End plate welding seam quality detection and judgment method and system based on image recognition

Abstract

The invention relates to the technical field of welding seam detection, in particular to an end plate welding seam quality detection and judgment method and system based on image recognition, and the method comprises the steps: S1, carrying out the multi-angle image collection of an end plate with a welding seam, and obtaining a multi-angle image of the end plate; s2, preprocessing the obtained multi-angle image to obtain preprocessed image data; s3, identifying a welding seam in the preprocessed image data; s4, performing feature extraction on each identified welding seam, and constructing a welding seam feature map; and S5, determining whether the welding seam quality of the end plate is qualified or not based on the constructed welding seam characteristic spectrum. According to the end plate welding seam quality detection and judgment method based on image recognition, in practical application, the welding seam quality of the end plate is judged by constructing the welding seam characteristic spectrum and based on the constructed welding seam characteristic spectrum, and rapid and accurate detection of the welding seam quality of the end plate can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of weld detection, and in particular to an end plate weld quality detection and determination method and system based on image recognition. Background Art

[0002] A weld is a seam formed by melting the welding rod and the metal at the joint using the high temperature of the welding heat source. After the weld metal cools, the two welds are connected into a whole. According to the shape of the weld metal and the relative position of the welds, there are butt welds, fillet welds, plug welds and electric riveting. In industrial production, the quality of welds directly affects the safety and reliability of products. Traditional end plate weld inspection methods mainly rely on manual visual inspection and some simple measuring tools, which are not only inefficient but also difficult to ensure accuracy.

[0003] Therefore, there is an urgent need for an end plate weld quality detection and determination method and system based on image recognition to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to solve the problems of low efficiency and low accuracy in the manual inspection of the quality of the end plate weld in the prior art, and to provide a method and system for inspecting and determining the quality of the end plate weld based on image recognition.

[0005] In order to achieve the above-mentioned object, the first aspect of the present invention provides an end plate weld quality detection and determination method based on image recognition, and the end plate weld quality detection and determination method based on image recognition comprises the following steps:

[0006] S1, collecting images of the end plate with the weld at multiple angles to obtain multi-angle images of the end plate;

[0007] S2, preprocessing the obtained multi-angle images to obtain image data after preprocessing;

[0008] S3, identifying the weld in the pre-processed image data;

[0009] S4, extracting features from each identified weld and constructing a weld feature map;

[0010] S5. Confirm whether the weld quality of the end plate is qualified based on the constructed weld characteristic map.

[0011] Preferably, in step S2, the preprocessing of the obtained multi-angle images specifically includes: performing normalization processing and noise reduction processing on the obtained multi-angle images in sequence.

[0012] Preferably, in step S3, identifying the weld in the pre-processed image data specifically includes:

[0013] S3.1, identifying the shape contour of each weld in the pre-processed image data, and comparing the identified shape contour of each weld with the standard weld shape contour;

[0014] S3.2. If the similarity value between the shape contour of the identified weld and the standard weld shape contour is greater than the set similarity threshold, it is identified as a weld, otherwise it is not a weld.

[0015] Preferably, in step S4, extracting features from each identified weld and constructing a weld feature map specifically includes:

[0016] The shape features and color features of each identified weld are extracted, and based on the shape features and color features of each weld, a weld feature map is constructed using a plane layout method.

[0017] Preferably, the shape feature extraction function is:

[0018]

[0019] Among them, S t is the shape characteristic, L is the length of the weld, and W is the width of the weld.

[0020] Preferably, the color feature extraction function is:

[0021]

[0022] Where D is the color feature; Cr, Cg and Cb are the red, green and blue components of the weld color respectively; Cs,r, Cs,g and Cs,b are the red, green and blue components of the standard weld color respectively.

[0023] Preferably, the construction function of the weld characteristic map includes:

[0024] Q(x)=G(X)×H(X)×K

[0025] G(X)=S t +D

[0026]

[0027] Among them, Q(x) is the weld feature map; G(X) is the function for marking the weld features; H(X) is the function for judging the weld features; and K is the feature judgment factor.

