A deep penetration argon arc welding image optimization method and system based on pulse triggered imaging

Abstract

The present invention proposes a deep penetration argon arc welding image optimization method and system using pulse-triggered imaging. By combining pulse-triggered imaging technology with an HDR camera and applying an adaptive algorithm for parameter optimization, the clarity, dynamic range, and detail display capability of the welding image are improved, and the technical problems of low contrast, difficulty in detecting keyholes and molten pool edges, and difficulty in selecting exposure time sequences, which exist in traditional deep penetration argon arc welding image digitization methods, are solved. The present invention can dynamically adjust the parameters and image processing algorithms of the HDR camera according to the quality of the welding image and the welding progress to adapt to different welding conditions and requirements. Therefore, the invention has high practicality and application prospects, and can be used for the digitization and analysis of welding images during deep penetration argon arc welding, providing a reliable technical means for welding quality control and process optimization.

Description

Technical Field

[0001] The present invention relates to the field of welding image processing, in particular to deep penetration argon arc welding image processing, and specifically to a deep penetration argon arc welding image optimization method and system using pulse triggered imaging. Background Art

[0002] To improve welding productivity, automation of the deep penetration argon argon (TIG) welding (K-TIG) process has become a key development direction in the shipbuilding and large vessel manufacturing industries. However, in K-TIG welding, the high heat input of the plasma arc generates a large number of photons, resulting in high brightness in the field of view of the weld area, which poses a challenge for visual monitoring of the weld pool during the welding process. In the images captured by the camera, some areas may appear overexposed or underexposed. For example, when the weld pool is well exposed, the keyhole may be overexposed; conversely, when the keyhole is fully exposed, the weld pool may be underexposed. Therefore, capturing images of deep penetration TIG welding is a relatively difficult problem.

[0003] Furthermore, the weld area in deep penetration TIG welding contains information such as the arc center, the center of the gap to be welded, and the state of the weld pool, all within varying brightness ranges. This poses a challenge for simultaneous detection. Traditional imaging techniques cannot effectively handle this high brightness difference, so high dynamic range imaging is required to obtain images that accommodate this variation.

[0004] High dynamic range imaging technology is one feasible solution to this problem. It generates images with a high dynamic range by acquiring multiple exposures. There are two common approaches to achieve this goal. One approach relies on the camera's inherent photoelectric conversion principles to reproduce the scene brightness and simultaneously capture details of different brightness areas in a single image. The other approach utilizes an image fusion algorithm with multiple exposure times. This involves acquiring images of the same scene at different exposure times, each containing details of different brightness areas. These images are then combined into a single image to display the details of these different brightness areas.

[0005] In existing research on camera monitoring during welding, exposure time is typically determined based on experience or power cost considerations. However, this approach struggles to adapt to variations in the welding process, resulting in overexposure or underexposure of the image. Furthermore, due to a lack of adequate consideration of the scene irradiance, it is difficult to ensure the stability and consistency of the captured welding image quality. Therefore, a method is needed that can adaptively adjust the exposure time based on the irradiance of the welding scene to ensure stable and consistent image quality. Summary of the Invention

[0006] In response to the technical problems existing in the prior art, the present invention provides a pulse-triggered imaging deep-penetration argon arc welding image optimization method and system, which can effectively solve the problems of low contrast, keyhole and molten pool edge detection difficulties, exposure time sequence selection difficulties and image fusion processing consistency in the digitization process of deep-penetration argon arc welding images, and can obtain high-quality, high-dynamic range welding images, and provide reliable technical support for welding quality control and process optimization.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A deep penetration argon arc welding image optimization method based on pulse triggered imaging comprises the following steps:

[0009] S1. Based on the material to be welded, the shielding gas, and the welding parameters, the HDR image processor initializes the exposure time sequence of the HDR camera and sends a control signal to the pulse generator to generate a trigger pulse corresponding to the exposure time sequence to control the HDR camera to capture the welding image;

[0010] S2. Based on the captured welding image, obtain the mapping relationship between the image pixel value and the welding scene irradiance and exposure time, that is, the HDR camera response function f(I) = ln AE + ln Δt, where I is the image pixel value, E is the welding scene irradiance, Δt is the exposure time, and A is a constant related to the HDR camera parameters and its spatial position;

[0011] S3. Based on the response function f(I), converting the captured welding image into an irradiance distribution map of the welding scene;

[0012] S4. Scan the irradiance distribution map line by line, record the irradiance value of each pixel in each line, and derive the irradiance gradient. Use the gradient extreme value method to accurately locate the edge points of the keyhole and the molten pool during the welding process.

