A method and device for screening weld defects and a detection system thereof

Abstract

The present application relates to a kind of screening method of weld defect, screening device and its detection system, for the screening of battery weld defect, wherein screening method includes the following steps: S1: obtain weld depth map and show top cover, shell side wall, weld and weld defect chart of suspected defect, wherein at least one suspected defect in the weld defect chart is marked as suspected pinhole defect, at least one suspected defect is marked as suspected pit defect;S2: respectively to the suspected pinhole defect and the suspected pit defect are screened to determine whether the suspected pinhole defect and the suspected pit defect are true defect;S3: according to the number of true defect, judge whether the quality of the weld meets the requirements.The screening method of the present application can improve the accuracy of the depth value calculation of suspected pinhole defect and suspected pit defect, so as to more accurately determine whether the suspected defect is a true defect, to reduce the occurrence of missed or overkill.

Description

Technical Field

[0001] The present invention relates to the technical field of visual algorithm recognition, and in particular to a method for screening weld defects, a screening device for implementing the screening method, and a detection system including the screening device. Background Art

[0002] With the booming new energy vehicle market, the demand for energy storage continues to rise, and the application of lithium-ion batteries is becoming more and more extensive, with a very promising future. From the structural dimension, the existing lithium-ion batteries mainly include cylindrical cells, square cells and soft-pack cells. Taking the cylindrical cell as an example, Figure 1 As shown in FIG, a cylindrical battery cell generally includes a shell sidewall 2 and a top cover 1 covering the shell sidewall 2. During the battery assembly process, the top cover 1 and the shell sidewall 2 need to be welded to form a weld 3 at the junction of the two. During welding, due to the presence of air in the solder or improper welding operation, the weld 3 may form the following Figure 2 The pinhole defect A shown in (a) is as follows Figure 2 The pit defect B shown in (b) of the figure. Pinhole defects A may become a channel for battery electrolyte leakage, resulting in a reduction in the electrolyte inside the battery cell, affecting the normal chemical reaction of the battery cell, and thus reducing the battery cell's capacity and performance. Meanwhile, pit defects B weaken the mechanical strength of the weld, making the battery cell prone to cracking or deformation at the weld when subjected to external forces. It also damages the seal of the weld, allowing external air and moisture to more easily enter the battery cell, causing performance degradation and shortening the battery cell life. Therefore, after welding, the weld needs to be inspected to identify welds with pinhole defects A and pit defects B.

[0003] In the past, the detection and screening of weld defects relied primarily on manual observation, a relatively inefficient method. With the advancement of AI technology, it's now possible to mimic the human eye's ability to detect and identify defects using AI models, image recognition, and algorithms. Taking cylindrical battery cell weld inspection as an example, the general approach is as follows: First, weld images are captured and preprocessed. Then, an AI model is trained on the preprocessed images to identify pinhole defects (A) and pit defects (B). The trained AI model then identifies defects in the weld under inspection and filters suspected defects. Specifically, the AI ​​model extracts the suspected defect areas from the weld depth map to create a weld defect map. The model then calculates the topographical features of the suspected defects in the weld defect map, including their depth. The calculated depth values ​​are then compared with a preset depth threshold. If the values ​​do not meet the preset threshold, they are considered true defects; otherwise, if they meet the threshold, they are considered non-defects. The screening process is completed based on the judgment results.

[0004] However, the existing AI-based screening method still has obvious shortcomings. Although it improves the detection efficiency to some extent, the judgment result is often difficult to match compared with experienced manual judgment, and the accuracy still needs to be improved. Especially in the face of some complex and boundary fuzzy defect cases, the AI model is prone to misjudgment when identifying pinhole defects A and pit defects B, thereby causing frequent problems of missed detection or overkill. When missed detection occurs, some defective welds will escape detection and flow into the subsequent process, which will cause hidden dangers to the quality of the battery cell and even the reliability of the entire battery product. Overkill will cause the AI model to misjudge some welds that do not have defects or have defects within an acceptable range as defects, which not only reduces the yield and increases the production cost, but also causes unnecessary production delay, reduces production efficiency, disrupts the production rhythm, and affects the smoothness of the entire production process. SUMMARY

[0005] Therefore, the purpose of the present application is to overcome the defects or deficiencies of the prior art. On the one hand, a weld defect screening method is provided, which can improve the accuracy of calculating the depth values of suspected pinhole defects and suspected pit defects, thereby reducing the occurrence of missed detection or overkill during screening.

[0006] A weld defect screening method, comprising the following steps:

[0007] S1: Obtain a weld depth map and a weld defect map showing a top cover, a shell side wall, a weld, and a suspected defect, wherein at least one suspected defect in the weld defect map is marked as a suspected pinhole defect, and at least one suspected defect is marked as a suspected pit defect;

[0008] S2: Screen the suspected pinhole defects and the suspected pit defects respectively to determine whether the suspected pinhole defects and the suspected pit defects are true defects;

[0009] S3: Determine whether the quality of the weld meets the requirements according to the number of true defects.

[0010] On the other hand, based on the same technical concept, the present application also provides a weld defect screening device, comprising:

[0011] A pinhole processing module is configured to obtain a weld depth map and a weld defect map showing a top cover, a shell side wall, a weld, and a suspected defect, and screen the suspected defect marked as a suspected pinhole defect in the weld defect map to determine whether the suspected pinhole defect is a true defect;

[0012] The concave processing module is used for acquiring a weld defect image showing a top cover, a shell sidewall, a weld seam and a suspected defect, and screening the suspected defect marked as a suspected concave defect in the weld defect image to determine whether the suspected concave defect is a true defect.

[0013] The weld quality analysis module is used for determining whether the quality of the weld seam meets the requirement according to the number of the true defects.

[0014] In another aspect, the present application further provides a weld defect detection system, which comprises an image acquisition device, a defect acquisition device and any of the screening devices described above; the image acquisition device forms a weld depth image and transmits the weld depth image to the defect acquisition device, wherein the weld depth image shows a top cover, a shell sidewall, a weld seam at the junction of the top cover and the shell sidewall and a suspected defect on the weld seam; the defect acquisition device acquires a weld defect image showing the suspected defect from the weld depth image and transmits the weld defect image and the weld depth image to the screening device.

[0015] Compared with the prior art, the screening method and the screening device of the present application are based on the shape features of the pinhole defects and the concave defects, and the appropriate depth reference datum is selected for the suspected pinhole defects and the suspected concave defects respectively, so as to improve the accuracy of the depth value calculation, make the depth values of the suspected pinhole defects and the suspected concave defects calculated more close to the actual depth values, and thus more accurately determine whether the suspected defects are true defects, reduce the occurrence of missed killing or overkilling, and further determine whether the quality of the weld seam meets the requirement according to the number of the true defects, thereby improving the detection accuracy.