[0028] Preferably, in step S5, confirming whether the welding quality of the end plate is qualified based on the constructed weld feature map specifically includes:

[0029] S5.1. Confirm whether the number of welds in the weld feature map is equal to the number of standard welds. If so, the number of welds is qualified, otherwise it is unqualified; confirm whether the ratio between the area occupied by all welds in the weld feature map on the image and the total area of ​​the plane image is within a preset ratio range. If so, it is qualified; if not, it is unqualified; confirm whether the color difference comparison value between the color of each weld in the weld feature map and the standard color of the weld is within a preset color difference range and whether the color feature D is within a preset color deviation range. If so, it is qualified; if not, it is unqualified;

[0030] S5.2. If the number of welds in the weld feature map is equal to the standard number of welds, the ratio of the area occupied by all welds on the image to the total area of ​​the plane image is within the preset ratio range, the color difference comparison value between the color of each weld and the standard color of the weld is within the preset color difference range, and the color feature D is within the preset color deviation range, then the weld quality of the end plate is confirmed to be qualified, otherwise it is unqualified.

[0031] Preferably, the calculation formula for the area occupied by a single weld on the image is:

[0032] S=WL

[0033] Among them, S is the area occupied by a single weld on the image, L is the length of the weld, and W is the width of the weld.

[0034] Preferably, the calculation formula for the color difference comparison value between a single weld color and the standard color of the weld is:

[0035]

[0036] Among them, Ω is the color difference comparison value between the weld color and the standard color of the weld, C is the weld color, and E is the standard color of the weld.

[0037] Preferably, the method further includes optimizing the weld in combination with the weld characteristic map when it is confirmed that the weld quality of the end plate is unqualified.

[0038] The second aspect of the present invention provides an end plate weld quality detection and determination system based on image recognition, which is applied to the end plate weld quality detection and determination method described above. The end plate weld quality detection and determination system comprises:

[0039] An image acquisition module is used to acquire images of the end plate with the weld from multiple angles to obtain images of the end plate from multiple angles;

[0040] An image preprocessing module is used to preprocess the obtained multi-angle images to obtain image data after preprocessing;

[0041] An image weld seam recognition module, used to recognize weld seams in pre-processed image data;

[0042] A weld feature map construction module is used to extract features from each identified weld and construct a weld feature map;

[0043] The end plate weld quality determination module is used to confirm whether the weld quality of the end plate is qualified based on the constructed weld feature map.

[0044] According to the above technical scheme, based on the end plate weld quality detection and judgment method and system based on image recognition, by identifying the welds in the image data after preprocessing, then extracting features for each identified weld, and constructing a weld feature map, and then further confirming whether the weld quality of the end plate is qualified based on the constructed weld feature map, and when the weld quality is unqualified, the weld is optimized in combination with the weld feature map. In actual application, the efficiency and accuracy of the end plate weld quality detection can be effectively improved, and the safety of the end plate can be ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flow chart of the end plate weld quality detection and judgment method based on image recognition. DETAILED DESCRIPTION

[0046] The specific implementation of the embodiment of the present invention is described in detail below. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0047] The first aspect of the present invention provides a method for detecting and judging the quality of end plate welds based on image recognition, such as Figure 1 As shown, the end plate weld quality detection and determination method based on image recognition includes the following steps:

[0048] S1, collecting images of the end plate with the weld at multiple angles to obtain multi-angle images of the end plate;

[0049] S2, preprocessing the obtained multi-angle images to obtain image data after preprocessing;

[0050] S3, identifying the weld in the pre-processed image data;

[0051] S4, extracting features from each identified weld and constructing a weld feature map;

[0052] S5. Confirm whether the weld quality of the end plate is qualified based on the constructed weld characteristic map.

[0053] According to the above technical scheme, based on the end plate weld quality detection and judgment method based on image recognition, by identifying the welds in the image data after preprocessing, then extracting features for each identified weld, and constructing a weld feature map, and then further confirming whether the weld quality of the end plate is qualified based on the constructed weld feature map, and when the weld quality is unqualified, the weld is optimized in combination with the weld feature map. In actual application, the efficiency and accuracy of the end plate weld quality detection can be effectively improved, and the safety of the end plate can be ensured.

[0054] In the end plate weld quality inspection and determination method based on image recognition described in the present invention, specifically, in step S1, a variety of high-definition imaging devices such as cameras or mobile phones can be used to obtain images of the end plate with welds at multiple different angles to ensure that the image can clearly display the shape, texture and surrounding environment of the weld and the overall structure of the end plate, thereby improving the accuracy of subsequent identification of the weld.

