A method for identifying macro-morphology defect types of an additive manufacturing dot matrix structure

Abstract

The application discloses a kind of additive manufacturing lattice structure macroscopic morphology defect type identification methods, comprising: obtaining vertical angle lattice cell image data;The lattice cell image data is calibrated and constructs lattice cell identification dataset, identifies single lattice cell image data;The single lattice cell image data is segmented and constructs lattice cell top arc region segmentation dataset, and four rod top arc regions of lattice cell are segmented out;The four rod top arc regions of lattice cell segmented out are handled, and cell feature information is extracted;Quantification is carried out to the extracted cell feature information, and quantitative index is determined;Based on the quantitative index, the defect type of the lattice cell image data is identified and classified.The application identifies lattice structure defect and quantifies defect type, effectively guarantees the printing quality and mechanical properties of large metal lattice.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of additive manufacturing lattice structure morphology defect identification, and particularly relates to a method for identifying macroscopic morphology defect types of an additive manufacturing lattice structure. BACKGROUND

[0002] With the development of science and technology, additive manufacturing technology has been deeply involved in people's production and life. Among them, the metal lattice structure has the structural characteristics of high porosity, lightweight, high strength, and functional characteristics of shock absorption, noise reduction, impact resistance and energy absorption. At present, the main additive process for large lattice structures is wire arc additive manufacturing (WAAM), which can eliminate the size limitations of parts and has lower cost and higher material utilization compared to metal powder technology. However, research has found that the WAAM forming process is highly sensitive to process parameters. The forming process is usually accompanied by gradient changes in multiple energy signals, resulting in significant thermal stress inside the part, which causes the formation of cracks, geometric faults and other defects. The existence of these defects seriously affects the mechanical properties of the lattice structure, and thus has a negative impact on the service life and ideal effect of the lattice. In the prior art, an industrial CT is used to scan and detect the internal structure to obtain a two-dimensional gray-scale image of the transverse fault section. According to the distribution characteristics of the gray-scale values of the pixels in the two-dimensional gray-scale image, it is determined whether there is an internal defect. In the prior art, a convolutional neural network and an industrial CT are combined to scan a multi-layer metal lattice structure material sample to obtain its three-dimensional structure and extract a two-dimensional gray-scale image of the transverse section. By learning the defect samples in the image, the defects are identified and located. The existing problems include that the generalization ability of the industrial CT is not ideal, the image is mainly identified by manual operation, the information dimension is low, the missed / incorrect detection rate is high, the time consumption is long, the cost is high, only the internal defects of the lattice are considered, and the detection of the macroscopic morphology defects of the lattice is not considered. Therefore, there is an urgent need for a method for identifying the macroscopic morphology defect types of an additive manufacturing lattice structure. SUMMARY

[0003] To solve the above technical problems, the present application provides a method for identifying the macroscopic morphology defect types of an additive manufacturing lattice structure, which identifies the defects of the lattice structure and quantifies the defect types, effectively ensuring the printing quality and mechanical properties of large metal lattices.

[0004] To achieve the above purpose, the present application provides a method for identifying the macroscopic morphology defect types of an additive manufacturing lattice structure, which includes:

[0005] Obtaining lattice cell image data at a vertical angle;

[0006] Calibrate the dot cell image data and construct a dot cell recognition data set to identify single dot cell image data;

[0007] Segment the single dot cell image data and construct a dot cell top arc region segmentation data set to segment the top arc region of the four rods of the dot cell;

[0008] Processing the segmented top arc region of the four rods of the dot cell to extract cell feature information;

[0009] Quantifying the extracted cell feature information to determine a quantization index;

[0010] Based on the quantization index, identifying and classifying the defect types of the dot cell image data.

[0011] Optionally, the dot cell image data of the vertical angle includes:

[0012] Measuring the printed dot structure, using a horizontal posture to scan according to a set S-shaped trajectory by a drone with a camera, using computer vision technology to frame the video data, if a complete dot cell is included in an image, the arc position of the four rods of the cell is obtained, and the center of the regular quadrilateral formed by the four vertices is taken as the calibration point, with the image center as the origin and a preset radius size, if the calibration point belongs to the inside of the circle, the cell image is the image taken vertically above, which is the dot cell image data of the vertical angle; if the calibration point does not belong to the inside of the circle, the image is discarded; if a number of different rods of a dot cell are included in adjacent image frames, but each does not include a complete cell, the image stitching technology is used to stitch adjacent image frames into dot cell image data of the vertical angle.

