Rotary arc welding detection method and system based on machine vision

Abstract

The invention is suitable for the technical field of welding detection, and provides a rotating electric arc welding detection method and system based on machine vision, and the system comprises a high-speed camera which is used for dynamically adjusting the view field range according to the size of a molten pool; the measuring wave band of the spectrograph is 370-1050 nm, the resolution ratio of the spectrograph is 0.75 nm, and the spectrograph is used for collecting spectral signals of welding arcs and a molten pool; the central wave band of the optical filter is 890 nm, and the optical filter is used for attenuating welding hard light and enhancing the characteristics of a molten pool; the image processing module is used for conducting median filtering, CLAHE contrast enhancement and ROI cutting preprocessing on the collected images, parameters such as the molten pool form and molten drop transition in the E307T0-1 flux-cored wire welding process can be monitored in real time, the problems of imaging blurring and noise interference in the high-temperature and strong-light environment are solved, and the welding quality detection precision and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of welding detection, and in particular to a rotary arc welding detection method and system based on machine vision. Background Art

[0002] During the rotary arc welding process, the morphological changes of the molten pool, the droplet transfer behavior and the size of the heat-affected zone are key factors affecting the welding quality.

[0003] However, the strong light radiation, high temperature splash and fast dynamic process in the welding environment cause traditional visual inspection methods to face the following problems: 1. The strong arc light can easily cause the camera sensor to overexpose and fail to capture the details of the molten pool; 2. The high-speed movement of the molten pool causes motion blur in ordinary shutter cameras; 3. Noise such as spatter and smoke interferes with the extraction of molten pool edge features.

[0004] Therefore, in view of the above situation, there is an urgent need to provide a rotary arc welding detection method and system based on machine vision to overcome the shortcomings in current practical applications. Summary of the Invention

[0005] The purpose of the present invention is to provide a rotary arc welding detection method and system based on machine vision, aiming to solve the problems in the above-mentioned background technology.

[0006] The present invention is implemented as follows: a rotary arc welding detection system based on machine vision comprises: High-speed camera, configured to achieve a frame rate of at least 2000 fps at a resolution of 1920 × 1080, a minimum exposure time of 1 μs, global shutter technology, a dynamic range of 60 dB, and a sensitivity of ISO 25000; Zoom lens with a focal length range of 17-50mm and an aperture value of F2.8-F22, used to dynamically adjust the field of view according to the size of the melt pool; Spectrometer, with a measurement band of 370-1050nm and a resolution of 0.75nm, used to collect spectral signals of the welding arc and molten pool; Filter, with a central wavelength of 890nm, is used to attenuate welding light and enhance weld pool characteristics; The image processing module is used to perform median filtering, CLAHE contrast enhancement and ROI cropping preprocessing on the collected images.

[0007] As a further solution of the present invention: the pixel size of the high-speed camera is 10×10 μm, and supports Gigabit Ethernet interface to transmit data.

[0008] As a further solution of the present invention: the field of view width and height of the zoom lens are calculated by the following formula:

[0009] When the working distance is 400mm, the field of view width at the wide-angle end is 451.8mm, and the field of view height is 254.1mm. The field of view width at the medium-telephoto end is 153.8mm, and the field of view height is 86.4mm.

[0010] As a further solution of the present invention: the spectrometer is equipped with a cosine corrector for eliminating the influence of the incident angle on the spectrum measurement.

[0011] As a further solution of the present invention: the median filtering formula of a two-dimensional image is as follows: For images Each pixel in , the filtered value Defined as: , in, is the original image matrix; The filtered image matrix; is the pixel coordinate currently being processed; is the relative offset within the sliding window; is the window radius; {·} is the median operation; the median filter of the image processing module uses a sliding window size of .

[0012] A method for detecting rotating arc welding based on machine vision is applied to the above-mentioned rotating arc welding detection system based on machine vision. The method comprises the following steps: Step 1: Real-time acquisition of melt pool images using a high-speed camera and zoom lens; Step 2: Use a spectrometer and filter to obtain the molten pool spectrum signal and filter out the interference of strong welding light; Step 3: Perform median filtering, CLAHE contrast enhancement and ROI algorithm image cropping preprocessing on the image in sequence; Step 4: Evaluate the melt pool morphology and temperature field distribution based on the preprocessed image.