[0055] In the end plate weld quality inspection and determination method based on image recognition described in the present invention, preferably, in step S2, the preprocessing of the obtained multi-angle images specifically includes: performing normalization processing and noise reduction processing on the obtained multi-angle images in sequence to eliminate the influence of scale differences between different images, and obtain image data with normalized scale features, and further performing denoising processing on the normalized image data to remove noise interference in the image, which can improve the image quality and thus improve the accuracy of subsequent weld identification.

[0056] In the end plate weld quality detection and determination method based on image recognition according to the present invention, preferably, in step S3, the identifying of the weld in the pre-processed image data specifically includes:

[0057] S3.1, identifying the shape contour of each weld in the pre-processed image data, and comparing the identified shape contour of each weld with the standard weld shape contour;

[0058] S3.2. If the similarity value between the shape contour of the identified weld and the standard weld shape contour is greater than the set similarity threshold, it is identified as a weld, otherwise it is not a weld.

[0059] Specifically, the similarity threshold is set to be no less than 90% to ensure the recognition accuracy of the weld in the image.

[0060] In the end plate weld quality detection and determination method based on image recognition described in the present invention, preferably, in step S4, the feature extraction of each identified weld and the construction of a weld feature map specifically include:

[0061] The shape features and color features of each identified weld are extracted, and based on the shape features and color features of each weld, a weld feature map is constructed using a plane layout method.

[0062] In a specific implementation, the shape feature extraction function is:

[0063]

[0064] Among them, S t is the shape characteristic, L is the length of the weld, and W is the width of the weld.

[0065] The color feature extraction function is:

[0066]

[0067] Where D is the color feature; Cr, Cg and Cb are the red, green and blue components of the weld color respectively; Cs,r, Cs,g and Cs,b are the red, green and blue components of the standard weld color respectively.

[0068] The construction functions of weld feature maps include:

[0069] Q(x)=G(X)×H(X)×K

[0070] G(X)=S t +D

[0071]

[0072] Among them, Q(x) is the weld feature map; G(X) is the function for marking weld features; H(X) is the function for judging weld features; K is the feature judgment factor. Specifically, the logic of H(X) for judging weld features is that if the shape feature S t In [S t min,S t max] and the color feature D is also in the range of [Dmin, Dmax], it is judged as qualified, and the judgment result of H(X) is 1, otherwise the judgment result of H(X) is 0.

[0073] In the present invention, the shape features and color features of each weld are extracted by using a shape feature extraction function and a color feature extraction function, and then a weld feature map is constructed based on the extracted weld shape features and color features through a weld feature map construction function, which can further effectively improve the accuracy of subsequent judgment on the quality of the end plate weld.

[0074] In the end plate weld quality detection and determination method based on image recognition of the present invention, preferably, in step S5, confirming whether the welding quality of the end plate is qualified based on the constructed weld feature map specifically includes:

[0075] S5.1. Confirm whether the number of welds in the weld feature map is equal to the number of standard welds. If so, the number of welds is qualified, otherwise it is unqualified; confirm whether the ratio between the area occupied by all welds in the weld feature map on the image and the total area of ​​the plane image is within a preset ratio range. If so, it is qualified; if not, it is unqualified; confirm whether the color difference comparison value between the color of each weld in the weld feature map and the standard color of the weld is within a preset color difference range and whether the color feature D is within a preset color deviation range. If so, it is qualified; if not, it is unqualified;

[0076] S5.2. If the number of welds in the weld feature map is equal to the standard number of welds, the ratio of the area occupied by all welds on the image to the total area of ​​the plane image is within the preset ratio range, the color difference comparison value between the color of each weld and the standard color of the weld is within the preset color difference range, and the color feature D is within the preset color deviation range, then the weld quality of the end plate is confirmed to be qualified, otherwise it is unqualified.

[0077] In a specific implementation, the calculation formula for the area occupied by a single weld on the image is:

[0078] S=WL

[0079] Among them, S is the area occupied by a single weld on the image, L is the length of the weld, and W is the width of the weld.