[0013] S5. Based on the welding images captured by the exposure time sequence, record the maximum irradiance of the keyhole area and the molten pool area, and calculate their average values, as well as the corresponding pixel points and pixel values, which are expressed as the keyhole average irradiance E kmax and pixel value I kmax , average irradiance of the molten pool E pmax and pixel value I pmax ;

[0014] S6. Based on the response function f(I) and the irradiance and pixel values ​​in step S5, the exposure time for optimal imaging of the keyhole and molten pool regions is calculated, a pulse trigger control signal is generated, and a pulse generator is used to control the HDR camera to capture the welding image, thereby obtaining images of optimal imaging of the keyhole and molten pool regions. These images are then fused to obtain a welding image that clearly images both the keyhole and molten pool regions.

[0015] S7. Repeat steps S3 to S6 until welding is completed to obtain and process multiple welding image sequences.

[0016] Furthermore, the exposure time sequence initialized in step S1 is generated based on the SVPWM technology, that is, a triangle wave and a sine wave are used for modulation so that the exposure time width varies according to a sine function, thereby obtaining images of different brightnesses in order to provide more image details.

[0017] Furthermore, the pulse triggering process of controlling the HDR camera shutter is to start the shutter at the rising edge of the pulse to start exposure, and close the shutter at the falling edge of the pulse to stop exposure.

[0018] Furthermore, the response function f(I) in step S2 is solved by the splicing method, that is, n welding images are captured using the exposure time sequence, k scene points are selected for each image, and n irradiances and n×k pixel values ​​are obtained, where n is the length of the exposure time sequence and k is a natural number. Then, each irradiance and its corresponding k pixel values ​​are used to describe a segment of the response function curve. The n segmented curves are spliced ​​into a complete curve through geometric transformation, and the response function f(I) is finally obtained through curve fitting.

[0019] Furthermore, the geometric transformation process includes:

[0020] SA1. First determine the reference irradiance E0, let f(I) = ln AE0 + lnΔt m =0, that is, AE0Δt m =1, E0=1 / (AΔt m ), where Δt m Take the exposure time as the median time width;

[0021] SA2. Convert n irradiances into relative irradiances, and keep the corresponding pixel values ​​unchanged. The conversion equation is E′ i =(E0 / E m )E i , where E m is the exposure time Δt m The corresponding irradiance, E′ i is the i-th relative irradiance, E i is the i-th irradiance, i=1,2,...,n;

[0022] SA3. Set the relative irradiance E′ i and their corresponding pixel values ​​are plotted as a curve.

[0023] Furthermore, when the exposure time series is sorted from small to large, when n is an odd number, Δt m =Δt (n+1) / 2 , when n is an even number, Δt m =Δt n / 2 .

[0024] Furthermore, the gradient extreme value method process is to first obtain the gradient sequence of irradiance changes, and then calculate the maximum and minimum values ​​of the gradient sequence, define the first extreme value as the left edge point of the molten pool, the second extreme value as the left edge point of the keyhole, the third extreme value as the right edge point of the keyhole, and the fourth extreme value as the right edge point of the molten pool, thereby determining the keyhole area and the molten pool area during the welding process.

[0025] Furthermore, the calculation to generate the pulse trigger control signal is based on the current exposure time and the historical exposure time sequence, and is calculated using the following formula:

[0026] Δt k =Σ(-1) j w j Δt k (xj)

[0027] Δt p =Σ(-1) j w j Δt p (xj)

[0028] w0=1,w j =(1-0.9 / j)w j-1

[0029] Where x is the current time, Δt k (xj) and Δt p (xj) are the exposure time of the keyhole and the molten pool at the xjth moment, w j is the weight of the exposure time at the xjth moment, j = 0, 1, 2, ..., 30.

[0030] A deep penetration argon arc welding image optimization system with pulse triggered imaging, comprising a deep penetration argon arc welding control system, an HDR camera, an HDR image processor and a pulse generator;

[0031] The deep penetration argon arc welding control system includes a deep penetration argon arc welding power supply, a welding gun, a water cooling device and a motion mechanism for carrying the welding gun;

[0032] The HDR camera is fixed on the motion mechanism and remains relatively stationary with the welding gun. It is connected to the pulse generator and the HDR image processor via a data line. According to the trigger signal transmitted by the pulse generator, the HDR camera sequentially captures the image information of the keyhole and the molten pool during the deep penetration argon arc welding process, and transmits the captured image information to the HDR image processor.