[0013] As a further solution of the present invention: in step 1, the depth of field formula is used:

[0014]

[0015]

[0016] in As a further solution of the present invention: in step 3, the specific steps of the CLAHE contrast enhancement are: Histogram clipping: Clip the histogram of each local area to limit its maximum value:

[0017] in, Redistribute the cropped pixels: Evenly distribute the cropped pixels to all gray levels:

[0018] in Histogram equalization: Equalize the cropped histogram and calculate the cumulative distribution function:

[0019] The mapping function is:

[0020] in, As a further solution of the present invention: In step 3, the ROI algorithm formula is obtained by the upper left corner coordinate (

[0021] in, As a further solution of the present invention: in step 4, a natural image quality assessor is used to score the images before and after preprocessing, and the score of the preprocessed image must be higher than the score of the original image.

[0022] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a machine vision monitoring system and method based on rotary arc welding. Through the collaborative work of multiple modules, high-precision real-time monitoring of the molten pool morphology and temperature field in harsh welding environments is achieved. The system has the advantages of strong anti-interference ability, high imaging quality, and good adaptability, which can significantly improve the level of welding quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 This is a schematic diagram of the visual detection principle in the present invention.

[0025] Figure 2This is the quantum efficiency diagram of the present invention.

[0026] Figure 3 This is the spectrum area distribution diagram of the present invention.

[0027] Figure 4 This is a flowchart of image preprocessing in the present invention.

[0028] Figure 5 This is a comparison chart of the pretreatment effects in the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0030] The present invention will be further explained below with reference to specific embodiments.

[0031] See also Figure 1-Figure 5 The embodiment of the present invention provides a rotary arc welding detection system based on machine vision, comprising: High-speed camera, configured to achieve a frame rate of at least 2000 fps at a resolution of 1920 × 1080, a minimum exposure time of 1 μs, global shutter technology, a dynamic range of 60 dB, and a sensitivity of ISO 25000; Zoom lens with a focal length range of 17-50mm and an aperture value of F2.8-F22, used to dynamically adjust the field of view according to the size of the melt pool; Spectrometer, with a measurement band of 370-1050nm and a resolution of 0.75nm, used to collect spectral signals of the welding arc and molten pool; Filter, with a central wavelength of 890nm, is used to attenuate welding light and enhance weld pool characteristics; The image processing module is used to perform median filtering, CLAHE contrast enhancement and ROI cropping preprocessing on the collected images.

[0032] See also Figure 1-Figure 5 The embodiment of the present invention provides a method for detecting rotary arc welding based on machine vision, which is applied to the above-mentioned rotary arc welding detection system based on machine vision. The method includes the following steps: Step 1: Real-time acquisition of melt pool images using a high-speed camera and zoom lens; Step 2: Use a spectrometer and filter to obtain the molten pool spectrum signal and filter out the interference of strong welding light; Step 3: Perform median filtering, CLAHE contrast enhancement and ROI algorithm image cropping preprocessing on the image in sequence; Step 4: Evaluate the melt pool morphology and temperature field distribution based on the preprocessed image.

[0033] Specifically: 1. Visual inspection technology testing principle In optical imaging, depth of field refers to the spatial range in front of and behind the focal plane of the lens within which a clear image can be maintained. Its size is affected by four factors: aperture number, imaging distance, focal length, and sensor size. In actual monitoring of weld pools, the strong welding light and high temperature environment limit the variability of imaging distance, focal length, and sensor size. Therefore, the only way to control depth of field is by adjusting the aperture size. The formula for depth of field is:

[0034]

[0035]

[0036] in, In complex welding scenarios, the visual inspection principle is as follows Figure 1 shown.