[0016] In order to better understand and implement, the present application is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a schematic view of an existing battery structure;

[0018] Figure 2 is a schematic view of different types of defects on the weld seam at the junction of the top cover and the shell sidewall of the battery: Figure 2 (a) in the figure is a weld defect image showing a pinhole defect; Figure 2 (b) in the figure is a weld defect image showing a concave defect;

[0019] Figure 3 is a schematic view of an embodiment of the detection system of the present application;

[0020] Figure 4 is a weld depth image acquired by the image acquisition device in the detection system of the present application;

[0021] Figure 5 isFigure 3 Flowchart of a screening method performed by a screening device in a detection system for suspected pinhole defects and suspected pit defects;

[0022] Figure 6 This is a structural schematic diagram of an embodiment of a pinhole processing module in a screening device of the present invention;

[0023] Figure 7 For Figure 4 Schematic diagram of the position of the ridge line of the weld calibration in the weld depth diagram shown;

[0024] Figure 8 for Figure 7 Edge contour image taken from the midplane K;

[0025] Figure 9 for Figure 8 Schematic diagram of the location of the ridge lines on the weld profile shown;

[0026] Figure 10 for Figure 9 Schematic diagram of the locations of the first and second inflection points;

[0027] Figure 11 Schematic diagram of suspected pinhole defects located in the weld area: Figure 11 (a) is the weld defect diagram when the suspected pinhole defect is located in the weld bead area; Figure 11 (b) is the plane K1 in Figure 11 The edge contour image taken from (a); Figure 11 (c) is a schematic diagram of the weld bead reference surface formed after the top cover, shell side wall and weld are not flattened;

[0028] Figure 12 Schematic diagram of one of the cases where the suspected pinhole defect is located in the inner edge area: Figure 12 (a) is the weld defect image when the suspected pinhole defect is located in the inner edge area; Figure 12 (b) is plane K 21 exist Figure 12 The edge contour image taken from (a);

[0029] Figure 13 This is a schematic diagram of another situation where the suspected pinhole defect is located in the inner edge area: Figure 13 (a) is the weld defect image when the suspected pinhole defect is located in the outer edge area; Figure 13 (b) is plane K 22 exist Figure 13 The edge contour image taken from (a);

[0030] Figure 14 Schematic diagram of a situation where a suspected pinhole defect is located in the outer edge area:Figure 14 (a) is a weld defect map when the suspected pinhole defect is located in the outer edge region in FIG. Figure 14 (b) is an edge profile map of the plane K3 intersecting (a) in FIG. Figure 14

[0031] Figure 15 Figure 6 The pinhole processing module shown in FIG.

[0032] Figure 16 FIG. 1 is a structural schematic diagram of a first embodiment of the pit processing module in the screening device of the present application.

[0033] Figure 17 Figure 16 FIG. 2 is a schematic diagram of a flattened reference surface formed after the suspected pit defect is flattened by the pit processing module shown in FIG. Figure 17 (a) is a schematic diagram of a weld with a suspected pit defect in FIG. Figure 17 (b) is a schematic diagram of a flattened reference surface formed after the top cover, the shell sidewall and the weld are flattened in FIG.

[0034] Figure 18 Figure 16 FIG. 3 is a flowchart of the process of flattening the suspected pit defect to determine whether the suspected pit defect is a true defect by the pit processing module shown in FIG.

[0035] Figure 19 FIG. 4 is a structural schematic diagram of a second embodiment of the pit processing module in the screening device of the present application.

[0036] Figure 20 Figure 19 FIG. 5 is a flowchart of the process of flattening and pit zoning the suspected pit defect to determine whether the suspected pit defect is a true defect by the pit processing module shown in FIG.

[0037] Figure 21 FIG. 6 is a schematic diagram of a demarcation line for dividing the inner region profile and the outer region profile on the edge profile. Figure 21 (a) is a weld depth map in FIG. Figure 21 (b) is an edge profile intersected by the plane K in FIG.

[0038] Figure 22 FIG. 7 is a schematic diagram of one of the cases when the suspected pit defect is located in the first region. Figure 22 (a) is a schematic diagram of obtaining the demarcation line between the inner region profile and the outer region profile in FIG. Figure 22 (b) is a schematic diagram of obtaining the first high point in FIG. Figure 22 (c) is a schematic diagram of obtaining the first inflection point reference line in FIG. ​​​​​

[0039] Figure 23 (a) is a schematic diagram for obtaining the second fold point reference line when the suspected pit defect is located in the second area;

[0040] Figure 24 (b) is a schematic diagram for obtaining the second fold point reference line when the suspected pit defect is located in the third area; Figure 24 (a) in FIG. 4 is a schematic diagram for obtaining the boundary line between the inner area profile and the outer area profile; Figure 24 (b) in FIG. 4 is a schematic diagram for obtaining the second high point; Figure 24 (c) in FIG. 4 is a schematic diagram for obtaining the third fold point reference line;

[0041] Figure 25 (a) is a schematic diagram for obtaining the second fold point reference line when the suspected pit defect is located in the second area;

[0042] Figure 26 FIG. 8 is a flowchart of a second embodiment of the pit processing module in the screening device of the present application, in which the suspected pit defect is subjected to pit partition processing to calculate the fold point depth value to determine whether the suspected pit defect is a true defect;

[0043] Figure 27 FIG. 6 shows a pit processing module according to the present application; Figure 19 FIG. 6 shows a pit processing module according to the present application, which obtains the end point reference line of the suspected pit defect; Figure 27 (a) in FIG. 6 is a weld defect image; Figure 27 (b) in FIG. 6 is an edge profile taken by a plane J;

[0044] Figure 28 FIG. 8 is a flowchart of a second embodiment of the pit processing module in the screening device of the present application, in which the suspected pit defect is subjected to pit partition processing to calculate the fold point depth value to determine whether the suspected pit defect is a true defect; DETAILED DESCRIPTION

[0045] In the current field of cylindrical battery weld seam detection, with the help of AI models and image recognition, algorithm and other advanced technical means, it has become an important detection method. The conventional idea is to first collect the images of pinholes and pits and other defects on the weld seam and pre-process the images (specifically, mark the images of different defect types), then use AI models to standardize the pre-processed images, train a large number of sample data, let the AI model learn the feature patterns of different defects, so as to have the ability to identify weld defects. After completing the training, the AI model extracts the area where the suspected defect is located on the pre-processed weld depth map, and further calculates the depth and other topographic features of the suspected defect in the area. Then the calculated depth value is compared with the preset threshold value, and according to the comparison result, it is determined whether the suspected defect is a true defect, and finally the screening of the weld defects is completed.

[0046] Although this AI-based screening method improves the detection efficiency to some extent, the judgment result is often difficult to fit the actual situation compared with experienced manual judgment, and the accuracy still needs to be improved, especially when facing some complex and boundary fuzzy defect situations, misjudgment often occurs when identifying pinhole defects, pits and other defects, thereby causing frequent problems of missing or over-killing.

[0047] The inventor of the present application found through research that the reason why the existing AI-based screening method is not accurate and prone to missing or over-killing is that the existing calculation of the depth value of the suspected pinhole defect and the suspected pit defect on the weld is by taking the vertical distance between the highest point and the lowest point of the suspected defect on the weld. However, since the highest point of the pinhole defect and the pit defect is usually located on the outer surface of the weld, and the outer surface of the weld of the cylindrical battery cell is a curved surface, the position of the highest point will change with the change of the curved surface, which means that the reference benchmark for calculating the depth value of the suspected defects located at different positions is different. The depth value obtained by this rough calculation method is often inaccurate, thus the defect cannot be accurately identified, and the missing or over-killing situation occurs.

