Vision Optimization Method, Medium and System for Welding Robot of Floating Plate of Oil Storage Tank

Abstract

The present invention provides a vision optimization method, medium and system for a petroleum storage tank pontoon plate welding robot, which belongs to the field of electrical digital data processing technology, including: through multi-dimensional and multi-scale data acquisition and decomposition strategies, comprehensively acquiring and deeply analyzing key information of the welding process, including weld images, robot posture, current, voltage, oil mist concentration and floating plate vibration frequency and other complex data. By constructing an image interference compensation model and a posture compensation model based on multivariate decomposition, the precise enhancement of the weld image quality and the dynamic correction of the robot posture are achieved, and the interference factors and system errors in the welding process are effectively eliminated. Finally, the optimized high-quality image data and precise posture data are used as the input of the machine vision module, which significantly improves the visual guidance accuracy of the petroleum storage tank pontoon plate welding robot, and solves the technical problem that the visual system of the welding robot is often affected by many interference factors, which seriously reduces the detection reliability.

Description

Technical Field

[0001] The invention belongs to the technical field of electric digital data processing, and in particular relates to a vision optimization method, medium and system for a welding robot for a petroleum storage tank pontoon plate. Background Art

[0002] In recent years, with the continuous advancement of oil exploration and production technology, the construction and application of large oil storage tanks have become more and more extensive. Among them, the pontoon plate is a key component of the floating roof of the tank, and its welding quality directly affects the safety and reliability of the entire tank system. Traditional manual welding has been difficult to meet the growing demand for oil storage tank construction due to its complex operation and low efficiency. Therefore, how to realize the automatic welding of the pontoon plate of the oil storage tank has become a key technical problem that needs to be solved urgently.

[0003] At present, industrial robots have been widely used in various welding operations due to their high efficiency and stability. In the field of oil tank pontoon plate welding, there are also automated welding solutions based on industrial robots. Such solutions usually use visual servo control technology, use machine vision systems to track and detect welds in real time, and feed back the detection results to the robot control system to achieve precise control of the welding trajectory.

[0004] However, in the complex oil tank welding environment, the vision system of the welding robot is often affected by many interference factors, such as oil mist generated during the welding process, welding arc interference, vibration of the floating board, etc. These factors will seriously reduce the detection reliability of the vision system, resulting in the robot being unable to accurately perform the welding task, thus affecting the welding quality. In addition, the posture fluctuation of the robot itself is also an important factor causing unstable welding quality. Summary of the invention

[0005] In view of this, the present invention provides a vision optimization method, medium and system for a petroleum storage tank pontoon plate welding robot, which can solve the technical problem that the vision system of the welding robot is often affected by many interference factors, seriously reducing the detection reliability of the vision system.

[0006] The present invention is achieved in that:

[0007] A first aspect of the present invention provides a vision optimization method for a petroleum tank pontoon plate welding robot, comprising the following steps:

[0008] S10, collecting weld image data, welding robot posture data, welding current data, welding voltage data, oil mist concentration data, and floating plate vibration frequency data during the welding process of the oil storage tank floating plate;

[0009] S20, performing multi-scale decomposition calculation on the weld image data to obtain a weld image steady-state feature matrix and a weld image dynamic feature matrix;

[0010] S30, performing multivariate decomposition calculation on the posture data of the welding robot to obtain a posture steady-state component and a posture fluctuation component;

[0011] S40, performing multivariate analysis and calculation on the oil mist concentration data to establish an oil mist density characteristic matrix;

[0012] S50, performing multi-dimensional decomposition on the welding current data, welding voltage data, and floating plate vibration frequency data to obtain a welding process parameter steady-state matrix and a welding process parameter fluctuation matrix;

[0013] S60, establishing an image interference compensation model based on the oil mist density characteristic matrix and the welding process parameter fluctuation matrix, and calculating an image enhancement coefficient matrix;

[0014] S70, optimizing the weld image data using the image enhancement coefficient matrix to obtain optimized weld image data;

[0015] S80, calculating the welding robot posture compensation amount based on the posture fluctuation component and the welding process parameter steady-state matrix to obtain corrected welding robot posture data: wherein the corrected welding robot posture data = original welding robot posture data + welding robot posture compensation amount;

[0016] S90, outputting the optimized weld image data and the corrected welding robot posture data for use as input of a machine vision module of a welding robot control system.

