Welding defect rapid detection system based on machine vision

Abstract

The invention discloses a rapid welding defect detection system based on machine vision, which belongs to the technical field of welding defect detection and comprises a workpiece turn-over module, an image acquisition module, a data processing module, a defect detection and analysis module, a defect classification and evaluation module and a defect result visualization module. The image collecting module is used for collecting video and image information data of a welding workpiece in real time when a worker shakes the workpiece to turn over the workpiece, the workpiece turning-over module is used for placing, clamping and fixing the workpiece, and the workpiece is manually rocked and turned over at a constant speed to turn over the workpiece at the inner side position of the image collecting module, and the image collecting module is used for collecting video and image information data of the welding workpiece in real time when the worker shakes the workpiece to turn over. A data processing module, a defect detection and analysis module and a defect classification and evaluation module which are integrated together carry out image preprocessing and defect detection, identification and classification in real time, and when a workpiece is static, a defect result visualization module completely displays a defect detection result in a current visual field range so as to realize rapid detection of workpiece welding defects.

Description

Technical Field

[0001] This invention relates to the field of welding defect detection technology, and more specifically to a rapid welding defect detection system based on machine vision. Background Technology

[0002] In the field of industrial welding, especially in the processing of automotive parts, traditional welding robots frequently exhibit four types of defects during operation: burn-through, porosity, incomplete welding, and missed welding. These defects directly affect the welding quality of the workpiece and its subsequent performance, and have a serious adverse impact on product qualification rate and production efficiency.

[0003] Currently, the industry's standard procedure for handling welding defects involves manually transferring the workpieces to another inspection and repair welding line after the welding operation is completed. Defects are then detected by visual inspection, and workers manually repair the defects based on the inspection results. When dealing with a large number of workpieces, tight deadlines, and high efficiency requirements, manual visual inspection inevitably suffers from inspection bias and low efficiency, leading to missed or false detections of welding defects, which seriously affects the quality of automotive parts leaving the factory. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides a rapid welding defect detection system based on machine vision. By setting up an image acquisition module, video and image information data of the welded workpiece are acquired in real time when the worker shakes and flips the workpiece. An integrated data processing module, defect detection and analysis module, and defect classification and evaluation module perform image preprocessing and defect detection and classification in real time. When the workpiece is stationary, the defect result visualization module displays the complete defect detection results within the current field of view, thereby achieving rapid detection of welding defects in the workpiece and solving the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A rapid welding defect detection system based on machine vision includes a workpiece flipping module, an image acquisition module, a data processing module, a defect detection and analysis module, a defect classification and evaluation module, and a defect result visualization module. The workpiece flipping module is used for placing and clamping the workpiece, which is then manually rotated at a uniform speed to flip it inside the image acquisition module. The image acquisition module includes a lifting bracket, vision sensors, an acquisition card, and tripods. Six vision sensors are used; the lifting bracket positions four of them directly above the workpiece, and two tripods position the remaining two vision sensors to the sides of the workpiece. The vision sensors are connected to the acquisition card to acquire and transmit welding video and image information to the data processing module. The data processing module preprocesses the video and image information. The defect detection and analysis module detects defects after image processing and extracts quantitative features of the defects. The defect classification and evaluation module classifies defects based on their features and calculates the severity level of the defects. The defect result visualization module displays the defect detection results and locations within the current field of view in real time when the workpiece is stationary, instructing workers to perform repair welding.

[0006] Furthermore, the workpiece flipping module is located inside the lifting bracket, and the lifting bracket has two longitudinally distributed crossbars inside. The top of the lifting bracket has two symmetrically distributed grooves, and the two ends of the crossbars are slidably connected inside the two grooves respectively.

[0007] Furthermore, two connecting shells are screwed to the outer sides of each of the two crossbars. A steering connecting rod is rotatably connected to the bottom of each connecting shell, and the other end of the steering connecting rod is fixedly connected to the corresponding vision sensor.