[0048] Based on this, the screening method of the present application is based on the shape characteristics of the pinhole defect and the pit defect. For the suspected pinhole defect, different regions are divided along the width direction of the weld, and different reference surfaces are selected for the suspected defects located in different regions, and the depth value of the suspected pinhole defect is calculated based on the corresponding reference surface. For the suspected pit defect, the region where the suspected pit defect is located is demarcated, and the part outside the region is flattened to make the top cover, the shell and the part of the weld outside the region where the suspected pit defect is located coplanar to form a pit flattening reference surface, and then the depth value of the suspected pit defect is calculated based on the flattening reference surface, thereby improving the accuracy of the depth value calculation, making the depth value of the suspected pinhole defect and the depth value of the suspected pit defect more close to the actual depth value, so that whether the suspected defect is a true defect can be more accurately judged, and the missing or over-killing situation can be reduced. Further, according to the number of true defects, it can be judged whether the quality of the weld meets the requirements, and the detection accuracy can be improved.

[0049] Based on the above improvement idea, the scheme of the present application will be described in detail below in combination with the drawings.

[0050] Figure 3 The specific structure of one embodiment of the detection system of the present application is shown. As shown in Figure 3 The detection system includes an image acquisition device, a defect acquisition device and a screening device.

[0051] The image acquisition device is specifically a 3D depth camera, which is arranged near the detection station. When the image acquisition device scans the weld seam 3 at the junction of the top cover 1 and the shell side wall 2 at a certain angle, a weld seam depth map as shown in Figure 4 is obtained, which shows the top cover 1, the shell side wall 2, and the weld seam 3 at the junction of the top cover 1 and the shell side wall 2. Then the formed weld seam depth map is transmitted to the defect acquisition device. The defect acquisition device is specifically an AI model that has been trained by a standard, which cuts out the area of the suspected defect in the weld seam depth map, outputs a weld defect map, and identifies and labels the type of suspected defect. Then the weld defect map and / or the weld seam depth map are transmitted to the screening device for further analysis. The suspected defects include suspected pinhole defects and suspected pit defects. In actual production, suspected pinhole defects and suspected pit defects may exist on the same weld seam. When the positions of the suspected pinhole defects and the suspected pit defects are relatively close, the boundaries of the two are relatively blurred, and the AI model cannot separately cut out the suspected pinhole defects or the suspected pit defects. At this time, the weld defect map output by the AI model shows at least one suspected pinhole defect and at least one suspected pit defect. Here, since the image of the area where the suspected defect is cut out by the AI model is prior art, it will not be described here.

[0052] The screening device is used to calculate the defect depth value of the suspected defect in the obtained weld defect map, compare the defect depth value with the preset depth value, determine whether the suspected defect is a true defect according to the comparison result, and then determine whether the quality of the weld seam 3 meets the requirements according to the number of true defects. The screening device comprises:

[0053] The pinhole processing module is used to acquire the weld seam depth map and the weld defect map showing the top cover, the shell side wall, the weld seam, and the suspected defect, and screen the suspected defects marked as suspected pinhole defects in the weld defect map to determine whether the suspected pinhole defects are true defects.

[0054] The pit processing module is used to acquire the weld defect map showing the top cover, the shell side wall, the weld seam, and the suspected defect, and screen the suspected defects marked as suspected pit defects in the weld defect map to determine whether the suspected pit defects are true defects.

[0055] The weld quality analysis module is used to determine whether the quality of the weld seam meets the requirements according to the number of true defects. When the number of true defects is ≥1, i.e., there are pinhole defects and / or pit defects on the weld seam 3, it is determined that the quality of the weld seam 3 does not meet the requirements, and the screening ends; when the number of true defects is 0, i.e., there are no pinhole defects and pit defects on the weld seam 3, it is determined that the quality of the weld seam 3 meets the requirements, and the screening ends or continues to perform other detection items according to production needs.

[0056] Figure 5 is shownFigure 3 The specific flow of the screening method performed by the screening device on the suspected pinhole defects and the suspected pit defects in the detection system shown.

[0057] Specifically, Figure 6 An embodiment of the specific structure of the pinhole processing module in the screening device of the present application is shown. As shown in the figure, Figure 6 The pinhole processing module includes:

[0058] The pinhole partition unit is configured to perform pinhole partition processing on the weld defect image obtained along the width direction of the weld 3, and then determine the defect type to which the suspected pinhole defect belongs according to the region where the suspected pinhole defect is located.

[0059] Specifically, the pinhole partition processing includes the following steps:

[0060] As shown in the figure, Figures 7-9 The pinhole partition unit divides the weld 3 into an outer region close to the top cover 1 and an inner region close to the shell side wall 2 in the weld depth image obtained. h , and the ridge line L h

[0061] The ridge line L h is calibrated in the following way: m edge contours are taken at intervals along the extension direction of the weld 3 from the weld depth image, where m≥1, and in this embodiment, m=10. The edge contours taken from the weld depth image are composed of the top cover contour 1a, the shell side wall contour 2a, and the weld contour 3a. The plane K on which the edge contours are located is perpendicular to the extension direction of the weld 3, and the ridge line L h The height value of any point on the weld contour 3a in each edge contour is calculated at the point on the edge contour, and the point with the maximum height value is taken as the highest point E. Then, the highest points E on each edge contour are fitted to form the ridge line L h .

[0062] Here, the highest point E is obtained in the following way: as shown in the figure, Figure 1 The top cover 1 and the shell side wall 2 of the battery cell are perpendicular to each other, so when the edge contour is taken along the extension direction of the weld 3, the included angle between the top cover contour 1a and the shell side wall 2a is 90°. Here, the plane with an included angle of 45° between the top cover 1 and the shell side wall 2 is set as the height reference plane P, and the height value of any point on the weld 3 can be obtained by calculating the distance from the height value of the point to the height reference plane P. In this embodiment, the image acquisition device is a 3D depth camera, and in the weld depth image output by the 3D depth camera, the top cover 1 and the shell side wall 2 have been automatically set as inclined planes inclined at 45° relative to the horizontal plane, that is, Figure 9 ​The height reference plane P shown is parallel to the horizontal plane, so when the profile along the extension direction of the weld 3 is taken, it is presented as shown in the edge profile Figures 9-10 The vertical distance from any point on the weld profile 3a to the height reference plane P is the height value of the point, and the point with the maximum height value is the highest point E.

[0063] The determination method of the defect type to which the suspected pinhole defect belongs is:

[0064] n edge profiles are taken at intervals along the extension direction of the weld 3 from the weld defect image, where n≥1, and in this embodiment, n=3. The edge profiles taken from the weld defect image are also composed of the profile 1a, the shell side wall profile 2a, and the weld profile 3a, and the suspected pinhole defect profile is located on the weld profile 3a. The ridge line L h In Figure 9 The point on the edge profile shown is the highest point E, and the highest point E is used as a dividing point to divide the edge profile into an inner region profile 10a and an outer region profile 10b.

[0065] In each edge profile, a first fold point Q1 and a second fold point Q2 are determined, where the first fold point Q1 can be regarded as the dividing point of the weld profile 3a and the shell side wall profile 2a, which is located on the inner region profile 10a; the second fold point Q2 can be regarded as the dividing point of the weld profile 3a and the top cover profile 1a, which is located on the outer region profile 10b. Then the distance between the first edge and the second edge on both sides of the suspected pinhole defect located on the inner region profile 10a along the width direction thereof and the first fold point Q1 is calculated respectively to obtain a first inner distance value and a second inner distance value; and / or the distance between the two side edges of the suspected defect located on the outer region profile 10b and the second fold point Q2 is calculated respectively to obtain a first outer distance value and a second outer distance value. According to the first and second inner distance values and / or the first and second outer distance values obtained, the following determination is made:

[0066] If the first inner distance value and the second inner distance value of the suspected pinhole defect located in the inner region are both less than 0.2 mm, it is determined that the suspected pinhole defect is an inner edge defect; otherwise, if any one of the first inner distance value and the second inner distance value is greater than or equal to 0.2 mm, the suspected pinhole defect is a weld defect.