[0037] 2. Monitoring system selection 2.1. Camera and lens selection The Mini series was developed by Qianyanlang for various applications in system integration, automotive collision analysis, non-contact measurement, and industrial research. Through a comprehensive redesign of its circuitry, structure, interfaces, exterior, and protective features, it features large memory, superior image quality, high performance, offline compatibility, and impact resistance. During arc additive manufacturing, the morphology and temperature field of the melt pool rapidly change with the arc and melting process. Dynamic behaviors include droplet transfer and thermal radiation variations. This camera achieves 2000 fps at 1920×1080 resolution, reaching up to 16,900 fps at lower resolutions, meeting the requirements for capturing these rapid dynamic changes. Due to the extremely high arc light intensity, the melt pool's inherent features may appear dim under strong light, requiring extreme exposure times to capture these details. The camera's minimum exposure time of 1μs enables precise capture of rapidly changing melt pool details, avoiding motion blur and overexposure. Melt pool morphology monitoring requires distortion-free imaging, while traditional rolling shutters can distort images of rapidly moving objects. This camera utilizes global shutter technology, ensuring accurate morphology in high-speed dynamic scenes. Under complex lighting conditions, such as strong arc light and ambient lighting, the edges of the weld pool and the heat-affected zone may appear dark. The IOS 25000's high sensitivity ensures signal acquisition in low-light conditions, and combined with its large 10μm pixel size, it further improves the signal-to-noise ratio. High-speed dynamic monitoring processes generate enormous amounts of image data. The Gigabit Ethernet interface effectively supports high frame rate data transmission, making it suitable for real-time acquisition and processing of monitoring data.

[0038] The parameters of the Qianyanlang M220 camera are shown in Table 1: Table 1 ThousandEyeWolf M220 camera parameters

[0039] Quantum efficiency is an important indicator that describes the response characteristics of camera sensors. It indicates the efficiency of the sensor in converting incident photons into electrons. Its calculation formula is:

[0040] Sigma lens parameters are shown in Table 2: Table 2 Sigma lens parameters

[0041]

[0042] The resolution used in this experiment is 1980×1080px, the frame rate is 2000fps, the sensor width resolution is 19.2mm, the longitudinal resolution is 10.8mm, and the weld size is (length: 7.5cm).

[0043] The field of view of the lens is calculated by the following formula:

[0044] Assuming a camera working distance D of 400mm, the object distance D must also take into account the high temperatures and molten liquid splashing generated during welding. At the wide-angle position (17mm focal length), the field of view is 451.8mm wide and 254.1mm high. At the mid-telephoto position (50mm focal length), the field of view is 153.8mm wide and 86.4mm high. The 17-50mm zoom range allows for flexible switching between panoramic observation and capturing local details. In arc additive manufacturing, the wide-angle position requires capturing the entire heat-affected zone (HAZ) of the melt pool and its surroundings. At the mid-telephoto position, the core area of ​​the melt pool can be magnified to analyze its morphology, droplet behavior, and defects. A zoom lens, by adjusting the focal length, can meet both requirements simultaneously, eliminating the need for frequent lens changes and improving monitoring efficiency. Arc additive manufacturing process parameters can alter the shape and size of the melt pool. Zoom lenses can quickly adjust the field of view to accommodate varying melt pool sizes.

[0045] 2.2. Spectral analysis and filter selection The fiber optic spectrometer used is a back-illuminated area array spectrometer model EK2000-Pro from Shanghai Chenchang Instrument Equipment Co., Ltd., with a measurement band of 370-1050nm, a slit width of 5μm, a resolution of 0.75nm, and an area array back-illuminated CCD detector. The straight-through fiber model is F-600-NIR straight-through pipeline, with an applicable band of 350-2500nm, a core diameter of 600μm, and a length of 2m. According to experimental detection, the spectrum consists of continuous radiation and very strong line spectra. The line spectrum peak in the spectral region is related to the element of the welding wire material, while the spectral intensity is related to the welding current intensity. The spectral region distribution measured in this experiment is as follows Figure 3 As shown, the spectrum is 210A welding current, 130mm / s welding speed, 14mm / s wire feeding speed, 2 average times and 50ms integration time; According to the spectral intensity distribution, the center band of the filter is selected as 890nm, and the molten pool photography under this band is tested.