[0008] Furthermore, the two triangular supports are located on the left and right sides of the workpiece flipping module, respectively, and the top of the triangular supports is rotatably connected to the corresponding vision sensor.

[0009] Furthermore, each of the six vision sensors is equipped with a supplementary light, which is snapped onto the front end of the corresponding vision sensor.

[0010] Furthermore, the image information preprocessing of the data processing module includes grayscale processing, Gaussian filtering, image sharpening, and image enhancement to improve image quality and amplify image features.

[0011] Furthermore, the defect detection and analysis module performs defect detection, including porosity, cracks, incomplete penetration and slag inclusion, after image processing according to traditional image processing methods. Based on geometric features, morphological features, texture features and positional features relative to the weld, it extracts measurable indicators of the defects.

[0012] Furthermore, the defect classification and evaluation module classifies the current welding defects into three levels based on the identification and detection of the current defects: porosity, burn-through, cracks, incomplete welding, and missing welding. The defects are then classified into three levels: minor, moderate, and severe according to their severity.

[0013] Furthermore, a power module is provided on the side of the workpiece flipping module, and the power module is connected to the vision sensor, the supplementary light, the data processing module and the defect result visualization module.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention utilizes four visual sensors positioned above and two on the sides of the workpiece. These sensors are pre-positioned manually to correspond to different areas and angles of the workpiece. As the workpiece is rotated at a constant speed within the image acquisition module, the visual sensors, in conjunction with a data acquisition card, acquire and transmit real-time video and image information of the welded workpiece to a data processing module. The data processing module preprocesses the video and image information to improve image quality and amplify the features of welding defect areas. A defect detection and analysis module performs defect detection after image processing and extracts quantitative features of the defects. A defect classification and evaluation module classifies defects based on these features and calculates the severity level of the defects. A defect result visualization module displays the detection results and locations of defects within the current field of view when the workpiece is stationary, effectively instructing workers to perform repair welding. This invention can effectively detect and identify target areas of welding defects on all surfaces of the workpiece during a single rotation, effectively replacing manual visual inspection for real-time welding defect detection without altering existing work procedures or equipment, thus achieving rapid detection of welding defects. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments, the accompanying drawings will be briefly described below.

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram of the system operation process of the present invention; Figure 3 This is a schematic diagram of the detection system of the present invention; Figure 4 This is a schematic diagram showing the structural distribution of the workpiece flipping module, image acquisition module, data processing module, and defect result visualization module in this invention. Figure 5 This is a schematic diagram of the image acquisition module in this invention; In the diagram: 1. Workpiece flipping module; 2. Image acquisition module; 21. Lifting bracket; 211. Crossbar; 212. Groove; 22. Vision sensor; 23. Acquisition card; 24. Triangular support; 25. Connecting shell; 251. Steering connecting rod; 26. Fill light; 3. Data processing module; 4. Defect detection and analysis module; 5. Defect classification and evaluation module; 6. Defect result visualization module; 7. Power supply module. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Specific mechanical structures of the present invention will be described in conjunction with the following references. Figures 1 to 5 The detailed description of the structure will be clearly presented. All structural details mentioned in the following embodiments are based on the accompanying drawings.

[0018] Please see Figures 1-5 In this embodiment of the invention, a rapid welding defect detection system based on machine vision includes a workpiece flipping module 1, an image acquisition module 2, a data processing module 3, a defect detection and analysis module 4, a defect classification and evaluation module 5, and a defect result visualization module 6. The workpiece flipping module 1 is used for placing and clamping the workpiece, and the workpiece is manually rotated at a uniform speed to flip it to the inner position of the image acquisition module 2. The image acquisition module 2 includes a lifting bracket 21, vision sensors 22, an acquisition card 23, and a triangular support 24. The number of vision sensors 22 is six. The lifting bracket 21 positions four vision sensors 22 directly above the workpiece, and the three... There are two corner brackets 24. The two corner brackets 24 position two additional vision sensors 22 on the side of the workpiece. The vision sensors 22 are connected to the acquisition card 23 to acquire and transmit welding video and image information to the data processing module 3. The data processing module 3 preprocesses the video image information. The defect detection and analysis module 4 is used for defect detection after image processing and extracts the quantitative features of the defects. The defect classification and evaluation module 5 classifies the defects according to the features and calculates the severity level of the defects. The defect result visualization module 6 displays the defect detection results and locations in the current field of view in real time when the workpiece is stationary, so as to instruct the worker to perform repair welding work.