[0067] Similarly, if the first outer distance value and the second outer distance value of the suspected pinhole defect located in the outer region are both less than 0.2 mm, it is determined that the suspected pinhole defect is an outer edge defect; otherwise, if any one of the first outer distance value and the second outer distance value is greater than or equal to 0.2 mm, the suspected pinhole defect is a weld defect.

[0068] In other words, if the suspected pinhole defect neither belongs to the inner edge defect nor the outer edge defect, the suspected pinhole defect is determined as the weld defect. If the suspected pinhole defect profile crosses the inner region profile 10a and the outer region profile 10b, since the first and second inner distance values and the first and second outer distance values of the suspected pinhole defect profile are all greater than 0.2 mm, the suspected defect must belong to the weld defect.

[0069] When n≥2, for the same suspected defect, in the intercepted multiple edge profiles, it can be determined to belong to different position types, but as long as one of them is determined to belong to the weld defect, the suspected pinhole defect belongs to the weld defect. For example, in two of the edge profiles, the suspected pinhole defect is determined to belong to the inner edge defect, and in the other edge profile, it is determined to belong to the weld defect. The suspected pinhole defect is identified as a weld defect. In other words, for the same suspected pinhole defect, in the intercepted n edge profiles, only when all are determined to belong to the inner edge defect or all are determined to belong to the outer edge defect, the suspected pinhole defect is identified as the inner edge defect or the outer edge defect. Otherwise, the suspected pinhole defect is identified as the weld defect.

[0070] The first fold point Q1 is obtained in the following manner:

[0071] First, the points on the shell side wall profile 2a and the top cover profile 1a are fitted to form a first fitting line L1 and a second fitting line L2, respectively; then, the vertical distance between each point on the inner region profile 10a and the first fitting line L1 is calculated as the first distance value, and the difference between the first distance values of two adjacent points (referred to as the first distance difference value) is calculated; at the same time, the included angle between the tangent of each point on the inner region profile 10a and the first fitting line L1 is calculated to obtain the first included angle value;

[0072] Then, according to the calculated first distance value and first included angle value of each point, and the difference between the first distance values of any two adjacent points, the first fold point Q1 is determined, and the specific process is as follows:

[0073] First, as shown in Figure 10 The intersection O of the first fitting line L1 and the second fitting line L2 to the orthographic projection on the weld line profile 3a is the starting point S, and the search direction is towards the shell side wall profile 2a. It is judged whether the first distance values of the next five points are less than the first preset value:

[0074] (1) If yes, it is judged whether the first distance difference values of the next five points after the fifth point are greater than one fifth of the collection resolution, wherein the collection resolution is the collection resolution of the 3D depth camera:

[0075] (a) If the subsequent five first distance difference values are greater than one fifth of the collection resolution, it indicates an increasing state, and the fifth point is not a true fitting point, and the search continues;

[0076] (b) If the subsequent five first distance difference values are less than or equal to one fifth of the collection resolution, the fifth point is recorded as the first fitting point, and it is determined whether the previous five first distance difference values of the first fitting point are less than -1 / 5 of the collection resolution:

[0077] a) If yes, the search ends;

[0078] b) If no, the search continues until the next fitting point is found, which is defined as the second fitting point.

[0079] Then, the vertical distance between the first fitting point and the second fitting point is calculated. If the vertical distance between the first fitting point and the second fitting point is greater than 7 (pixel units), the second fitting point is taken as the pre-selected point. If the vertical distance between the first fitting point and the second fitting point is less than or equal to 7, the first fitting point is taken as the pre-selected point.

[0080] Finally, from the position of the pre-selected point to the highest point E, if the included angle between the tangent of the three consecutive points and the first fitting line L1 is greater than 0.08°, the pre-selected point is determined as the first fold point Q1.

[0081] Similarly, the second fold point Q2 is obtained in the following way:

[0082] First, the points on the shell side wall profile 2a and the top cover profile 1a are fitted to form a first fitting line L1 and a second fitting line L2, respectively. Then, the vertical distance between each point on the outer side area profile 10b and the second fitting line L2 is calculated to obtain a second distance value, and the difference value (hereinafter referred to as the second distance difference value) of the second distance values of adjacent two points is calculated. At the same time, the included angle between the tangent of each point on the outer side area profile 10b and the second fitting line L2 is calculated to obtain a second included angle value.

[0083] Then, according to the calculated second distance value and second included angle value of each point, and the difference value between the second distance values of any two adjacent points, the same search logic as determining the first fold point Q1 is executed to finally determine the second fold point Q2, and the specific search process is not repeated here.

[0084] The pinhole reference forming unit is configured to form a reference surface corresponding to a defect type to which the suspected pinhole defect belongs, wherein the reference surface is any one of a welding bead reference surface P1 corresponding to a welding bead defect, an inner side reference surface P2 corresponding to an inner side edge defect, and an outer side reference surface P3 corresponding to an outer side edge defect.

[0085] Specifically, for weld defects, the weld reference plane P1 is obtained by defining the location of the suspected defect in the obtained weld defect map as the target area. 目 , except the target area on the top cover 1, the shell side wall 2 and the weld 3 目 The position outside the target area is the non-target area. The top cover 1, the shell side wall 2 and the weld 3 are flattened so that the top cover 1, the shell side wall 2 and the weld 3 are coplanar. The non-target area after flattening is taken as the weld reference plane P1. It should be noted here that although the target area 目 is also flattened, but the target area 目 The depth information of each point in the image will be retained after flattening, and this depth information will be used for subsequent depth value calculations.

[0086] The following Figure 11 The process of forming the weld bead reference surface P1 by the pinhole reference forming unit is specifically described by taking one of the cases where the suspected pinhole defect is in the weld bead area as an example:

[0087] exist Figure 11 middle, Figure 11 (b) in the figure is the cross section K1. Figure 11 The edge contour captured on (a) in the figure is Figure 11 (c) in the figure shows the weld bead reference surface formed after the top cover, shell side wall and weld are not flattened. Figure 11 As shown in (a)-(c) in the figure, the location of the suspected pinhole defect in the weld defect map is defined as the target area. 目 , except for the target area in the weld defect diagram on the top cover 1, shell side wall 2 and weld 3 目 The areas outside the target area are defined as non-target areas. Next, the top cover 1, shell sidewall 2, and weld 3 are flattened to make them coplanar. The flattened non-target area is then used as the weld bead reference plane P1. Flattening a curved surface is known to be a well-known technique (e.g., "Flattening of Triangular Mesh Surfaces" by Song Wanzhong and Su Xianyu, Journal of Sichuan University), so this process will not be described in detail here.

[0088] For inner edge defects, the inner reference plane P2 is obtained by fitting the points on the shell side wall 2 to form a shell fitting surface, and taking the shell fitting surface as the inner reference plane P2.