[0013] Optionally, the design process of the S-shaped trajectory includes:

[0014] Measuring the overall physical size of the dot structure, wherein the dot cells are printed according to the settings of the G code and are arranged equidistantly in turn, calculating the effective projection range according to the size of the dot cell and combining the pinhole model to obtain the length and interval of the S-shaped trajectory.

[0015] Optionally, identifying single dot cell image data includes:

[0016] Obtaining a number of dot cell image data taken vertically above by scanning, marking based on the dot cell image data through a label, setting the size of the input image, training the dot cell recognition target detection neural network to obtain the trained dot cell recognition target detection neural network, and identifying all single dot cell image data.

[0017] Optionally, the four-rod top arc region of the dot matrix cell includes:

[0018] The dot matrix cell image data is divided into a dot matrix cell rod top arc region, the size of the input image is set, a dot matrix cell rod top arc region segmentation network is trained, and a trained dot matrix cell rod top arc region segmentation network is obtained. The four-rod top arc region of the dot matrix cell is segmented out.

[0019] Optionally, the cell feature information extraction includes:

[0020] The segmented four-rod top arc region of the dot matrix cell is processed, the image is binarized and edge detected, the boundaries of the four-rod top arc regions are obtained, each rod top arc region is fitted into a standard circle, the centers of the four circles are connected, and a quadrilateral cell feature information is obtained.

[0021] Optionally, the determination of the quantitative indicators includes:

[0022] Indicator 1 is that the quadrilateral is a convex quadrilateral;

[0023] Indicator 2 is whether the ratio of the quadrilateral area to the dot matrix rod diameter is greater than 625;

[0024] Indicator 3 is whether the shortest side length of the quadrilateral is less than 5mm;

[0025] Indicator 4 is whether the longest side length of the quadrilateral is greater than 40mm;

[0026] Indicator 5 is whether the logarithmic value of the ratio of the longest side length to the shortest side length of the quadrilateral is greater than 0.9;

[0027] Indicator 6 is whether the diagonal of the quadrilateral is less than 5mm;

[0028] Indicator 7 is whether the diagonal of the quadrilateral is greater than 60mm;

[0029] Indicator 8 is whether the ratio of the two diagonals of the quadrilateral is 1.

[0030] Optionally, the identification of the defect type of the dot matrix cell image data includes:

[0031] The defect type of the dot matrix cell image data is identified. If none of the eight indicators is met, it is determined to be a normal cell; if indicator 1 is met and other indicators are not met, it is determined that the defect type is rod offset; if indicators 2 or 4 or 7 are met and other indicators are not met, it is determined that the defect type is rod missing; if indicators 3 or 5 or 6 are met and other indicators are not met, it is determined that the defect type is rod adhesion; and if indicator 8 is met and other indicators are not met, it is determined that the defect type is asymmetric dot matrix.

[0032] Technical effects of this invention: This invention discloses a method for identifying macroscopic morphological defect types in additive manufacturing lattice structures. Although some existing technologies and research can identify some defects in 3D printed products or manufacturing processes, research on macroscopic morphological problems of lattice structures is very scarce. Furthermore, this invention proposes specific quantitative indicators for determining whether lattice cells are qualified and the types of lattice cell defects, which is more scientific. This invention combines the S-shaped trajectory scanning technology of UAVs to achieve a high-precision method for capturing vertical lattice images. Through steps such as scanning methods, lattice cell identification, feature extraction, defect quantification, and defect classification, it achieves the purpose of identifying lattice structure defects and quantifying defect types, effectively ensuring the printing quality and mechanical properties of large metal lattices. Attached Figure Description

[0033] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0034] Figure 1 This is a flowchart illustrating the method for identifying macroscopic morphological defect types in additive manufacturing lattice structures according to an embodiment of the present invention.

[0035] Figure 2 This is a flowchart of the module for obtaining vertical lattice cell images according to an embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram illustrating how to determine whether a calibration point is within the judgment range according to an embodiment of the present invention;

[0037] Figure 4 This is a flowchart of the module for extracting cell feature information using computer vision technology, as described in an embodiment of the present invention.