[0033] The HDR image processor receives the original image from the HDR camera, generates a high dynamic welding image after analysis-processing-fusion operations, generates a control signal for controlling the exposure time of the HDR camera according to the exposure characteristics of the generated image, and sends it to the pulse generator;

[0034] The pulse generator receives a control signal sent by the HDR image processor, generates a pulse trigger signal, and controls the action of the image capture shutter of the HDR camera.

[0035] Furthermore, it includes at least three HDR cameras, among which HDR-1 is installed in front of the movement direction of the welding gun, tilted horizontally at 30°, and is used to photograph the front of the molten pool, arc shape, electrode position and weld gap information; HDR-2 is installed in the rear of the movement direction of the welding gun, tilted horizontally at 45°, and is used to photograph the arc shape, keyhole information, the center of the molten pool and the tail of the molten pool; HDR-3 is installed on the side of the movement direction of the welding gun, tilted horizontally at 45°, and is used to photograph the center of the molten pool and the rear of the molten pool.

[0036] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0037] (1) Improving the quality and clarity of welding images: Through the application of pulse-triggered imaging and HDR cameras, the present invention can effectively solve the problems of overexposure and underexposure in welding images, provide clearer and more accurate welding images, and make the monitoring and quality control of the welding process more reliable.

[0038] (2) Improve the accuracy of edge detection of the welding area: Through line-by-line scanning and gradient peak detection methods, the present invention can accurately determine the edge of the welding area, achieve precise positioning of the welding area, and provide accurate data support for the analysis and evaluation of the welding process.

[0039] (3) Adaptive parameter optimization and processing: Through the application of adaptive algorithms, the present invention can automatically adjust the exposure time sequence and image processing parameters according to the irradiance of the welding scene to adapt to different welding conditions and requirements, and improve the stability and consistency of image quality.

[0040] (4) Improve the monitoring and evaluation capabilities of the welding process: Through consistent image fusion processing and adaptive welding image processing, the present invention can provide clear and accurate welding images, provide a reliable basis for real-time monitoring and quality evaluation of the welding process, and help operators make timely adjustments and optimizations.

[0041] In summary, the present invention solves multiple technical problems in the digitization process of deep-penetration argon arc welding images, improves the quality and clarity of welding images, enhances the monitoring and evaluation capabilities of the welding process, and provides important technical support for the automation and efficiency of deep-penetration argon arc welding. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the pulse-triggered imaging deep penetration argon arc welding image optimization method;

[0043] Figure 2 It is a schematic diagram of a pulse-triggered imaging deep-penetration argon arc welding image optimization system;

[0044] Figure 3 This is a schematic diagram of the exposure time series generated by the SVPWM technology;

[0045] Figure 4 is the segmented calibration result of the HDR camera response function f(I);

[0046] Figure 5 is the fitting result of the HDR camera response function f(I);

[0047] Figure 6 It is an irradiance distribution map of the position changes of pixels in different rows. DETAILED DESCRIPTION

[0048] The present invention will be described in further detail below.

[0049] like Figure 1 As shown in FIG, a pulse-triggered imaging deep penetration argon arc welding image optimization method mainly includes seven steps:

[0050] Step 1: Based on the material to be welded, shielding gas, and welding parameters, the HDR image processor initializes the exposure time sequence of the HDR camera and sends the generated control signal to the pulse generator. The pulse generator generates a trigger pulse corresponding to the exposure time sequence based on the control signal and controls the HDR camera shutter to capture the welding image.

[0051] Step 2: Based on the captured welding image, obtain the mapping relationship between the image pixel value and the welding scene irradiance and exposure time, that is, the response function of the HDR camera.