[0089] The following Figures 12-13 The process of forming the inner reference surface P2 by the pinhole reference forming unit is specifically described by taking two cases of suspected pinhole defects in the inner edge area as examples:

[0090] Figure 12The suspected pinhole defect is shown in the inner edge region and on the outside of the first folding point Q1 (i.e. the side away from the shell side wall 2); Figure 13 The suspected pinhole defect is shown in the inner edge region and on the inside of the first folding point Q1 (i.e. the side close to the shell side wall 2). For the above two cases, the determination of the inner side reference surface P2 is the same, that is, the points on the shell side wall 2 in the weld defect image are fitted to form the inner side reference surface P2.

[0091] For the outer edge defect, the outer side reference surface P3 is formed in the same way as the weld bead reference surface P1, that is, it is obtained by flattening processing. Specifically, the position of the suspected pinhole defect in the weld defect image is divided into a target area area 目 , and the positions other than the target area on the top cover 1, the shell side wall 2 and the weld 3 are divided into a non-target area. The top cover 1, the shell side wall 2 and the weld 3 are flattened to make them coplanar, and the flattened non-target area is taken as the outer side reference surface P3.

[0092] The following will be described in detail with an example of one of the cases shown in Figure 14 where the suspected pinhole defect is in the outer edge region:

[0093] As shown in (a)-(b) in Figure 14 , the position of the suspected pinhole defect in the weld defect image is divided into a target area area 目 , and the positions other than the target area area 目 on the top cover 1, the shell side wall 2 and the weld 3 are divided into a non-target area. Then, the top cover 1, the shell side wall 2 and the weld 3 are flattened to make them coplanar, and the flattened non-target area is taken as the outer side reference surface P3.

[0094] The pinhole feature calculation unit is configured to calculate the perpendicular distance between each point on the suspected pinhole defect and the corresponding reference surface and take the maximum value to obtain the pinhole defect depth value. Here, each point on the suspected pinhole defect refers to each point within the target area area 目 (the target area area 目 is usually set to be larger than the area range of the orthographic projection of the suspected pinhole defect). When calculating the inner edge region defect, the suspected pinhole defect position in the weld defect image can also be divided into a target area area 目 as shown in Figures 12-13 .

[0095] The pinhole depth value analysis unit is used to compare the pinhole defect depth value with the preset depth value and determine whether the suspected pinhole defect is a true defect based on the comparison results. Specifically, the calculated depth values ​​of the weld bead defect, the inner edge defect depth value, and the outer edge defect depth value are compared with the corresponding pinhole depth preset values. If the pinhole defect depth value is greater than the preset pinhole depth value, the suspected pinhole defect is determined to be a true defect; otherwise, it is determined to be a non-defect. Here, different depth preset values ​​can be set for these three defect types, or the same depth preset value can be set, depending on actual production needs.

[0096] Figure 15 Shown Figure 6 The pinhole processing module shown in FIG. 1 is a specific process of the screening method performed by performing pinhole partitioning processing on suspected pinhole defects. Figure 15 As shown, for suspected pinhole defects, the suspected pinhole defects are divided into three defect types according to the location of the suspected pinhole defects on the weld 3, namely weld bead defects, inner edge defects and outer edge defects. Different reference planes are formed for suspected pinhole defects at different positions to calculate their depth values, so that the calculated pinhole defect depth value is closer to the actual depth value, and then the calculated pinhole defect depth value is compared with the preset depth value, which can more accurately determine whether the suspected pinhole defect is a true defect. Compared with the existing method of calculating the depth value by taking the vertical distance between the highest point and the lowest point of the suspected pinhole defect regardless of the location of the suspected pinhole defect, the present invention distinguishes suspected pinhole defects at different positions, selects a more appropriate reference base according to the different position types of the suspected pinhole defects, and regardless of the weld bead reference plane P formed after the flattening process 1、 Neither the outer reference plane P3 nor the inner reference plane P2 formed after fitting processing will change with the change of the curved surface. Therefore, the calculated depth value of the suspected defect is closer to the actual depth value, thereby improving the accuracy of screening suspected pinhole defects and reducing the occurrence of missed or over-killed defects.

[0097] Figure 16 FIG. 2 shows the specific structure of the first embodiment of the pit processing module in the screening device of the present invention. Figure 16 As shown, the pit treatment module includes:

[0098] The pit reference forming unit is used to obtain a weld defect map showing the top cover 1, the shell side wall 2, the weld 3 and the suspected pit defect, and define the location of the suspected pit defect in the weld defect map as the target area. 目 , except the target area on the top cover 1, shell side wall 2 and weld 3 目The non-target area is determined by the position outside the target area, and the top cover 1, the shell side wall 2 and the weld 3 are flattened, and the flattened non-target area is taken as the flattened reference surface P 展 As shown in Figure 17 ).

[0099] The pit feature calculation unit is configured to calculate the vertical distance between each point on the suspected pit defect and the flattened reference surface P 展 , and take the maximum value to obtain the pit defect depth value.

[0100] The pit depth value analysis unit is configured to compare the pit defect depth value with the pit depth preset value, and determine whether the suspected pit defect is a true defect according to the comparison result: if the pit defect depth value is greater than the pit depth preset value, it is determined that the suspected pit defect is a true defect; otherwise, it is determined to be a non-defect.

[0101] Figure 18 The pit processing module shown in Figure 16 determines the specific process of the suspected pit defect by flattening the suspected pit defect to determine whether the suspected pit defect is a true defect.

[0102] In this way, the target area area 目 is formed at the position of the suspected pit defect, and the top cover 1, the shell side wall 2 and the weld 3 are flattened to make them coplanar, and the non-target area outside the flattened target area area 目 is taken as the flattened reference surface P 展 , and the depth value of the suspected pit defect is calculated based on the flattened reference surface P 展 . Compared with the existing method of calculating the depth value by taking the vertical distance between the highest point and the lowest point of the suspected pit defect, the flattened reference surface P 展 obtained after flattening in the depth value calculation method of the present application does not change with the change of the curved surface, so that the depth value of the suspected pit defect calculated is closer to the actual depth value, thereby improving the accuracy of the suspected pit defect screening and reducing the occurrence of missed or over-killing.

[0103] The inventors of the present application have found in practice that, since the pit can have a slowly changing slope surface which cannot be identified, the flattening processing is flattened, thus easily causing missed killing. To avoid this situation, for the suspected pit defect, the screening device of the present application, in addition to the suspected pit defect, performs partition processing on the suspected pit defect, divides the inner and outer regions of the weld 3 along the width direction of the weld 3 through the partition processing, and forms different reference bases for the suspected pit defects located at different positions to calculate the depth value, then compares the depth value calculated after the partition processing with the depth value calculated after the flattening processing and takes the maximum value, that is, the depth value of the suspected defect, and then compares the calculated pit defect depth value with the pit depth preset value.

[0104] To achieve the above technical concept, Figure 19 The specific structure of the second embodiment of the pit processing module in the screening device of the present application is shown. As shown in Figure 19 The pit processing module of the screening device includes two pit reference forming units, namely a first pit reference forming unit and a second pit reference forming unit; two pit feature calculation units, namely a first pit feature calculation unit and a second pit feature calculation unit; and a pit feature analysis unit. Among them, the first pit reference forming unit and the first pit feature calculation unit jointly realize the calculation of the first pit defect depth value of the suspected pit defect through flattening processing; the pit partition unit, the second pit reference forming unit and the second pit feature calculation unit jointly realize the calculation of the second pit defect depth value of the suspected pit defect through pit partition processing. Figure 20 The specific process of the pit processing module shown in Figure 19 The specific process of the pit processing module shown in Figure 20 As shown, the first pit defect depth value calculated by the flattening processing is compared with the second pit defect depth value calculated by the pit partition processing, and the larger value is taken to obtain the pit defect depth value; then the pit defect depth value is compared with the pit depth preset value, and whether the suspected pit defect is a true defect is determined according to the comparison result. In this way, by forming reference bases for suspected pit defects in two different ways, the finally calculated pit defect depth value is closer to the true value, so that whether the suspected defect is a true defect can be more accurately judged, so as to reduce the occurrence of missed killing.