[0038] Figure 5 This is a flowchart illustrating the classification of lattice defect types according to an embodiment of the present invention. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0041] like Figure 1 As shown, this embodiment provides a method for identifying macroscopic morphological defect types in additive manufacturing lattice structures, including:

[0042] measuring the dot matrix structure after printing to obtain dot cell image data at a vertical angle;

[0043] calibrating the dot cell image data and constructing a dot cell recognition data set to identify single dot cell image data;

[0044] segmenting the dot cell image data and constructing a dot cell top arc region segmentation data set to segment the top arc region of the four rods of the dot cell;

[0045] processing the segmented top arc region of the four rods of the dot cell to extract cell feature information;

[0046] quantifying the extracted cell feature information to determine a quantization index;

[0047] based on the quantization index, identifying and classifying the defect types of the dot cell image data.

[0048] Further, obtaining dot cell image data at a vertical angle includes:

[0049] measuring the dot matrix structure after printing, using a drone with a camera to scan according to a set S-shaped trajectory with a horizontal posture, using computer vision technology to frame the video data, if an image contains a complete dot cell, the arc position of the four rods of the cell is obtained, and the center of the regular quadrilateral formed by the four vertices is taken as the calibration point, with the image center as the origin and a preset radius size, if the calibration point belongs to the inside of the circle, the cell image is the image taken vertically above, which is the dot cell image data at a vertical angle; if the calibration point does not belong to the inside of the circle, the image is discarded; if the adjacent image frames contain several different rods of a dot cell, but each does not contain a complete cell, the image stitching technology is used to stitch the adjacent image frames into dot cell image data at a vertical angle.

[0050] Specifically, as Figure 2 To obtain the vertical dot cell picture flow chart, a drone carrying a camera is used to scan the dot matrix structure according to an S-shaped trajectory. The specific design of the S-shaped trajectory is as follows: the overall physical size of the dot matrix structure is measured, and the dot cells are printed according to the G code setting and arranged equidistantly in turn, such as ten rows and ten columns. Then, according to the size (x, y, z) of the dot cell and combining the pinhole model, the effective projection range is calculated, and the length and spacing of the S-shaped trajectory are obtained. For example, the preset flight height of the drone is H, and the focal length f of the drone camera and the image sensor size (width W and height H s ) are known.

[0051] The UAV vertically takes a picture, the distance between the image plane and the camera is the focal length f, so the position of the image plane is:

[0052] Z = H - f,

[0053] According to the pinhole model, the projection of an object on the image plane can be calculated by the following formula:

[0054]

[0055] Where (X, Y) is the coordinate of the dot cell in the world coordinate system, and (x, y) is the coordinate of the dot cell on the image plane.

[0056] Subsequently, the width W of the effective projection range is calculated p :

[0057]

[0058] Where d is the diagonal length of the image sensor, FOV is the angle of view of the camera, and W is the width of the image sensor.

[0059] Based on the UAV sensor and flight control, the UAV can maintain the body as horizontal as possible during shooting. For example, using a rotor UAV, the rotation speed of the UAV can be measured by the built-in gyroscope, so as to judge the attitude. If the gyroscope reading shows that the UAV is continuously rotating or tilting, it means that the UAV is not in a horizontal attitude; when the UAV is in a non-horizontal attitude, the rotation speed of the propeller is changed by adjusting the rotation speed of the multiple motors, so as to change the lift, and then achieve the purpose of horizontal flight control. Based on this, the UAV completes the shooting task in the end according to the S-shaped trajectory and keeps the horizontal attitude all the time.

[0060] Further, the single dot cell image data includes:

[0061] A plurality of dot cell image data taken at a right angle above are obtained by scanning, based on the dot cell image data, the dot cell recognition target detection neural network is trained by marking with labels and dividing training set, verification set and test set, setting the size of the input image, and the trained dot cell recognition target detection neural network is obtained, and all single dot cell image data is recognized.