[0080] The calculation formula for the color difference comparison value between a single weld color and the standard weld color is:

[0081]

[0082] Among them, Ω is the color difference comparison value between the weld color and the standard color of the weld, C is the weld color, and E is the standard color of the weld.

[0083] In the present invention, by evaluating the quantity, shape and color of the end plate welds based on the weld characteristic map, it is possible to accurately determine whether the weld quality of the end plate is qualified in practical applications.

[0084] In the end plate weld quality detection and determination method based on image recognition described in the present invention, preferably, it also includes optimizing the weld in combination with the weld feature map when it is confirmed that the weld quality of the end plate is unqualified. Specifically, when the weld quality of the end plate is unqualified, the weld is optimized in terms of the number, shape and color of the weld in combination with the weld feature map, so as to ensure the safety of the end plate in use.

[0085] The second aspect of the present invention provides an end plate weld quality detection and determination system based on image recognition, which is applied to the end plate weld quality detection and determination method described above. The end plate weld quality detection and determination system based on image recognition includes:

[0086] An image acquisition module is used to acquire images of the end plate with the weld from multiple angles to obtain images of the end plate from multiple angles;

[0087] An image preprocessing module is used to preprocess the obtained multi-angle images to obtain image data after preprocessing;

[0088] An image weld seam recognition module, used to recognize weld seams in pre-processed image data;

[0089] A weld feature map construction module is used to extract features from each identified weld and construct a weld feature map;

[0090] The end plate weld quality determination module is used to confirm whether the weld quality of the end plate is qualified based on the constructed weld feature map.

[0091] According to the above technical scheme, based on the end plate weld quality detection and judgment system based on image recognition, by identifying the welds in the preprocessed image data, then extracting features for each identified weld, and constructing a weld feature map, and then further confirming whether the weld quality of the end plate is qualified based on the constructed weld feature map, and when the weld quality is unqualified, the weld is optimized in combination with the weld feature map. In actual application, the efficiency and accuracy of the end plate weld quality detection can be effectively improved, and the safety of the end plate can be ensured.

[0092] The present invention will be described in detail below through examples, but the protection scope of the present invention is not limited thereto.

[0093] Example 1

[0094] Use Figure 1 The end plate weld quality detection and determination method based on image recognition is used for end plate weld detection. Specifically, the end plate weld quality detection and determination method based on image recognition includes the following steps:

[0095] S1, collecting images of the end plate with the weld at multiple angles to obtain multi-angle images of the end plate;

[0096] S2, preprocessing the obtained multi-angle images to obtain image data after preprocessing;

[0097] S3, identifying the weld in the pre-processed image data;

[0098] S4, extracting features from each identified weld and constructing a weld feature map;

[0099] S5. confirming whether the weld quality of the end plate is qualified based on the constructed weld characteristic map;

[0100] Specifically, in step S1, a camera is used to collect images of the end plate with the weld from multiple angles;

[0101] In step S2, the preprocessing of the obtained multi-angle images specifically includes: performing normalization processing and noise reduction processing on the obtained multi-angle images in sequence;

[0102] In step S3, the identifying of the weld in the pre-processed image data specifically includes:

[0103] S3.1, identifying the shape contour of each weld in the pre-processed image data, and comparing the identified shape contour of each weld with the standard weld shape contour;

[0104] S3.2. If the similarity between the shape contour of the identified weld and the standard weld shape contour is greater than the set similarity threshold, it is identified as a weld, otherwise it is not a weld; the similarity threshold is set to 90%;

[0105] In step S4, the feature extraction is performed on each identified weld and a weld feature map is constructed, which specifically includes:

[0106] The shape and color features of each identified weld are extracted, and based on the shape and color features of each weld, a weld feature map is constructed using a plane layout method;

[0107] In step S5, confirming whether the welding quality of the end plate is qualified based on the constructed weld feature map specifically includes:

[0108] S5.1. Confirm whether the number of welds in the weld feature map is equal to the number of standard welds. If so, the number of welds is qualified, otherwise it is unqualified; confirm whether the ratio between the area occupied by all welds in the weld feature map on the image and the total area of ​​the plane image is within a preset ratio range. If so, it is qualified; if not, it is unqualified; confirm whether the color difference comparison value between the color of each weld in the weld feature map and the standard color of the weld is within a preset color difference range and whether the color feature D is within a preset color deviation range. If so, it is qualified; if not, it is unqualified; wherein the preset ratio range is [0.3, 0.35], the preset color difference range is [0.9, 1.1], and the preset color deviation range is (0, 25);