[0105] Specifically, the crater partition unit is configured to acquire a weld defect image showing the top cover 1, the shell sidewall 2, the weld seam 3, and the suspected crater defect, and to obtain an edge profile of the suspected crater defect in the acquired weld defect image along the extension direction of the weld seam, and then to perform crater partition processing on the obtained edge profile along the width direction of the edge profile, and to determine the position type of the suspected crater defect according to the region where the suspected crater defect is located.

[0106] The specific process of the crater partition processing will be described below with reference to the weld seam 3 shown in (a)-(b) in FIG. 6. Figure 21 The specific process of the crater partition processing will be described below with reference to the weld seam 3 shown in (a)-(b) in FIG. 6.

[0107] The edge profile is obtained by cutting the weld seam depth image shown in (a) in FIG. 6 along the extension direction of the weld seam 3, as shown in (b) in FIG. 6. Figure 21 The edge profile is obtained by cutting the weld seam depth image shown in (a) in FIG. 6 along the extension direction of the weld seam 3, as shown in (b) in FIG. 6. Figure 21 The edge profile is obtained by cutting the weld seam depth image shown in (a) in FIG. 6 along the extension direction of the weld seam 3, as shown in (b) in FIG. 6.

[0108] The first fold point Q1 and the second fold point Q2 are determined for each edge profile. Since the first fold point Q1 and the second fold point Q2 are obtained in the same way as the first fold point Q1 and the second fold point Q2 obtained by the pinhole reference forming unit, no further description is given here.

[0109] The points on the shell sidewall profile 2a and the top cover profile 1a are fitted respectively to form a first fitting line L1 and a second fitting line L2. The first fitting line L1 and the second fitting line L2 are divided by the vertical line of the intersection O along the vertical direction (the dashed line part in (b) in FIG. 6) relative to the point O, and the weld seam profile 3a is divided into a first part close to the shell sidewall profile 2a and a second part close to the top cover profile 1a. Figure 21

[0110] Further, the first fold point Q1 divides the first part of the weld seam profile 3a into a first region close to the intersection O and a second region close to the shell sidewall profile 2a, and the second fold point Q2 divides the second part of the weld seam profile 3a into a third region close to the intersection O and a fourth region close to the top cover profile 1a.

[0111] When the edge profile obtained has a suspected crater defect profile, and the suspected crater defect profile is located in the first region, the suspected crater defect profile is a first position defect; when the suspected crater defect profile is located in the second region, it is a second position defect; when the suspected crater defect profile is located in the third region, it is a third position defect; and when the suspected crater defect profile is located in the fourth region, it is a fourth position defect.

[0112] ​The second pit reference forming unit is configured to form a corresponding reference line according to a position type to which the suspected pit defect belongs, specifically:

[0113] Figure 22 Figures (a)-(c) in the first area show one of the cases where the suspected pit defect is located. As shown in (a)-(c) in Figure 22, when the suspected pit defect is a first position defect, a line between the first fold point Q1 and the second fold point Q2 is obtained, which is a first line; then, the point on the weld contour 3a with the maximum perpendicular distance from the first line is found along the weld contour 3a, which is a first high point N1; then, a line between the first fold point Q1 and the first high point N1 is obtained, which is a first fold point reference line l1.

[0114] Figure 23 Figures (a)-(c) in the second area show one of the cases where the suspected pit defect is located. As shown in (a)-(c) in Figure 22, when the suspected pit defect is a second position defect, the first fitting line L1 is taken as the second fold point reference line. Figure 23

[0115] Figures (a)-(c) in the third area show one of the cases where the suspected pit defect is located. As shown in (a)-(c) in Figure 22, when the suspected pit defect is a third position defect, a line between the first fold point Q1 and the second fold point Q2 is obtained, which is a second line. The point on the weld contour 3a with the maximum perpendicular distance from the second line is found along the weld contour 3a, which is a second high point N2, and then a line between the second fold point Q2 and the second high point N2 is obtained, which is a third fold point reference line l3. Figure 24 Figure 24 Figures (a)-(c) in the fourth area show one of the cases where the suspected pit defect is located. As shown in (a)-(c) in Figure 22, when the suspected pit defect is a fourth position defect, the second fitting line L2 is taken as the fourth fold point reference line.

[0116] Figure 25 Figure 25 Figures (a)-(c) in the fourth area show one of the cases where the suspected pit defect is located. As shown in (a)-(c) in Figure 22, when the suspected pit defect is a fourth position defect, the second fitting line L2 is taken as the fourth fold point reference line.

[0117] ​​Of course, in actual production process, when the suspected pit defect profile possibly crosses the first region and the third region (i.e. the suspected pit defect profile has its two side endpoints of the weld 3 width located in the first region and the third region respectively), the distances from the two side endpoints of the suspected pit defect profile to the intersection O orthogonal projection line are calculated. When the distance from the endpoint of the suspected pit defect profile located in the first region to the intersection O orthogonal projection line is greater than the distance from the endpoint of the suspected pit defect profile located in the third region to the intersection O orthogonal projection line, it is considered that the suspected pit defect profile belongs to the first position defect, and then the first fold point reference line l1 is taken to calculate the defect depth value. Conversely, it is considered that the suspected pit defect profile belongs to the third position defect, and then the second fold point reference line l2 is taken to calculate the defect depth value. If the distance from the endpoint of the suspected pit defect profile located in the first region to the intersection O vertical perpendicular line (orthogonal projection line) is equal to the distance from the endpoint of the suspected pit defect profile located in the third region to the intersection O vertical perpendicular line (orthogonal projection line), it is considered that the suspected pit defect profile belongs to the first position defect or the third position defect, and then the first fold point reference line l1 or the second fold point reference line l2 can be used to calculate the defect depth value.

[0118] The second pit feature calculation unit is used to calculate the vertical distance between each point on the suspected pit defect profile and the corresponding reference line and take the maximum value to obtain the pit defect depth value.

[0119] Figure 26 The specific flow of the above pit processing module for pit partition processing of the suspected pit defect to calculate the fold point depth value to determine whether the suspected pit defect is a true defect is shown. As shown in Figure 26 For the suspected pit defect, the suspected pit defect is divided into four position types according to the position of the suspected pit defect on the weld 3, which are the first position defect, the second position defect, the third position defect and the fourth position defect. For different suspected pit defects in different positions, different reference lines are used to calculate the depth value, so that the calculated second pit defect depth value is closer to the actual depth value.

[0120] The above first position defect, second position defect, third position defect and fourth position defect are distinguished and the depth value is calculated by the first fold point Q1 and the second fold point Q2, so the depth value calculated by the above first position defect, second position defect, third position defect and fourth position defect is also called the fold point depth value of the suspected pit defect.

[0121] Since the first pit reference forming unit and the first pit feature calculation unit jointly realize the calculation of the first pit defect depth value of the suspected pit defect by the flattening processing, which has been described in detail in the foregoing, it will not be repeated here.