[0062] Specifically, the video stream data shot by the unmanned aerial vehicle is processed by using computer vision technology to take pictures frame by frame. First, it is analyzed whether a complete dot matrix cell is contained in a single frame picture. The specific judgment process is as follows: first, the dot matrix cell is identified by the dot matrix cell identification algorithm, and then the target tracking algorithm based on feature points is used to track in several consecutive frames to find the frame in which the center point of the cell is closest to the center point of the image, and the cell image is taken, combined with the flight data of the unmanned aerial vehicle and the sequence recording serial number (Xth row, Yth column). Based on the above process, a picture containing a complete dot matrix cell is obtained, the arc starting position of the four rods of the cell is obtained, and the center of the regular quadrilateral formed by the four vertices is taken as the calibration point. The image pixel size is converted into physical size, and the center of the image is taken as the origin, and a circle (such as 35mm radius) is drawn within a certain range. If the calibration point belongs to the inside of the circle, it is considered that the cell is the image taken directly above, i.e. the image without projection transformation; if the calibration point does not belong to the inside of the circle, the picture is discarded. For example Figure 3 A schematic diagram for judging whether the calibration point is within the judgment range.

[0063]

[0064] In the formula, O2 represents the calibration point, S O1 represents a circle with a certain range as the radius with the center of the image as the origin.

[0065] Secondly, if several different rods of a dot matrix cell are contained in adjacent frame pictures, but each does not contain a complete cell, then the image splicing technology is used to splice these adjacent image frames into a complete vertical angle dot matrix cell. The detailed process of the image splicing technology is as follows: first, feature detection, such as detecting the feature points of the four rods of the dot matrix cell, and using the SIFT feature detection algorithm to extract key points and descriptors from each cell image. Then, feature matching is performed, and by comparing the descriptors (such as vertical angle dot matrix cell images), matching point pairs between different images are found. Common methods include brute-force matching (Brute-Force Matching) and FLANN (Fast Library for Approximate Nearest Neighbors). In addition, the distance between the matching points also needs to be calculated, which is generally obtained by Euclidean distance:

[0066]

[0067] In the formula, x i and y i represent different matching point coordinates.

[0068] Subsequently, the matching points are screened by a ratio test (such as Lowe's ratio test) to improve the accuracy of matching. Image mapping is achieved through an image transformation structure. Then, the APAP algorithm is used to align the feature points of the dot cell. The graph cut method is used to automatically select appropriate cell images for splicing. Finally, the multi-band blending strategy is used to realize dot cell image fusion:

[0069] I fused (x,y)=α·I1(x,y)+(1-α)·I2(x,y),

[0070] where I1 and I2 are the transformed image and the target image, and a is the fusion weight. The specific weight setting will vary depending on the characteristics of the image.

[0071] A high-quality dot cell image dataset is obtained by shooting from a vertical angle. In order to meet the requirements of the training dataset and improve the learning and generalization ability of the model, data enhancement methods such as changing brightness, contrast, saturation, and rotating pictures are used to expand the samples in the original dataset. Finally, multiple dot cell images are obtained. In addition, the images are labeled, and the training set, validation set and test set are divided according to 7:2:1, and finally a complete dataset is obtained. The input image size is 500*500*3, or the input picture is resized to the same size. The small sample dot cell recognition target detection algorithm is used as the backbone network for training, such as Meta-DETR, UniT, etc., to obtain the optimal network hyperparameters. The trained dot cell recognition target detection neural network is used to recognize all dot cells in the image, with the input being the image and the output being the position of the dot cell in the image.

[0072] y=F(x),

[0073] where F represents the backbone neural network, such as the Meta-DETR model, x is the input image, and y is the detection result.

[0074] The process of converting pixel size to physical size: In order to ensure that the length of a certain area in the dot cell image taken by a camera with different resolutions and pixels remains consistent in physics, a physical size calibration method is used to unify the standard. First, the physical size needs to be calibrated. When taking an image, use an object with a known size as a calibration object (such as a standard ruler). Ensure that the calibrated dot cell is clearly visible in the image and on the same plane as the object being photographed. Then calculate the relationship between pixels and physical size:

[0075] Physical length = pixel length x physical length per pixel.