[0017] The multi-scale decomposition calculation process of the weld image in S20 is specifically expressed as follows:

[0018] ;

[0019] In the formula, is the original weld image; For the Layer approximation component; For the details; is the number of decomposition layers, the value range is 3-5; is the image noise error term; is the image coordinate.

[0020] Based on the above decomposition, the steady-state feature matrix of the weld image and the dynamic feature matrix The calculation is expressed as:

[0021] ;

[0022] ;

[0023] in, and Respectively represent the image Line Steady-state eigenvalues ​​and dynamic eigenvalues ​​of the column;

[0024] ;

[0025] ;

[0026] In the formula, is the average value of detail components.

[0027] The specific calculation process of the pose multivariate decomposition in S30 is as follows:

[0028] ;

[0029] In the formula, is the robot posture time series data; is the steady-state component of posture; is the posture fluctuation component; is the pose measurement error term; is the time variable.

[0030] Position fluctuation component The calculation is expressed as:

[0031] ;

[0032] In the formula, is the amplitude coefficient, ranging from 0 to 1; is the angular frequency; is the phase; is the attenuation coefficient; is the time constant; is the acceleration coefficient.

[0033] For the oil mist density characteristic matrix in S40 The establishment process is specifically expressed as follows:

[0034] ;

[0035] In the formula, is the oil mist concentration sampling point data; is the oil mist diffusion coefficient; are the number of rows and columns of sampling points respectively; is the oil mist measurement error term.

[0036] The specific decomposition process of welding process parameters in S50 is as follows:

[0037] ;

[0038] ;

[0039] In the formula, is the steady-state matrix of process parameters; is the process parameter fluctuation matrix; are the steady-state components of current, voltage, and vibration frequency respectively; They are the fluctuation components of current, voltage and vibration frequency respectively.

[0040] For the image enhancement coefficient matrix in S60 The calculation process is specifically expressed as:

[0041] ;

[0042] In the formula, is the enhancement coefficient, and its value range is 0.5-2; is the compensation factor.

[0043] The image optimization process in S70 is specifically expressed as follows:

[0044] ;

[0045] In the formula, For the optimized image; is the error term of the optimization process.

[0046] For the posture compensation in S80 The calculation process is specifically expressed as:

[0047] ;

[0048] In the formula, For the compensation cycle; To compensate for the error term.

[0049] The principle of the above calculation process is explained as follows:

[0050] 1. Multi-scale decomposition is used to separate different frequency components in the image. The high-frequency part reflects noise and interference, and the low-frequency part reflects the basic characteristics of the weld;

[0051] 2. The posture fluctuation components are in the form of three superpositions, which respectively represent the periodic vibration, attenuated vibration and acceleration effects;

[0052] 3. The diffusion coefficient is introduced into the oil mist density characteristic matrix, taking into account the diffusion characteristics of oil mist in space;

[0053] 4. The image enhancement coefficient matrix adopts the partial derivative form, which reflects the influence of the dynamic changes of oil mist concentration and process parameters on image quality;

[0054] 5. The posture compensation is in integral form to achieve compensation for the cumulative effect of posture fluctuations.

[0055] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned oil tank pontoon plate welding robot visual optimization method.

[0056] A third aspect of the present invention provides a petroleum storage tank pontoon plate welding robot visual optimization system, which includes the above-mentioned computer-readable storage medium.

[0057] Compared with the prior art, the beneficial effects of the oil tank pontoon plate welding robot vision optimization method, medium and system provided by the present invention are:

[0058] 1. The multi-sensor fusion method is used to comprehensively reflect the key parameters of the welding process, laying a solid foundation for subsequent modeling and optimization.