[0122] The pit feature analysis unit is configured to compare the first pit defect depth value calculated by the first pit feature calculation unit with the second pit defect depth value calculated by the second pit feature calculation unit and take the larger one as the pit defect depth value.

[0123] In the pit processing module as shown in Figure 19 The pit depth value analysis unit is configured to compare the pit defect depth value output by the pit feature analysis unit with the pit depth preset value, and if the pit defect depth value is greater than the depth preset value, it is determined that the suspected pit defect is a true defect, otherwise, it is determined to be a non-defect.

[0124] However, the inventors of the present application found in production practice that there is an error between the calculated inflection point depth value of the first position defect and the third position defect and the actual depth value of the defect. Based on this, in order to further improve the accuracy of the obtained defect depth value and ensure that the calculated depth value can truly reflect the defect condition, the inventors of the present application form an end point reference line from the two end points of the suspected pit defect profile to calculate the depth value.

[0125] Specifically, the second pit reference forming unit is further configured to obtain a first end point and a second end point on both sides of the suspected pit defect profile along the width direction in the intercepted edge profile, and obtain a line connecting the first end point and the second end point as the end point reference line l 端 , wherein the edge profile is composed of the top cover profile 1a, the shell side wall profile 2a, the weld profile 3a and the suspected pit defect profile.

[0126] Correspondingly, the second pit feature calculation unit is further configured to calculate the perpendicular distance between each point on the suspected pit defect profile and the end point reference line l 端 , and take the maximum value to obtain the end point depth value of the suspected pit defect, and then compare the end point depth value of the suspected pit defect with the aforementioned inflection point depth value and take the larger one. The larger one obtained is the second pit defect depth value. That is, when the suspected defect is the first position defect or the third position defect, different reference lines are formed by the obtained inflection point of the edge profile and the end point of the suspected defect profile, and then the inflection point depth value and the end point depth value are calculated and compared. In this way, by using two different reference line forming methods, the accuracy of calculating the pit defect depth value can be improved, and the calculated depth value is closer to the actual depth value.

[0127] , the first end point and the second end point of the suspected pit defect profile are obtained as follows: as shown in (a)-(b) in Figure 27 , identify the suspected pit defect in the weld defect image and mark the defect area area 缺 When the edge profile is intercepted at the position of the suspected pit defect, the defect area area缺 The two side boundaries along the width direction of the weld contour 3a intersect with the first edge and the second edge of the suspected pit defect contour on the edge contour along the width direction of the weld contour 3a at a point respectively, forming a first intersection point and a second intersection point. The defect area area 缺 may be obtained by AI grayscale processing and recognition of the weld defect map. Since this belongs to the prior art, it will not be described in detail here.

[0128] However, since the defect area area 缺 identified and obtained by AI is usually smaller than the actual range of the suspected pit defect (the range of the orthographic projection area), it may make the positions of the first intersection point and the second intersection point obtained closer to the lowest point of the suspected defect contour, thereby causing the end point depth value calculated based on the line between the first end point M1 and the second end point M2 as the reference to be lower than the actual depth value. Based on this, preferably, after the edge contour is intercepted, the first intersection point and the second intersection point of the suspected pit defect contour of the defect area area 缺 are extended outward along the edge contour by 15 pixel points, and the points after the outward extension are taken as the first end point M1 and the second end point M2, and then the end point reference line l 端 is formed based on the line between the first end point M1 and the second end point M2 after the outward extension to calculate the end point depth value of the suspected pit defect, so that the end point depth value calculated is closer to the actual depth value.

[0129] Figure 28 The specific flow of the above pit processing module for calculating the inflection point depth value and the end point depth value of the suspected pit defect belonging to the first position defect and the third position defect respectively to determine whether the suspected pit defect is a true defect is shown. As Figure 27 shown, for the first position defect and the third position defect, the inflection point depth value and the end point depth value are obtained respectively, then the obtained inflection point depth value and the end point depth value are compared, and the larger one is taken and compared with the depth preset value, so that the pit defect depth value calculated can be further ensured to be closer to the actual depth value, thereby the suspected defect can be more accurately judged to be a true defect, so as to further reduce the occurrence of the missed killing situation.

[0130] Compared with the prior art, the screening method and screening device of the present invention are based on the shape characteristics of pinhole defects and pit defects. For suspected pinhole defects, the weld is divided into different areas along its width direction, and different reference planes are selected for suspected defects located in different areas, and the depth value of the suspected pinhole defect is calculated based on the corresponding reference plane; at the same time, for suspected pit defects, the area where the suspected pit defect is located is delineated, and the top cover, shell side wall and weld are flattened so that the top cover, shell side wall and weld are coplanar, and the area outside the flattened area where the suspected pit defect is located is taken as the flattened reference plane, and then the depth value of the suspected pit defect is calculated based on the flattened reference plane, thereby improving the accuracy of the depth value calculation, so that the calculated depth values ​​of the suspected pinhole defect and the suspected pit defect are closer to the actual depth value, so that the suspected defect can be more accurately judged whether it is a true defect, which can reduce the occurrence of missed or over-killed, and further judge whether the quality of the weld meets the requirements based on the number of true defects, thereby improving the accuracy of detection.

[0131] Based on the same inventive concept, the present invention also provides an electronic device, which can be a terminal device such as a server, desktop computing device, or mobile computing device (e.g., laptop computing device, handheld computing device, tablet computer, netbook). The device includes one or more processors and a memory, wherein the processor is configured to execute the screening method of an embodiment of the program-implemented method; and the memory is configured to store a computer program executable by the processor.

[0132] Based on the same inventive concept, the present invention also provides a computer-readable storage medium, corresponding to the embodiment of the aforementioned screening method, wherein the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the screening method described in any of the aforementioned embodiments.

[0133] The application can take the form of a computer program product implemented on one or more storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so on) having computer readable program code embodied thereon. Computer readable storage media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible to computing devices.

[0134] The above-described embodiments only express several implementation manners of the present application, which are described in a more specific and detailed manner, but cannot be understood as a limitation on the scope of the patent application. It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, and the present application also intends to include these modifications and improvements.

Claims

1. A method for screening weld defects, characterized in that: The following steps are involved: S1: Obtaining a weld depth map and a weld defect map showing the top cover, the shell sidewall, the weld, and suspected defects, wherein at least one suspected defect in the weld defect map is marked as a suspected pinhole defect, and at least one suspected defect is marked as a suspected pit defect; S2: Screening the suspected pinhole defect and the suspected pit defect respectively to determine whether the suspected pinhole defect and the suspected pit defect are true defects, wherein: Screening for the suspected pinhole defect includes the following steps: Performing pinhole partitioning processing on the weld in the weld defect map along the width direction of the weld, and then determining the defect type of the suspected pinhole defect based on the area where the suspected pinhole defect is located, wherein the defect type of the suspected pinhole defect is any one of a weld bead defect, an inner edge defect, and an outer edge defect; forming a corresponding reference plane according to the defect type of the suspected pinhole defect, wherein the reference plane is any one of a weld bead reference plane corresponding to the weld bead defect, an inner side reference plane corresponding to the inner edge defect, and an outer side reference plane corresponding to the outer edge defect; Calculate the vertical distance between each point on the suspected pinhole defect and the corresponding reference plane and take the maximum value to obtain the pinhole defect depth value; Comparing the pinhole defect depth value with a preset pinhole depth value, and determining whether the suspected pinhole defect is a true defect based on the comparison result; Screening the suspected pit defects includes the following steps: The position of the suspected pit defect in the weld defect map is defined as a target area, and the positions on the top cover, the shell side wall, and the weld other than the target area are defined as non-target areas. The top cover, the shell side wall, and the weld are flattened to make them coplanar, and the non-target area after flattening is used as a flattening reference plane. Calculating the vertical distance between each point on the suspected pit defect and the flattened reference plane and taking the maximum value to obtain the pit defect depth value; Comparing the pit defect depth value with a preset pit depth value, and determining whether the suspected pit defect is a true defect based on the comparison result; S3: Determine whether the quality of the weld meets the requirements based on the number of true defects.