[0052] f(I)=lnAE+lnΔt

[0053] Where I is the image pixel value, E is the irradiance of the welding scene, Δt is the exposure time, and A is a constant related to the HDR camera parameters and its spatial position;

[0054] Step 3: Based on the response function f(I), the captured welding image is converted into an irradiance distribution map of the welding scene;

[0055] Step 4: Scan the irradiance distribution map line by line, record the irradiance value of each pixel in each row, and derive the irradiance change gradient. Use the gradient extreme value method to accurately locate the edge points of the keyhole and molten pool area during the welding process;

[0056] Step 5: Based on the welding images captured by the exposure time sequence, record the maximum irradiance of the keyhole area in the irradiance distribution diagram of each image and calculate the average value, which is recorded as E kmax ; Record the maximum irradiance of the molten pool area in the irradiance distribution map of each image and calculate its average value, which is recorded as E pmax ; Find the pixel point and pixel value corresponding to the maximum irradiance in the keyhole area, and record it as I kmax ; Find the pixel point and pixel value corresponding to the maximum irradiance in the molten pool area, and record it as I pmax ;

[0057] Step 6: Based on the response function f(I) and the irradiance and pixel value in step S5, the exposure time for optimal imaging of the keyhole and molten pool areas is calculated, a pulse trigger control signal is generated, and the HDR camera is controlled by the pulse generator to capture the welding image to obtain the images of the optimal imaging of the keyhole area and the molten pool area. These images are then fused to obtain a welding image in which both the keyhole area and the molten pool area are clearly imaged.

[0058] Step 7: Repeat steps 3 to 6 until welding is completed.

[0059] like Figure 2As shown, a deep-penetration argon arc welding image optimization system with pulse-triggered imaging includes a deep-penetration argon arc welding control system, an HDR camera, an HDR image processor, and a pulse generator. The deep-penetration argon arc welding control system includes a deep-penetration argon arc welding power supply, a welding gun, a water-cooling device, and a motion mechanism for carrying the welding gun. The welding gun is fixed to the motion mechanism by a clamp, and the motion mechanism controls the welding trajectory through the teaching setting. The HDR camera is fixed on the motion mechanism and remains relatively stationary with the welding gun. It is connected to the pulse generator and the HDR image processor through a data line. According to the trigger signal transmitted by the pulse generator, the keyhole and molten pool image information of the deep-penetration argon arc welding process is captured in sequence, and the captured image information is transmitted to the HDR image processor. The HDR image processor receives the original image of the HDR camera, generates a high-dynamic welding image after analysis-processing-fusion operations, and generates a control signal for controlling the exposure time of the HDR camera according to the exposure characteristics of the generated image, and sends it to the pulse generator. The pulse generator receives the control signal sent by the HDR image processor, generates a pulse trigger signal, and controls the action of the image capture shutter of the HDR camera. As shown Figure 2 As shown, at least three HDR cameras are arranged, among which HDR-1 is installed in front of the moving direction of the welding gun, tilted horizontally at 30°, and is used to photograph the front of the molten pool, arc shape, electrode position and weld gap information; HDR-2 is installed in the rear of the moving direction of the welding gun, tilted horizontally at 45°, and is used to photograph the arc shape, keyhole information, the center of the molten pool and the tail of the molten pool; HDR-3 is installed on the side of the moving direction of the welding gun, tilted horizontally at 45°, and is used to photograph the center of the molten pool and the rear of the molten pool.

[0060] The exposure time is related to the characteristics of the welding material, the type and flow of the shielding gas, the welding current and voltage, etc. In order to avoid underexposure or overexposure, it is necessary to determine the minimum and maximum values ​​of the exposure time. The minimum value should be able to capture important details in the welding process, while the maximum value should avoid overexposure and loss of image details. At the same time, the length of the exposure time sequence also needs to be determined. Make sure that the sequence length is long enough to capture the key stages and changes of the entire welding process. Through experimental verification and continuous optimization, the optimal exposure time sequence setting suitable for specific deep penetration argon arc welding applications can be obtained. However, before welding begins, the appropriate range of exposure time is not known. The present invention uses SVPWM technology to initialize the exposure time sequence, such as Figure 3As shown. Triangular and sine waves are used to modulate the pulse width. When the sine wave level is greater than the triangular wave level, the pulse generator outputs a high level. When the sine wave level is less than the triangular wave level, the pulse generator outputs a low level. After this modulation, the pulse width changes according to a sinusoidal function, first from small to large and then from large to small, and so on in a cyclical manner. The HDR camera activates the shutter at the rising edge of the pulse to begin exposure and closes the shutter at the falling edge of the pulse to stop exposure. In this way, the pulse width is equal to the image exposure time, thereby obtaining images that alternate between low and high brightness. This will ensure that clear, accurate, and sufficiently detailed welding images are obtained, providing a good foundation for subsequent image digitization processing.