[0059] 2. The analysis and modeling method based on multi-dimensional features can accurately characterize the dynamic influence of oil mist, process parameters, posture, etc. on welding quality, providing a reliable basis for the design of compensation measures.

[0060] 3. The dynamic compensation image optimization algorithm and posture correction algorithm effectively reduce visual interference and robot motion errors, and significantly improve the clarity of weld images and robot motion accuracy.

[0061] 4. The closed-loop control solution integrating vision and posture has greatly improved the stability and repeatability of the welding quality of the floating plate of the oil storage tank.

[0062] In general, the visual optimization method proposed in the present invention makes full use of multi-source information fusion and dynamic compensation technology, effectively solves the key technical problems in the existing automatic welding of oil storage tank floating plates, significantly improves the level of welding quality control, and solves the technical problem that the visual system of the welding robot is often affected by many interference factors, which seriously reduces the detection reliability of the visual system. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0065] like Figure 1As shown, it is a flow chart of the visual optimization method of the oil tank pontoon plate welding robot provided by the present invention, and the method comprises the following steps:

[0066] S10, collecting weld image data, welding robot posture data, welding current data, welding voltage data, oil mist concentration data, and floating plate vibration frequency data during the welding process of the oil storage tank floating plate;

[0067] S20, performing multi-scale decomposition calculation on the weld image data to obtain a steady-state feature matrix and a dynamic feature matrix of the weld image;

[0068] S30, performing multivariate decomposition calculation on the welding robot posture data to obtain a posture steady-state component and a posture fluctuation component;

[0069] S40, performing multivariate analysis and calculation on the oil mist concentration data to establish an oil mist density characteristic matrix;

[0070] S50, performing multi-dimensional decomposition on the welding current data, the welding voltage data, and the buoy plate vibration frequency data to obtain a welding process parameter steady-state matrix and a welding process parameter fluctuation matrix;

[0071] S60, based on the oil mist density characteristic matrix and the welding process parameter fluctuation matrix, an image interference compensation model is established, and an image enhancement coefficient matrix is ​​calculated;

[0072] S70, optimizing the weld image data using an image enhancement coefficient matrix to obtain optimized weld image data;

[0073] S80, based on the posture fluctuation component and the steady-state matrix of the welding process parameters, calculate the posture compensation amount of the welding robot to obtain the corrected posture data of the welding robot: wherein the corrected posture data of the welding robot = the original posture data of the welding robot + the posture compensation amount of the welding robot;

[0074] S90, outputting the optimized weld image data and the corrected welding robot posture data, which are used as inputs of the machine vision module of the welding robot control system.

[0075] The specific implementation methods of the above steps are described in detail below:

[0076] The specific implementation of step S10 is as follows:

[0077] Firstly, a high-definition camera is used to collect weld image data in real time to obtain the appearance characteristics and morphological changes of the weld. For the weld image I(x,y), (x,y) represents the image coordinates. It reflects the real-time status of the weld, including weld shape, surface texture and other information.

[0078] At the same time, the robot body sensor is used to collect the posture data P(t) of the welding robot, where t represents the time variable. P(t) includes multi-dimensional information such as the position coordinates (x, y, z) and posture angles (α, β, γ) of the robot end effector, which can fully describe the motion trajectory of the robot during the welding process.

[0079] In addition, the output current I(t) and voltage U(t) data of the welding machine power supply are also collected to reflect the changes in welding process parameters. Current I(t) and voltage U(t) are key indicators of welding quality, and their dynamic changes will have an important impact on the weld morphology.

[0080] In addition, vibration sensors are used to obtain the vibration frequency F(t) data of the floating plate to monitor the dynamic response of the structure during welding. The vibration characteristics F(t) of the floating plate is also an important factor affecting the welding quality and needs to be considered.

[0081] Finally, a smoke detector is used to measure the oil mist concentration C(x,y,t) generated during welding, which provides a basis for subsequent image optimization. C(x,y,t) reflects the distribution of oil mist at different positions and times, which is one of the main reasons for the decline in weld image quality.