2. The method for screening weld defects according to claim 1, wherein: The pinhole partitioning process comprises the following steps: Marking a ridge line for the weld in the weld depth map, and dividing the weld into an inner area and an outer area along the width direction of the weld with the ridge line as a boundary; For the suspected pinhole defect in the weld defect map, m edge contours are intercepted along the extension direction of the weld, where m ≥ 1, and the edge contours are composed of the top cover contour, the shell side wall contour, the weld contour, and the suspected pinhole defect contour; Dividing the weld profile into an inner region profile and an outer region profile based on the ridge line, obtaining a first inflection point on the inner region profile, and obtaining a second inflection point on the outer region profile, wherein the first inflection point is a boundary point between the weld profile and the shell sidewall profile, and the second inflection point is a boundary point between the weld profile and the top cover profile; Calculating the distances between the first inflection point and both side edges of the suspected pinhole defect contour located on the inner region contour along the width direction thereof to obtain a first inner distance value and a second inner distance value; and / or calculating the distance values ​​between the two side edges of the suspected pinhole defect contour located on the contour of the outer region along the width direction and the second inflection point to obtain a first outer distance value and a second outer distance value; The first and second inner distance values ​​and / or the first and second outer distance values ​​are respectively compared with the preset distance values, and the defect type of the suspected pinhole defect is determined according to the comparison results.

3. The method for screening weld defects according to claim 1, wherein: The maximum vertical distance between each point on the suspected pit defect and the flattened reference plane is taken as the first pit defect value; In step S2, screening the suspected pit defects further includes the following steps: Intercepting an edge contour of the position where the suspected pit defect is located in the obtained weld defect image along the weld extension direction, and performing pit partitioning processing on the edge contour along its width direction, wherein the edge contour is composed of a top cover contour, a shell sidewall contour, a weld contour, and a suspected pit defect contour, and then determining the position type of the suspected pit defect based on the area where the suspected pit defect contour is located, wherein the position type of the suspected pit defect is any one of a first position defect, a second position defect, a third position defect, and a fourth position defect; forming a corresponding reference line according to the position type of the suspected pit defect, wherein the reference line is any one of a first inflection point reference line corresponding to the first position defect, a second inflection point reference line corresponding to the second position defect, a third inflection point reference line corresponding to the third position defect, and a fourth inflection point reference line corresponding to the fourth position defect; Calculate the vertical distance between each point on the suspected pinhole defect and the corresponding reference line and take the maximum value to obtain the second pit defect depth value; The second pit defect depth value is compared with the first pit defect depth value and the larger value is taken to obtain the pit defect depth value.

4. A device for screening weld defects, characterized in that: include: a pinhole processing module, configured to obtain a weld depth map and a weld defect map showing the top cover, the shell sidewall, the weld, and suspected defects, and to screen at least one suspected defect marked as a suspected pinhole defect in the weld defect map to determine whether the suspected pinhole defect is a true defect; Wherein, the pinhole processing module includes: a pinhole partitioning unit, configured to perform pinhole partitioning processing on the weld in the weld defect map along the width direction of the weld, and then determine the defect type of the suspected pinhole defect based on the area where the suspected pinhole defect is located, wherein the defect type of the suspected pinhole defect is any one of a weld bead defect, an inner edge defect, and an outer edge defect; a pinhole reference forming unit, configured to form a corresponding reference plane according to the defect type to which the suspected pinhole defect belongs, wherein the reference plane is any one of a weld bead reference plane corresponding to the weld bead defect, an inner side reference plane corresponding to the inner side edge defect, and an outer side reference plane corresponding to the outer side edge defect; a pinhole feature calculation unit, configured to calculate the vertical distance between each point on the suspected pinhole defect and the corresponding reference plane and take the maximum value to obtain the pinhole defect depth value; a pinhole depth value analysis unit, configured to compare the pinhole defect depth value with a preset pinhole depth value, and determine whether the suspected pinhole defect is a true defect based on the comparison result; a pit processing module, configured to obtain a weld defect map showing a top cover, a shell sidewall, a weld, and suspected defects, and screen at least one suspected defect marked as a suspected pit defect in the weld defect map to determine whether the suspected pit defect is a true defect; Wherein, the pit processing module includes: a first pit reference forming unit, configured to define a location of the suspected pit defect in the weld defect map as a target area, define locations on the top cover, the shell sidewall, and the weld other than the target area as a non-target area, flatten the top cover, the shell sidewall, and the weld so that they are coplanar, and use the flattened non-target area as a flattening reference surface; a first pit feature calculation unit, configured to calculate a vertical distance between each point on the suspected pit defect and the flattened reference plane and take the maximum value to obtain a pit defect depth value; a pit depth value analysis unit, configured to compare the pit defect depth value with a preset pit depth value, and determine whether the suspected pit defect is a true defect based on the comparison result; The weld quality analysis module is used to determine whether the quality of the weld meets the requirements based on the number of true defects.

5. The device for screening weld defects according to claim 4, characterized in that: The pit processing module further includes: a pit partitioning unit, configured to intercept an edge contour of the position of the suspected pit defect in the acquired weld defect image along the weld extension direction, and perform pit partitioning processing on the edge contour along its width direction, wherein the edge contour is composed of a top cover contour, a shell sidewall contour, a weld contour, and a suspected pit defect contour, and then determine a position type of the suspected pit defect based on the area where the suspected pit defect contour is located, wherein the position type of the suspected pit defect is any one of a first position defect, a second position defect, a third position defect, and a fourth position defect; a second pit reference forming unit, configured to form a corresponding reference line according to the position type of the suspected pit defect, wherein the reference line is any one of a first inflection point reference line corresponding to the first position defect, a second inflection point reference line corresponding to the second position defect, a third inflection point reference line corresponding to the third position defect, and a fourth inflection point reference line corresponding to the fourth position defect; a second pit feature calculation unit, configured to calculate the vertical distance between each point on the suspected pit defect and the corresponding reference line and take the maximum value to obtain a pit defect depth value; The pit feature analysis unit is used to compare the pit defect depth value calculated by the second pit feature calculation unit with the pit defect depth value calculated by the first pit feature calculation unit and take the larger value to obtain the pit defect depth value for the pit depth value analysis unit to analyze whether the suspected pit defect is a true defect.

6. A weld defect detection system, characterized by: comprising an image acquisition device, a defect acquisition device and a screening device as claimed in claim 4 or 5; The image acquisition device forms a weld depth map and transmits the map to the defect acquisition device, wherein the weld depth map shows the top cover, the side wall of the shell, the weld at the junction of the top cover and the side wall of the shell, and suspected defects located on the weld; The defect acquisition device acquires the weld defect map showing the suspected defect from the weld depth map, and transmits the weld defect map and the weld depth map to the screening device.

Images