[0061] The existing technology selects the length and interval of the exposure time sequence based on experience and fixed settings. This fixed setting is difficult to adapt to changes in different welding processes, resulting in overexposure or underexposure of the image. The present invention solves the response function f(I) in step S2 through a splicing method. That is, n welding images are captured using an exposure time sequence, k scene points are selected for each image, and n irradiances and n×k pixel values ​​are obtained, where n is the length of the exposure time sequence and k is a natural number. Each irradiance and its corresponding k pixel values ​​are then used to describe a segment of the response function curve. The n segmented curves are spliced ​​into a complete curve through geometric transformation, and the response function f(I) is finally obtained through curve fitting.

[0062] The present invention uses four different exposure times of Δt1, Δt2, Δt3, and Δt4 to capture images of the deep penetration argon arc welding process, and selects five different scene points from these images, where each scene point corresponds to a different pixel point. Figure 4 As shown, the obtained correspondence is analyzed in the same figure. It is found that the measurement values ​​obtained at different scene points can show a certain trend of change, and they are all equivalent to the change law of a certain section of a curve, which is equivalent to a certain section of the camera response function curve. The difference in irradiance between different scene points depends on the selection of the reference irradiance. Before performing curve fitting, a geometric transformation operation is performed first to obtain the reference irradiance. The irradiance of different scene points is adjusted according to this reference irradiance. Let

[0063] f(I)=lnAE0+lnΔt m =0

[0064] AE0Δt m =1, E0=1 / (E0Δt m )

[0065] Where Δt m Take the exposure time of the median time width. When the exposure time sequence is sorted from small to large, when n is an odd number, Δt m =Δt(n+1) / 2 , when n is an even number, Δt m =Δt n / 2 .

[0066] Next, the n irradiances are converted to relative irradiances, and the corresponding pixel values ​​remain unchanged. The conversion equation is E′ i =(E0 / E m )E i , where E m is the exposure time Δt m The corresponding irradiance, E′ i is the i-th relative irradiance, E i is the ith irradiance, i=1,2,...,n. Then the relative irradiance E′ i and its corresponding pixel values ​​are plotted as a curve. Figure 5 As shown in Figure 1, the discrete points determine the direction of the complete response function, so the curve fitting method can be used to fit the final response function f(I).

[0067] After the response function f(I) is obtained, the captured welding image can be converted into the irradiance distribution diagram of the welding scene. According to f(I) = ln AE + lnΔt,

[0068] E=e f(I) / (AΔt)

[0069] According to the above formula, the exposure time of the image and the pixel values ​​of different pixels are known and converted into corresponding irradiance in turn.

[0070] The present invention proposes a pulse-triggered imaging technology, which calculates the appropriate exposure time through the response function f(I). First, the irradiance distribution diagram of deep penetration argon arc welding is analyzed to distinguish the keyhole area and the molten pool area. Figure 6 As shown in the figure, the irradiance distribution diagram is scanned line by line. Each line corresponds to the irradiance change of the welding scene along the row coordinate direction. The irradiance value of each row of pixels is recorded and plotted as a curve with the change of row pixel position. As can be seen from the figure, there are two positions where the irradiance change curve suddenly increases and two positions where the value suddenly decreases. After the irradiance suddenly increases, there is a horizontal area, which indicates that the irradiance index is 2 and its amplitude is about 10 2 , then suddenly rises and has a horizontal area, which means the irradiance index is 4 and its amplitude is about 10 4, and then undergoes two sudden decreases. Obviously, the sudden change position points of the irradiance can be considered as the edge points of the keyhole and the molten pool. As the number of rows in the image gradually increases, the spacing between the sudden change points becomes smaller and smaller, and the inverted U-shaped keyhole shape and the molten pool shape have smaller and smaller lateral edge distances. Therefore, the edges of the keyhole and the weld can be determined by searching for points on the curve where the gradient changes significantly, thereby distinguishing the keyhole area and the weld area. The present invention uses the gradient extreme value method to accurately locate the edge points of the keyhole and the molten pool area during the welding process. The process is: take the irradiance corresponding to 5 continuously distributed pixel points and take the average value as the irradiance of the pixel point, then obtain the gradient sequence of the irradiance change with the pixel point position, and then calculate the maximum and minimum values ​​of the gradient sequence. According to the previous analysis, the first extreme value is defined as the left edge point of the molten pool, the second extreme value is the left edge point of the keyhole, the third extreme value is the right edge point of the keyhole, and the fourth extreme value is the right edge point of the molten pool, thereby determining the keyhole area and the molten pool area during the welding process.