[0019] The workpiece flipping module 1 is used to place the welding workpiece. The workpiece flipping module 1 can be a welding flipping table. The workpiece is clamped and fixed by the fixture on the welding flipping table. The welding flipping table is manually shaken to flip the surface of the workpiece, so as to realize the flipping inspection of each side of the welding workpiece.

[0020] Image acquisition module 2 is located outside workpiece flipping module 1. Image acquisition module 2 consists of lifting bracket 21, vision sensor 22, acquisition card 23, and triangular support 24. Lifting bracket 21 stands on the ground in the work area, and workpiece flipping module 1 is located inside lifting bracket 21. Manually rotating the workpiece at a uniform speed allows the entire workpiece to flip inside image acquisition module 2, so that each side of the workpiece faces the vision sensor 22 located inside lifting bracket 21. Two longitudinally distributed crossbars 211 are provided inside lifting bracket 21. Two symmetrically distributed grooves 212 are formed at the top of lifting bracket 21. The two ends of crossbars 211 are slidably connected inside the two grooves 212, allowing the two crossbars 211 to move sequentially along the grooves 212 inside the rectangular frame at the top of lifting bracket 21, facilitating adjustment of the crossbars 211's position at the top of lifting bracket 21. There are six vision sensors 22. Two connecting shells 25 are screwed to the outer sides of the two crossbars 211. A steering connecting rod 251 is rotatably connected to the bottom of the connecting shell 25. The other end of the steering connecting rod 251 is fixedly connected to the corresponding vision sensor 22. The four vision sensors 22 are fixed to the top position of the lifting bracket 21 through the connecting shells 25. The steering connecting rod 251 can rotate with the corresponding vision sensor 22 on the outside of the connecting shell 25 to adjust the lens orientation of the vision sensor 22. The connecting shell 25 can be adjusted in position on the outside of the crossbars 211 and re-fixed by screwing. When the lifting bracket 21 is activated, the rectangular frame at the top of the lifting bracket 21, along with the crossbars 211 and the four vision sensors 22, can move vertically above the workpiece flipping module 1 to adjust the position of the four vision sensors 22 directly above the workpiece. Two triangular supports 24 are used, located on the left and right sides of the workpiece flipping module 1, respectively. The top of each triangular support 24 is rotatably connected to a corresponding vision sensor 22, allowing the other two vision sensors 22 to be positioned on the side of the workpiece via the two triangular supports 24. The triangular supports 24 can move to the side of the workpiece flipping module 1, and the vision sensors 22 connected to the top of the triangular supports 24 can rotate at the top of the triangular supports 24 to adjust the position and status of the two vision sensors 22 on the side of the workpiece flipping module 1. The vision sensors 22 are fixedly installed according to the worker's visual inspection angle for welding defects and their positions are automatically adjusted. The vision sensors 22 are connected to a data acquisition card 23 to acquire the workpiece video stream data collected by the vision sensors 22. Each of the six vision sensors 22 is equipped with a supplementary light 26, which is snapped onto the front end of the corresponding vision sensor 22. The supplementary light 26 provides a uniform, non-reflective light source to ensure clear welding images.

[0021] Data processing module 3 connects to acquisition card 23. Data processing module 3 uses a high-performance computing host. Acquisition card 23 acquires and transmits welding video and image information to data processing module 3. Data processing module 3 performs image preprocessing and defect detection on the video information acquired by acquisition card 23. Image preprocessing in data processing module 3 includes grayscale processing, Gaussian filtering, image sharpening, and image enhancement to improve image quality and amplify image features. Acquisition card 23 and data processing module 3 are placed in a corner away from personnel passageways to ensure their safe use.