[0046] 3. Welding experiment The welding wire used was 1.6 mm E307T0-1 flux-cored wire, which exhibits excellent crack resistance, impact resistance, and work hardening properties. This significantly increases surface hardness and enhances wear resistance under repeated loading. The elemental content of the E307T0-1 flux-cored wire is shown in Table 3. The substrate used for cladding was 45# steel. The test utilized fully automatic TIG rotary arc welding equipment. This equipment demonstrated stable operation, ease of operation, and high welding efficiency. The equipment primarily comprises a welding torch, welding power supply, automatic wire feed mechanism, water cooling system, PLC control system, and a three-axis welding platform. To acquire spectral signals from the welding arc and weld pool light, an EK2000 spectrometer from Shanghai Chenchang Instrument Co., Ltd. was used, equipped with an STD-CC cosine corrector.

[0047] Table 3 Chemical composition of E307T0-1 flux-cored welding wire (mass fraction%)

[0048] 4. Monitoring system construction and testing The filter center band is 890nm, the attenuator OD is 0.5, the gain is 2, the resolution is 1920×1080, the frame rate is 1000fps, the exposure time is 170μs, the welding speed is 120mm / s, and the wire feeding speed is 12mm / s.

[0049] 5. Image Preprocessing and Evaluation Welding is a process of drastic physical and chemical changes. The most common interferences to the acquired molten pool image are arc light, spatter, and smoke noise, which increases the difficulty of detecting the molten pool morphology. Therefore, it is necessary to preprocess the smoke and spatter noise in the molten pool image.

[0050] For image denoising, common methods are to use Gaussian blur or median filtering. Gaussian filtering is suitable for removing Gaussian noise, effectively smoothing and denoising, and has fast calculation speed, but it will produce fuzzy noise, blurring details and edges. Median filtering is suitable for removing salt and pepper noise, has strong adaptability, and can effectively protect image details and edges, but there is a contradiction between suppressing noise and maintaining image details. Median filtering works well for salt and pepper noise, while Gaussian blur is more effective for Gaussian noise. The welding process generates more random noise, so median filtering would be more suitable. Median filtering is a nonlinear filtering method. Its core idea is to replace the center pixel value with the median of the pixel values ​​in the sliding window to effectively remove noise. The median filter formula for a two-dimensional image is as follows: For images Each pixel in , the filtered value Defined as:

[0051] in,

[0052] In order to enhance the edge features of the melt pool, contrast enhancement, CLAHE, is used. The following are the steps of the algorithm: Histogram clipping: Clip the histogram of each local area to limit its maximum value:

[0053] in, Redistribute the cropped pixels: Evenly distribute the cropped pixels to all gray levels:

[0054] in Histogram equalization: Equalize the cropped histogram and calculate the cumulative distribution function (CDF):

[0055] The mapping function is:

[0056] in, The ROI algorithm image cropping is then performed. The following is the ROI algorithm formula: By the upper left corner coordinates ( , )( , ) and the lower right corner coordinates ( , )( , ) defines a rectangular area:

[0057] in, Image quality refers to the visual quality of an image and is evaluated both subjectively and objectively. Due to the intense thermal radiation, laser radiation, and plasma interference generated during the welding process, visual sensors cannot capture high-quality images of the weld pool, resulting in a lack of reference images. Therefore, the Natural Image Quality Evaluator (NIQE), a no-reference image quality assessment method, was selected. NIQE is based on a set of "quality-aware" features and fits them to an MVG (Multivariate Gaussian) model. These quality-aware features are derived from a simple but highly regularized NSS (Natural Scene Statistic) model. The NIQE metric for a given test image is expressed as the distance between the MVG model of the NSS features extracted from the test image and the MVG model of the quality-aware features extracted from a corpus of natural images. Image evaluation using the Natural Image Quality Evaluator (NIQE) yielded a score of 13.575 for the original image and 29.216 for the preprocessed weld pool image.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rotary arc welding detection system based on machine vision, characterized in that: include: High-speed camera, configured to achieve a frame rate of at least 2000 fps at a resolution of 1920 × 1080, a minimum exposure time of 1 μs, global shutter technology, a dynamic range of 60 dB, and a sensitivity of ISO 25000; Zoom lens with a focal length range of 17-50mm and an aperture value of F2.8-F22, used to dynamically adjust the field of view according to the size of the melt pool; Spectrometer, with a measurement band of 370-1050nm and a resolution of 0.75nm, used to collect spectral signals of the welding arc and molten pool; Filter, with a central wavelength of 890nm, is used to attenuate welding light and enhance weld pool characteristics; The image processing module is used to perform median filtering, CLAHE contrast enhancement and ROI cropping preprocessing on the collected images.