[0076] Further, the related feature information of the lattice cell rod top end arc region needs to be obtained to facilitate quantifying the defect lattice. First, the lattice cell image obtained is divided into a lattice cell rod top end arc region, and the training set, the verification set and the test set are divided in the same manner as 7:2:1, and finally a complete data set is obtained. The input image size is 500*500*3, or the input picture is resized to the same size. The zero-shot segmentation algorithm is used as the backbone network for training, such as the FastSAM and ViTDet models, to obtain the optimal network hyperparameters. The trained neural network is used to segment the lattice cell rod top end arc region from the image, and the input is the image and the output is the lattice cell rod top end arc region in the image. For example, the lattice cell rod top end arc region is assigned a value of 1 and the remaining regions are assigned a value of 0 to obtain a segmentation mask.

[0077] y = G(x),

[0078] where G represents a backbone neural network, such as the FastSAM model, x is an input image, and y is a segmentation result.

[0079] Further, the four rod top arc regions of the lattice cell are segmented, including:

[0080] The lattice cell image data is divided into a lattice cell rod top end arc region, and the training set, the verification set and the test set are divided. The size of the input image is set, and the lattice cell rod top end arc region segmentation network is trained to obtain the trained lattice cell rod top end arc region segmentation network, and the four rod top arc regions of the lattice cell are segmented.

[0081] Further, the cell feature information is extracted, including:

[0082] The segmented four rod top arc regions of the lattice cell are processed, the image is binarized and edge detection is performed to obtain the boundaries of the four rod top arc regions. Each rod top arc region is fitted into a standard circle, and the centers of the four circles are connected to obtain a quadrilateral cell feature information.

[0083] Specifically, as Figure 4 is a flowchart for processing the rod top arc region using computer vision technology, and the detailed processing process is shown as follows:

[0084] According to the segmentation mask, the image is binarized to facilitate obtaining the boundaries of the four rod top arc regions. The RGB value of each point in the image matrix is different. Binarization is to set a threshold value so that the RGB value of each point in the matrix becomes (0, 0, 0) [black] or (255, 255, 255) [white].

[0085] Edge detection is performed on the binary image. Edge detection aims to identify the regions in the image where the intensity changes significantly, i.e. edges. The edges here correspond to the boundary contours of the top arc regions of the rods. The main steps include five processes: smoothing, gradient calculation, non-maximum suppression, double thresholding, and edge linking. Among them, Gaussian filtering is often used for smoothing, and the Gaussian function is:

[0086]

[0087] where σ is the standard deviation, controlling the smoothing degree.

[0088] The Sobel operator is often used for gradient calculation, which has two direction convolution kernels:

[0089]

[0090] The gradient amplitude and direction are calculated:

[0091]

[0092] Each top arc region of the rod is fitted into a standard circle by the Hough circle detection algorithm. The Hough circle detection algorithm actually maps the points in the image space to the parameter space, thereby identifying the existence of a circle. Each edge point can be regarded as a candidate center of a circle, and the algorithm determines the final circle by accumulating the parameters of these candidate circles. For a circle, its parameters can be represented by three values: the center coordinates (a, b) and the radius r. Therefore, the parameter space of Hough circle detection is three-dimensional, represented as (a, b, r). Then, a voting mechanism needs to be used to select the best approximate circle. For each edge point (x0, y0), the Hough circle detection algorithm calculates all possible center and radius combinations. For each possible center (a, b), the following formula is used to calculate:

[0093]

[0094] A vote is added to each (a, b, r) combination in the parameter space, indicating that the combination is a possible circle. A three-dimensional accumulator is used to record the number of votes for each (a, b, r) combination. The higher the number of votes, the greater the likelihood of the circle.

[0095] For each edge point, update the accumulator:

[0096] accumulator[a][b][r] += 1,

[0097] By setting a threshold, find the point with the highest number of votes in the accumulator, i.e. (a*, b*, r*) is the parameter of the best detected circle:

[0098] argmax_abr = (a* b * ,r * ),

[0099] Finally, four circles are obtained, and the centers of the circles are connected to obtain a quadrilateral.

[0100] Further, the quantitative indicators include:

[0101] Indicator 1 is that the quadrilateral is a convex quadrilateral;

[0102] Indicator 2 is whether the ratio of the area of the quadrilateral to the diameter of the lattice rod is greater than 625;

[0103] Indicator 3 is whether the shortest side length of the quadrilateral is less than 5 mm;

[0104] Indicator 4 is whether the longest side length of the quadrilateral is greater than 40 mm;

[0105] Indicator 5 is whether the logarithmic value of the ratio of the longest side length to the shortest side length of the quadrilateral is greater than 0.9;

[0106] Indicator 6 is whether the diagonal of the quadrilateral is less than 5 mm;

[0107] Indicator 7 is whether the diagonal of the quadrilateral is greater than 60 mm;

[0108] Indicator 8 is whether the ratio of the two diagonals of the quadrilateral is 1.