[0022] The processed parameters are input into the defect detection and analysis module 4, where the YOLO model performs target recognition and defect detection. Based on traditional image processing methods, defects including porosity, cracks, incomplete penetration, and slag inclusions are detected after image processing. Porosity detection is based on roundness analysis to identify and detect porosity data in the weld area. Crack detection is based on weld edge detection and morphological acquisition and recognition to detect crack data in the weld area. Incomplete penetration detection is based on the grayscale distribution of the weld area, analyzing the grayscale characteristics of the molten pool area on the back of the weld. By comparing the calculated grayscale statistics with a preset threshold, the weld penetration state is determined. A stable molten pool morphology and grayscale values ​​typically within a normal and relatively stable range indicate normal penetration; abnormal molten pool morphology, resulting in grayscale values ​​significantly lower than the threshold for normal penetration, indicates incomplete penetration; abnormally high grayscale values ​​indicate burn-through. Slag inclusion detection is based on the shape analysis of the weld area to identify and detect slag inclusion data. Point-like slag inclusions appear as individual black dots with angular and irregular shapes; strip-like slag inclusions appear as wide, short, thick lines; and elongated strip-shaped slag inclusions have wider and inconsistent lines. After defect detection and identification, measurable indicators of the defects are extracted based on geometric features, morphological features, texture features, and positional features relative to the weld.

[0023] The defect classification and assessment module 5, based on the measurable indicators of the extracted defects and the identification and detection of current defects, classifies the current welding defects into categories including porosity, burn-through, cracks, incomplete welds, and missed welds, and calculates the severity level of the defects, classifying them into three levels: minor, moderate, and severe. The defect result visualization module 6 can be a display screen, positioned directly in front of the workpiece flipping module 1 and the workpiece, ensuring that the welding technician can clearly and quickly view the workpiece defects displayed on the screen. The workpiece is manually flipped; when the workpiece is stationary, the defect result visualization module 6 displays the defect detection results and locations within the current field of view in real time, instructing the worker to perform the welding repair work quickly and accurately.

[0024] A power module 7 is located on the side of the workpiece flipping module 1. The power module 7 connects to the vision sensor 22, the supplementary light 26, the data processing module 3, and the defect result visualization module 6. The power module 7 provides a stable power supply to these components. A power button is located on the power module 7, allowing for one-button control of the entire system's power supply and shutdown. The worker presses the power button to start the entire inspection system.

[0025] The workpiece is placed and clamped on the workpiece flipping module 1 inside the image acquisition module 2. The workpiece is manually rotated at a constant speed to flip it inside the image acquisition module 2. Four vision sensors 22, positioned above the workpiece and two on its sides, are pre-adjusted by the operator to correspond to different areas and angles of the workpiece. The vision sensors 22, in conjunction with the acquisition card 23, acquire and transmit real-time video and image information of the welded workpiece to the data processing module 3. The data processing module 3 preprocesses the video image information to improve image quality and amplify the features of the welding defect area. The defect detection and analysis module 4 performs defect detection after image processing and extracts the quantitative features of the defects. The defect classification and evaluation module 5 classifies defects based on their features and calculates the severity level of the defects. The defect result visualization module 6 displays the defect detection results and locations within the current field of view in real time when the workpiece is stationary, effectively instructing the worker to perform repair welding. During the process of manually rotating the workpiece 360°, the target area detection and identification of all welding defects on all surfaces of the workpiece can be effectively completed. This effectively replaces the real-time detection of welding defects by human eyes, without changing the original tooling equipment and manual welding process, and without changing the workers' operating habits when rotating the workpiece for welding. It is low-cost, economical and practical, and easy to operate, achieving accurate and efficient detection of welding defects in the workpiece.

