, and is used to control the three-coordinate displacement platform and the image acquisition device to acquire the image to be detected and execute the detection method for tearing and crushing defects on the surface of honeycomb material processing as described in claim 1 or 2.

4. The detection device according to claim 3, characterized in that, Also includes: The camera clamping part includes a mounting part and a connecting part that are connected to each other. The mounting part is disposed on the execution end of the three-coordinate displacement platform, and the connecting part is used to mount the image acquisition device and adjust the tilt angle of the image acquisition device.

5. The detection device according to claim 3, characterized in that, The image acquisition device uses a structured light camera; And / or, a stage is also provided in the area corresponding to the movement range of the execution end of the three-coordinate displacement platform. The workpiece clamping part includes a bracket, a pressure block and a base plate. The bracket is arranged above the stage, and the base plate is arranged above the bracket. The pressure block is used to fix the bracket and the stage. The base plate is made of light-absorbing material and is used to place the honeycomb material to be tested.

6. A computer device, characterized in that, include: The device includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the detection method according to claim 1 or 2.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the detection method of claim 1 or 2.

Images