[0109] Specifically, the judgment basis for whether the dot matrix is qualified and the defect type includes eight dot matrix defect quantitative indicators:

[0110] Indicator 1: whether the quadrilateral is a convex quadrilateral;

[0111] Indicator 2: whether the ratio of the area of the quadrilateral to the diameter of the lattice rod is greater than 625;

[0112]

[0113] Indicator 3: whether the shortest side length of the quadrilateral is less than;

[0114] l min <5mm,

[0115] Indicator 4: whether the longest side length of the quadrilateral is greater than;

[0116] l max >40mm,

[0117] Indicator 5: whether the logarithmic value of the ratio of the longest side length to the shortest side length of the quadrilateral is greater than 0.9;

[0118] v l =log(l min / lmax ) > 0.7,

[0119] Indicator 6: whether the diagonal of the quadrilateral is less than

[0120] diag min < 5mm,

[0121] Indicator 7: whether the diagonal of the quadrilateral is greater than

[0122] diag max > 60mm,

[0123] Indicator 8: whether the ratio of the two diagonals of the quadrilateral is 1;

[0124] v diag = diag1 / diag2≠ 1,

[0125] wherein diag1 and diag2 represent the two diagonals respectively.

[0126] Further, the type of the defect of the dot matrix cell image data is identified, including:

[0127] The type of the defect of the dot matrix cell image data is identified, if none of the eight indicators is met, it is determined as a normal cell; if indicator 1 is met and none of the other indicators is met, it is determined as a rod member offset; if indicator 2 or 4 or 7 is met and none of the other indicators is met, it is determined as a rod member missing; if indicator 3 or 5 or 6 is met and none of the other indicators is met, it is determined as a rod member bonding; if indicator 8 is met and none of the other indicators is met, it is determined as an asymmetric dot matrix.

[0128] Specifically, the types of the dot matrix cell defects are classified in combination with the defect quantification indicators. For example, Figure 5 is a flow chart for classifying the types of the dot matrix defects. The specific definitions of the types of the dot matrix defects are as follows:

[0129] Rod member offset: the stacking direction of one or more rod members in the dot matrix cell deviates from the preset direction to a great extent;

[0130] Rod member missing: one or more rod members in the dot matrix cell are broken;

[0131] Rod member bonding: the top arc regions of some rod members in the dot matrix cell are bonded together;

[0132] Asymmetric dot matrix: the four rod members of the dot matrix cell present an asymmetric shape.

[0133] If the above eight indexes are not met, the cell is determined to be normal, and the cell state is assigned as s(i)=0; if index 1 is met and other indexes are not met, the defect type is determined to be a rod offset, and the cell state is assigned as s(i)=1; if index 2 or 4 or 7 is met and other indexes are not met, the defect type is determined to be a rod missing, and the cell state is assigned as s(i)=2; if index 3 or 5 or 6 is met and other indexes are not met, the defect type is determined to be a rod sticking, and the cell state is assigned as s(i)=3; if index 8 is met and other indexes are not met, the defect type is determined to be an asymmetric lattice, and the cell state is assigned as s(i)=4.

[0134] The application discloses a kind of additive manufacturing lattice structure macroscopic morphology defect type identification method, although some technologies and research of existing can also identify some defects of 3D printing manufacturing product or manufacturing process, but the research for lattice structure macroscopic morphology problem is very rare, and the present application proposes specific determination lattice cell whether pass quantitative index and lattice cell defect type, more scientific;The application combines the S-shaped trajectory scanning technology of unmanned plane, realizes the method of high-precision vertical lattice image shooting, reaches the purpose of identifying lattice structure defect and quantifying defect type by scanning method, lattice cell identification, defect quantification, feature extraction, (feature extraction before defect quantification) defect classification and other steps, effectively guarantee the printing quality and mechanical properties of large metal lattice.

[0135] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be limited by the protection scope of claims.