[0071] After determining the keyhole area and the molten pool area, the maximum irradiance of the keyhole area and the molten pool area is recorded according to the welding images captured by the exposure time sequence, and their average values, as well as the corresponding pixel points and pixel values, are calculated, which are expressed as the keyhole average irradiance E and E respectively. kmax and pixel value I kmax , average irradiance of the molten pool E pmax and pixel value I pmax The exposure time for optimal imaging of the keyhole and molten pool area can be calculated based on the response function f(I)lnAE+lnΔt.

[0072] In order to avoid excessive fluctuations in exposure time, an adaptive algorithm is used to generate the pulse trigger control signal. The algorithm is based on the current exposure time and the historical exposure time series. The calculation formula is:

[0073] Δt k =∑(-1) j w j Δt k (xj)

[0074] Δt p =∑(-1) j w j Δt p (xj)

[0075] w0=1,w j =(1-0.9 / j)w j-1

[0076] Where x is the current time, Δt k (xj) and Δt p(xj) are the exposure time of the keyhole and the molten pool at the xjth moment, w j is the weight of the exposure time at time xj, where j = 0, 1, 2, ..., 30. The above formula shows that exposure time is constrained by the state over a period of time. The closer to the current moment, the greater the weight of the exposure time. This ensures the continuity of exposure time and the algorithm converges quickly.

[0077] After generating the pulse trigger control signal, the pulse generator is used to control the HDR camera to capture the welding image, and the optimal imaging images of the keyhole area and the molten pool area are obtained. These images are then fused to obtain a welding image in which the keyhole area and the molten pool area are clearly imaged at the same time.

[0078] In summary, the present invention achieves the digitization and optimization of deep penetration argon arc welding images by introducing pulse-triggered imaging and an HDR camera, combined with an adaptive algorithm to adjust the exposure time sequence. This method can provide clear and accurate welding images, solving the problems of underexposure and overexposure in traditional imaging technologies. The present invention solves multiple technical problems in the process of digitizing deep penetration argon arc welding images, improves the quality and clarity of welding images, enhances the monitoring and evaluation capabilities of the welding process, and provides important technical support for automation, efficiency, and quality control in the field of deep penetration argon arc welding.

[0079] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A deep penetration argon arc welding image optimization method based on pulse triggered imaging, characterized in that: The steps include: S1. Based on the material to be welded, the shielding gas, and the welding parameters, the HDR image processor initializes the exposure time sequence of the HDR camera and sends a control signal to the pulse generator to generate a trigger pulse corresponding to the exposure time sequence to control the HDR camera to capture the welding image; S2. Based on the captured welding image, obtain the mapping relationship between the image pixel value and the welding scene irradiance and exposure time, that is, the response function of the HDR camera f ( I )=ln AE +ln ∆ t ,in, I is the image pixel value, E is the irradiance of the welding scene, ∆ t is the exposure time, A are constants related to HDR camera parameters and their spatial positions; S3. Based on response function f ( I ), converting the captured welding image into an irradiance distribution map of the welding scene; S4. Scan the irradiance distribution map line by line, record the irradiance value of each pixel in each line, and derive the irradiance gradient. Use the gradient extreme value method to accurately locate the edge points of the keyhole and the molten pool during the welding process. S5. Based on the welding images captured by the exposure time sequence, record the maximum irradiance of the keyhole area and the molten pool area, and calculate their average values, as well as the corresponding pixel points and pixel values, which are expressed as the keyhole average irradiance E kmax and pixel values I kmax , average irradiance of the melt pool E pmax and pixel values I pmax ; S6. According to the response function f ( I ) and the irradiance and pixel value thereof in step S5, calculate the exposure time for optimal imaging of the keyhole and molten pool areas, generate a pulse trigger control signal, and control the HDR camera to capture the welding image through the pulse generator to obtain the images of the optimal imaging of the keyhole area and the molten pool area, and fuse these images to obtain a welding image in which the keyhole area and the molten pool area are clearly imaged at the same time; S7. Repeat steps S3 to S6 until welding is completed to obtain and process multiple welding image sequences.

2. The deep penetration argon arc welding image optimization method according to claim 1, characterized in that: The exposure time sequence initialized in step S1 is generated based on the SVPWM technology, that is, a triangle wave and a sine wave are used for modulation so that the exposure time width varies according to a sine function, thereby obtaining images of different brightnesses in order to provide more image details.