[0109] S5.2. If the number of welds in the weld feature map is equal to the standard number of welds, the ratio of the area occupied by all welds on the image to the total area of ​​the plane image is within a preset ratio range, and the color difference comparison value between the color of each weld and the standard color of the weld is within a preset color difference range, and the color feature D is within a preset color deviation range, then the weld quality of the end plate is confirmed to be qualified, otherwise it is unqualified;

[0110] The method also includes optimizing the weld in combination with the weld characteristic map when it is confirmed that the weld quality of the end plate is unqualified.

[0111] After testing, it was found that the end plate weld quality inspection and judgment method based on image recognition described in the present invention can effectively improve the efficiency and accuracy of end plate weld quality inspection and judgment when applied in practice, and ensure the safety of end plate use, compared with the solution of manual inspection and judgment of end plate weld quality.

[0112] Example 2

[0113] Refer to Example 1, except that the shape feature extraction function is:

[0114]

[0115] Among them, S t is the shape feature, L is the length of the weld, and W is the width of the weld;

[0116] The color feature extraction function is:

[0117]

[0118] Where D is the color feature; Cr, Cg and Cb are the red, green and blue components of the weld color respectively; Cs,r, Cs,g and Cs,b are the red, green and blue components of the standard weld color respectively;

[0119] The construction functions of weld feature maps include:

[0120] Q(x)=G(X)×H(X)×K

[0121] G(X)=S t +D

[0122]

[0123] Wherein, Q(x) is the weld feature spectrum; G(X) is the function for marking weld features; H(X) is the function for determining weld features; K is the feature determination factor, specifically 0.65; Wherein, S t min is 50, S t max is 60; Dmin is 0, and Dmax is 25.

[0124] After testing, it was found that the end plate weld quality inspection and determination method based on image recognition described in the present invention can further effectively improve the accuracy of the end plate weld quality inspection and determination compared with the scheme in Example 1, thereby further effectively improving the safety of the end plate.

[0125] Example 3

[0126] Referring to Example 2, the difference is that the calculation formula for the area occupied by a single weld on the image is:

[0127] S=WL

[0128] Among them, S is the area occupied by a single weld on the image, L is the length of the weld, and W is the width of the weld;

[0129] The calculation formula for the color difference comparison value between a single weld color and the standard weld color is:

[0130]

[0131] Among them, Ω is the color difference comparison value between the weld color and the standard color of the weld, C is the weld color, and E is the standard color of the weld.

[0132] After testing, it is found that the end plate weld quality inspection and determination method based on image recognition described in the present invention can further effectively improve the accuracy of the end plate weld quality inspection and determination compared with the scheme in Example 2, thereby further effectively improving the safety of the end plate.

[0133] The end plate weld quality inspection and determination method and system based on image recognition provided by the present invention, by identifying the welds in the preprocessed image data, further extracting features for each identified weld, and constructing a weld feature map, further confirming whether the weld quality of the end plate is qualified based on the constructed weld feature map, can effectively improve the efficiency and accuracy of the end plate weld quality inspection in actual application.

[0134] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited thereto. Within the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of simple modifications. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. However, these simple modifications and combinations should also be regarded as the contents disclosed by the present invention and belong to the protection scope of the present invention.

Claims

1. A method for detecting and judging the quality of end plate welds based on image recognition, characterized in that: The end plate weld quality detection and determination method based on image recognition comprises the following steps: S1, collecting images of the end plate with the weld at multiple angles to obtain multi-angle images of the end plate; S2, preprocessing the obtained multi-angle images to obtain image data after preprocessing; S3, identifying the weld in the pre-processed image data; S4, extracting features from each identified weld and constructing a weld feature map; S5. Confirm whether the weld quality of the end plate is qualified based on the constructed weld characteristic map.

2. The method for detecting and judging the quality of end plate welds based on image recognition according to claim 1 is characterized in that: In step S2, the preprocessing of the obtained multi-angle image specifically includes: performing normalization processing and noise reduction processing on the obtained multi-angle image in sequence.