[0026] The working principle of this invention is as follows: First, the worker presses the power button to start the entire detection system. During the process of manually or using equipment to rotate and flip the workpiece at a constant speed, the vision sensor 22 collects the welding image information of the workpiece in real time. The acquisition card 23 acquires the video and image information data of the welded workpiece and transmits it to the data processing module 3. The data processing module 3, the defect detection and analysis module 4, and the defect classification and evaluation module 5 perform image preprocessing and defect detection identification and classification. When the workpiece is stationary, the defect result visualization module 6 displays the defect detection results and locations within the current field of view in real time to instruct the worker to perform repair welding.

[0027] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A machine vision based rapid detection system for weld defects, characterized in that, It includes a workpiece flipping module (1), an image acquisition module (2), a data processing module (3), a defect detection and analysis module (4), a defect classification and evaluation module (5), and a defect result visualization module (6). The workpiece flipping module (1) is used for placing and clamping the workpiece, and the workpiece is flipped inside the image acquisition module (2) by manual rotation at a uniform speed. The image acquisition module (2) includes a lifting bracket (21), a vision sensor (22), an acquisition card (23), and a triangular support (24). There are six vision sensors (22). The lifting bracket (21) positions four vision sensors (22) directly above the workpiece. There are two triangular supports (24). The two triangular supports (24) position the other two vision sensors (22) on the side of the workpiece. The vision sensors (22) are connected to the acquisition card (23) to acquire and transmit welding video and image information to the data processing module (3). The data processing module (3) preprocesses the video image information, and the defect detection and analysis module (4) is used for defect detection after image processing and extracts the quantitative features of the defects. The defect classification and evaluation module (5) classifies defects according to their characteristics and calculates the severity level of the defects. The defect result visualization module (6) displays the defect detection results and locations within the current field of view in real time when the workpiece is stationary, so as to instruct the worker to perform repair welding work.

2. The machine vision based welding defect rapid detection system according to claim 1, wherein, The workpiece flipping module (1) is located inside the lifting bracket (21). The lifting bracket (21) has two longitudinally distributed crossbars (211) inside. The top of the lifting bracket (21) has two symmetrically distributed grooves (212). The two ends of the crossbars (211) are slidably connected inside the two grooves (212).

3. The machine vision-based welding defect rapid detection system of claim 2, wherein, Two connecting shells (25) are screwed to the outside of the two crossbars (211). A steering connecting rod (251) is rotatably connected to the bottom of the connecting shell (25). The other end of the steering connecting rod (251) is fixedly connected to the corresponding vision sensor (22).

4. The machine vision based welding defect rapid detection system of claim 1, wherein, The two triangular supports (24) are located on the left and right sides of the workpiece flipping module (1), respectively, and the top of the triangular supports (24) is rotatably connected to the corresponding vision sensor (22).

5. The machine vision based welding defect rapid detection system of claim 1, wherein, Each of the six vision sensors (22) is equipped with a fill light (26), which is snapped onto the front end of the corresponding vision sensor (22).

6. The machine vision based welding defect rapid detection system of claim 1, wherein, The image information preprocessing of the data processing module (3) includes grayscale processing, Gaussian filtering, image sharpening, and image enhancement to improve image quality and amplify image features.

7. The machine vision based welding defect rapid detection system of claim 1, wherein, The defect detection and analysis module (4) performs defect detection, including porosity, cracks, incomplete penetration and slag inclusion, after image processing according to traditional image processing methods. Based on geometric features, morphological features, texture features and positional features relative to the weld position, it extracts measurable indicators of defects.

8. The rapid welding defect detection system based on machine vision according to claim 1, characterized in that, The defect classification and evaluation module (5) classifies the current welding defects into three levels: minor, moderate and severe, based on the identification and detection of the current defects.

9. A rapid welding defect detection system based on machine vision according to claim 5, characterized in that, The workpiece flipping module (1) is equipped with a power module (7) on its side. The power module (7) is connected to the vision sensor (22), the supplementary light (26), the data processing module (3), and the defect result visualization module (6).

Images