2. The machine vision-based rotary arc welding detection system according to claim 1, characterized in that: The pixel size of the high-speed camera is 10×10 μm, and supports Gigabit Ethernet interface to transmit data.

3. The machine vision-based rotary arc welding detection system according to claim 1, characterized in that: The width and height of the field of view of the zoom lens are calculated using the following formula: ； When the working distance is 400mm, the field of view width at the wide-angle end is 451.8mm, and the field of view height is 254.1mm. The field of view width at the medium-telephoto end is 153.8mm, and the field of view height is 86.4mm.

4. The machine vision-based rotary arc welding detection system according to claim 1, characterized in that: The spectrometer is equipped with a cosine corrector to eliminate the influence of the incident angle on spectrum measurement.

5. The machine vision-based rotary arc welding detection system according to claim 1, characterized in that: The median filter formula for a two-dimensional image is as follows: For images Each pixel in , the filtered value Defined as: ， in, is the original image matrix; The filtered image matrix; is the pixel coordinate currently being processed; is the relative offset within the sliding window; is the window radius; {·} is the median operation; the median filter of the image processing module uses a sliding window size of .

6. A method for detecting rotating arc welding based on machine vision, characterized in that: Applied to the machine vision-based rotary arc welding detection system according to any one of claims 1 to 5, the method comprises the following steps: Step 1: Real-time acquisition of melt pool images using a high-speed camera and zoom lens; Step 2: Use a spectrometer and filter to obtain the molten pool spectrum signal and filter out the interference of strong welding light; Step 3: Perform median filtering, CLAHE contrast enhancement and ROI algorithm image cropping preprocessing on the image in sequence; Step 4: Evaluate the melt pool morphology and temperature field distribution based on the preprocessed image.

7. The method for detecting rotary arc welding based on machine vision according to claim 6, characterized in that: In step 1, the depth of field formula is: ； ； ； in, is the depth of field length, is the depth of foreground, is the depth of field length, is the object distance, To allow the diameter of the circle of confusion, is the lens aperture value, is the focal length of the lens; as the aperture size increases, the depth of field becomes larger, and conversely, as the aperture size decreases, the depth of field becomes smaller, ensuring clear imaging of the entire molten pool area.

8. The method for detecting rotary arc welding based on machine vision according to claim 6, characterized in that: In step 3, the specific steps of the CLAHE contrast enhancement are: Histogram clipping: Clip the histogram of each local area to limit its maximum value: ； in, is the original histogram The frequency of gray levels, is the clipping threshold, = α , α is the crop factor, is the number of pixels in the local area; Redistribute the cropped pixels: Evenly distribute the cropped pixels to all gray levels: ； in is the total number of gray levels; Histogram equalization: Equalize the cropped histogram and calculate the cumulative distribution function: ； The mapping function is: ； in, is the input pixel value, yes The minimum non-zero value of , where ⌊⋅⌋ indicates rounding down.

9. The method for detecting rotary arc welding based on machine vision according to claim 6, wherein: In step 3, the ROI algorithm formula is calculated by the upper left corner coordinate ( , )( , ) and the lower right corner coordinates ( , )( , ) defines a rectangular area: ； in, is the original image matrix.

10. The method for detecting rotary arc welding based on machine vision according to claim 6, characterized in that: In step 4, a natural image quality assessor is used to score the images before and after preprocessing. The score of the preprocessed image must be higher than that of the original image.

Images