3. The end plate weld quality detection and determination method based on image recognition according to claim 1 is characterized in that: In step S3, the identifying of the weld in the pre-processed image data specifically includes: S3.1, identifying the shape contour of each weld in the pre-processed image data, and comparing the identified shape contour of each weld with the standard weld shape contour; S3.

2. If the similarity value between the shape contour of the identified weld and the standard weld shape contour is greater than the set similarity threshold, it is identified as a weld, otherwise it is not a weld.

4. The end plate weld quality detection and determination method based on image recognition according to any one of claims 1 to 3, characterized in that: In step S4, the feature extraction is performed on each identified weld and a weld feature map is constructed, which specifically includes: The shape features and color features of each identified weld are extracted, and based on the shape features and color features of each weld, a weld feature map is constructed using a plane layout method.

5. The method for detecting and judging the quality of end plate welds based on image recognition according to claim 4 is characterized in that: The shape feature extraction function is: Among them, S t is the shape characteristic, L is the length of the weld, and W is the width of the weld.

6. The method for detecting and judging the end plate weld quality based on image recognition according to claim 5 is characterized in that: The color feature extraction function is: Among them, D is the color feature; Cr, Cg and Cb are the red, green and blue components of the weld color respectively; Cs,r, Cs,g and Cs,b are the red, green and blue components of the standard weld color respectively.

7. The method for detecting and judging the quality of end plate welds based on image recognition according to claim 6 is characterized in that: The construction functions of weld feature maps include: Q(x)=G(X)×H(X)×K G(X)=S t +D Among them, Q(x) is the weld feature map; G(X) is the function for marking the weld features; H(X) is the function for judging the weld features; and K is the feature judgment factor.

8. The method for detecting and judging the quality of end plate welds based on image recognition according to claim 4 is characterized in that: In step S5, confirming whether the welding quality of the end plate is qualified based on the constructed weld feature map specifically includes: S5.

1. Confirm whether the number of welds in the weld feature map is equal to the number of standard welds. If so, the number of welds is qualified, otherwise it is unqualified; confirm whether the ratio between the area occupied by all welds in the weld feature map on the image and the total area of ​​the plane image is within a preset ratio range. If so, it is qualified; if not, it is unqualified; confirm whether the color difference comparison value between the color of each weld in the weld feature map and the standard color of the weld is within a preset color difference range and whether the color feature D is within a preset color deviation range. If so, it is qualified; if not, it is unqualified; S5.

2. If the number of welds in the weld feature map is equal to the standard number of welds, the ratio of the area occupied by all welds on the image to the total area of ​​the plane image is within the preset ratio range, the color difference comparison value between the color of each weld and the standard color of the weld is within the preset color difference range, and the color feature D is within the preset color deviation range, then the weld quality of the end plate is confirmed to be qualified, otherwise it is unqualified.

9. The method for detecting and judging the quality of end plate welds based on image recognition according to claim 8, characterized in that: The calculation formula for the area occupied by a single weld on the image is: S=WL Among them, S is the area occupied by a single weld on the image, L is the length of the weld, and W is the width of the weld.

10. The method for detecting and judging the quality of end plate welds based on image recognition according to claim 8, characterized in that: The calculation formula for the color difference comparison value between a single weld color and the standard weld color is: Among them, Ω is the color difference comparison value between the weld color and the standard color of the weld, C is the weld color, and E is the standard color of the weld.

11. The method for detecting and judging the quality of end plate weld based on image recognition according to claim 1, characterized in that: The method also includes optimizing the weld in combination with the weld characteristic map when it is confirmed that the weld quality of the end plate is unqualified.

12. An end plate weld quality detection and determination system based on image recognition, characterized in that: The method for detecting and judging the quality of the end plate weld seam according to any one of claims 1 to 11 above is applied, and the end plate weld seam quality detection and judgment system comprises: An image acquisition module is used to acquire images of the end plate with the weld from multiple angles to obtain images of the end plate from multiple angles; An image preprocessing module is used to preprocess the obtained multi-angle images to obtain image data after preprocessing; An image weld seam recognition module, used to recognize weld seams in pre-processed image data; A weld feature map construction module is used to extract features from each identified weld and construct a weld feature map; The end plate weld quality determination module is used to confirm whether the weld quality of the end plate is qualified based on the constructed weld feature map.

Images