Claims

1. A method for identifying macro-morphology defect types of an additive manufacturing dot array structure, characterized in that, The method comprises the following steps: Obtain dot cell image data at a vertical angle; Calibrate the dot cell image data and construct a dot cell recognition data set to identify single dot cell image data; Segment the single dot cell image data and construct a dot cell top arc region segmentation data set to segment the top arc region of the four rods of the dot cell; Process the segmented top arc region of the four rods of the dot cell to extract cell feature information; Quantify the extracted cell feature information to determine a quantitative index; Based on the quantitative index, identify and classify the defect types of the dot cell image data.

2. The method of claim 1, wherein the macro-topography defect type of the additive manufacturing dot-matrix structure is identified by: Obtaining dot cell image data at a vertical angle comprises the following steps: Measuring the printed dot structure, using a horizontal posture to scan the dot structure according to a preset S-shaped trajectory by a UAV with a camera, using computer vision technology to frame the video data, if an image contains a complete dot cell, obtaining the arc starting position of the four rods of the dot cell, and taking the center of a regular quadrilateral formed by four vertices as a calibration point, taking the center of the image as the origin and a preset radius size, if the calibration point is inside the circle, the dot cell image is an image taken vertically above, which is the dot cell image data at a vertical angle, if the calibration point is not inside the circle, the image is discarded, if the adjacent image frames contain several different rods of a dot cell, but each does not contain a complete dot cell, using image stitching technology to stitch the adjacent image frames into dot cell image data at a vertical angle.

3. The method of claim 1, wherein the macro-topography defect type of the additive manufacturing dot-matrix structure is identified by: The design process of the S-shaped trajectory comprises the following steps: Measuring the overall physical size of the dot structure, wherein the dot cells are printed according to the G code setting and arranged equidistantly, calculating the effective projection range according to the size of the dot cell and combining the pinhole model to obtain the length and interval of the S-shaped trajectory.

4. The method of claim 2, wherein the macro-topography defect type of the additive manufacturing dot-matrix structure is identified by: Identifying single dot cell image data comprises the following steps: Obtaining several dot cell images taken vertically above by scanning, marking the dot cell image data based on the dot cell image data, setting the size of the input image, training the dot cell recognition target detection neural network to obtain the trained dot cell recognition target detection neural network, and identifying all single dot cell image data.

5. The method of claim 1, wherein, Segmenting the top arc region of the four rods of the dot cell comprises the following steps: Dividing the dot cell image data into a dot cell rod top arc region, setting the size of the input image, training the dot cell rod top arc region segmentation network to obtain the trained dot cell rod top arc region segmentation network, and segmenting the top arc region of the four rods of the dot cell.

6. The method of claim 1, wherein, Extracting cell feature information comprises the following steps: Processing the segmented top arc region of the four rods of the dot cell, performing binaryzation processing and edge detection on the image to obtain the boundaries of the four rod top arc regions, fitting each rod top arc region into a standard circle, connecting the centers of the four circles to obtain a quadrilateral cell feature information.

7. The method of claim 1, wherein, Determining the quantitative index comprises the following steps: Index 1: the quadrilateral is a convex quadrilateral; Index 2: whether the ratio of the area of the quadrilateral to the diameter of the dot rod is greater than 625; Index 3 is whether the shortest side length of the quadrilateral is less than 5mm; Index 4 is whether the longest side length of the quadrilateral is greater than 40mm; Index 5 is whether the logarithmic value of the ratio of the longest side length to the shortest side length of the quadrilateral is greater than 0.9; Index 6 is whether the diagonal of the quadrilateral is less than 5mm; Index 7 is whether the diagonal of the quadrilateral is greater than 60mm; Index 8 is whether the ratio of the two diagonals of the quadrilateral is 1.

8. The method of claim 7, wherein the macro-topography defect type of the additive manufacturing dot-matrix structure is identified by: The defect type of the dot matrix cell image data is identified, including: The defect type of the dot matrix cell image data is identified, if none of the eight indexes is met, it is determined as a normal cell; if index 1 is met and other indexes are not met, it is determined as a rod member offset; if index 2 or 4 or 7 is met and other indexes are not met, it is determined as a rod member missing; if index 3 or 5 or 6 is met and other indexes are not met, it is determined as a rod member bonding; if index 8 is met and other indexes are not met, it is determined as an asymmetric dot matrix.

Images