3. The deep penetration argon arc welding image optimization method according to claim 1, characterized in that: The pulse triggering process of controlling the HDR camera shutter is to open the shutter at the rising edge of the pulse to start exposure, and close the shutter at the falling edge of the pulse to stop exposure.

4. The deep penetration argon arc welding image optimization method according to claim 1, characterized in that: Solve the response function in step S2 by splicing method f ( I ), that is, using exposure time series to capture n welding images, each image is selected k scene points, get n irradiance and n × k pixel values, where n is the exposure time series length, k is a natural number, and then each irradiance and its corresponding k The pixel value describes a segment of the response function curve, and the n The segmented curves are spliced ​​into a complete curve, and the response function is finally obtained through curve fitting. f ( I ).

5. The deep penetration argon arc welding image optimization method according to claim 4, characterized in that: The geometric transformation process includes: SA1. Determine the reference irradiance first E 0, let f ( I )=ln AE 0+ln ∆ t m =0, that is AE 0∆ t m =1, E 0=1 / ( A ∆ t m ), where ∆ t m Take the exposure time as the median time width; SA2. n The irradiance is converted to relative irradiance, and the corresponding pixel value remains unchanged. The conversion equation is: E′ i =( E 0 / E m ) E i ,in, E m is the exposure time ∆ t m The corresponding irradiance, E′ i For the i The relative irradiance, E i For the i Irradiance, i =1,2,..., n ; SA3. Relative irradiance E′ i and their corresponding pixel values ​​are plotted as a curve.

6. The deep penetration argon arc welding image optimization method according to claim 5, characterized in that: When the exposure time series is sorted from small to large, n is an odd number, ∆ t m =∆ t (n+1) / 2 ,when n is an even number, ∆ t m =∆ t n / 2 .

7. The deep penetration argon arc welding image optimization method according to claim 1, characterized in that: The gradient extreme value method process is as follows: first, a gradient sequence of irradiance changes is obtained, and then the maximum and minimum values ​​of the gradient sequence are calculated. The first extreme value is defined as the left edge point of the molten pool, the second extreme value is the left edge point of the keyhole, the third extreme value is the right edge point of the keyhole, and the fourth extreme value is the right edge point of the molten pool, thereby determining the keyhole area and the molten pool area during the welding process.

8. The deep penetration argon arc welding image optimization method according to claim 1, characterized in that: The calculation of the pulse trigger control signal is based on the current exposure time and the historical exposure time sequence, and is calculated using the following formula: in, x is the current moment, ∆ t k ( x - j ) and ∆ t p ( x - j ) are respectively x - j Exposure time of keyhole and molten pool, w j For the x - j The weight of the moment exposure time, j=0,1,2,...,30.

9. A system for implementing the deep penetration argon arc welding image optimization method according to any one of claims 1 to 8, characterized in that: Includes deep penetration argon arc welding control system, HDR camera, HDR image processor and pulse generator; The deep penetration argon arc welding control system includes a deep penetration argon arc welding power supply, a welding gun, a water cooling device and a motion mechanism for carrying the welding gun; The HDR camera is fixed on the motion mechanism and remains relatively stationary with the welding gun. It is connected to the pulse generator and the HDR image processor via a data line. It is used to sequentially capture the image information of the keyhole and the molten pool during the deep penetration argon arc welding process according to the trigger signal transmitted by the pulse generator, and transmit the captured image information to the HDR image processor. The HDR image processor is used to receive the original image of the HDR camera and generate a high dynamic welding image, generate a control signal for controlling the exposure time of the HDR camera according to the exposure characteristics of the generated image, and send the control signal to the pulse generator; The pulse generator is used to receive a control signal sent by the HDR image processor, generate a pulse trigger signal, and control the action of the image capture shutter of the HDR camera.

10. The system according to claim 9, characterized in that There are at least three HDR cameras, the first of which is installed in front of the welding gun's movement direction, tilted horizontally at 30°, and is used to photograph the front of the molten pool, arc shape, electrode position and weld gap information; the second is installed behind the welding gun's movement direction, tilted horizontally at 45°, and is used to photograph the arc shape, keyhole information, molten pool center and molten pool tail; the third is installed on the side of the welding gun's movement direction, tilted horizontally at 45°, and is used to photograph the molten pool center and the rear of the molten pool.

Images