[0082] In summary, step S10 obtains key parameter data such as weld image I(x, y), robot posture P(t), welding current I(t) and voltage U(t), floating plate vibration frequency F(t) and oil mist concentration C(x, y, t) through the method of multi-sensor collaborative acquisition. These data lay the foundation for subsequent modeling and optimization.

[0083] The specific implementation of step S20 is as follows:

[0084] In order to extract steady-state features and dynamic features from the weld image I(x,y), wavelet transform is used to perform multi-scale decomposition on the original image. The expression of wavelet decomposition is as follows:

[0085] ;

[0086] in, represents the approximate component of the i-th layer, Represents the detail component, n is the number of decomposition layers, and the value range is 3-5. is the image noise error term.

[0087] Approximate Components reflects the low-frequency steady-state characteristics of the weld image, while the detail component It contains high-frequency dynamic features. Based on this, the steady-state feature matrix of the weld image can be constructed. and the dynamic feature matrix :

[0088] ;

[0089] ;

[0090] in, and They represent the steady-state eigenvalue and dynamic eigenvalue of the i-th row and j-th column of the image respectively.

[0091] This multi-scale decomposition method based on wavelet transform can effectively distinguish the noise components and effective information in the weld image, laying the foundation for subsequent image optimization.

[0092] The specific implementation of step S30 is as follows:

[0093] In order to analyze the steady-state and dynamic characteristics of the welding robot posture P(t), the time series decomposition algorithm is used to analyze it. The time series decomposition model is as follows:

[0094] ;

[0095] in, represents the steady-state component of posture, represents the pose fluctuation component, is the pose measurement error term.

[0096] For the pose fluctuation component , can be further divided into the following three parts:

[0097] ;

[0098] in, is the amplitude coefficient, is the angular frequency, is the phase, is the attenuation coefficient, is the time constant, is the acceleration coefficient. These three parts respectively characterize the effects of periodic vibration, attenuated vibration and acceleration change on posture.

[0099] Through this time series decomposition method, the steady-state characteristics of the welding robot posture can be accurately characterized. and dynamic characteristics , providing a basis for subsequent posture compensation.

[0100] The specific implementation of step S40 is as follows:

[0101] In order to establish the oil mist density characteristic matrix ,First, oil mist concentration sampling points were arranged at different locations in the welding area, and the Represents the oil mist concentration data at each sampling point. Then the oil mist diffusion coefficient is introduced , established through the multivariate regression method matrix:

[0102] ;

[0103] in, and represent the number of rows and columns of sampling points respectively, is the oil mist measurement error term.

[0104] The matrix can describe the spatial distribution characteristics of oil mist and reflect the oil mist concentration at different positions during welding, which provides an important basis for subsequent image optimization.

[0105] The specific implementation of step S50 is as follows:

[0106] In order to analyze the steady-state and dynamic characteristics of welding process parameters, the welding current ,Voltage and the vibration frequency of the pontoon The three parameters are decomposed into multiple dimensions:

[0107] ; ;

[0108] in, represents the steady-state matrix of process parameters, Represents the fluctuation matrix of process parameters. are the steady-state components of current, voltage, and vibration frequency, respectively. They are the fluctuation components of current, voltage and vibration frequency respectively.

[0109] This decomposition method can accurately characterize the steady-state and dynamic characteristics of welding process parameters, help analyze their impact on welding quality, and provide a basis for subsequent compensation measures. Mathematical methods such as harmonic analysis and time series decomposition can be used to achieve the decomposition of the above parameters.

[0110] The specific implementation of step S60 is as follows:

[0111] Based on the oil mist density characteristic matrix obtained above and process parameter fluctuation matrix , establish the image enhancement coefficient matrix :

[0112] ;

[0113] in, express The partial derivative of the matrix with respect to time reflects the dynamic change characteristics of the oil mist concentration; express The partial derivative of the matrix with respect to time reflects the dynamic characteristics of process parameter fluctuations; is the enhancement coefficient, and its value range is 0.5-2; is the compensation factor.

[0114] This partial derivative-based modeling method can accurately characterize the dynamic impact of oil mist and process parameters on image quality, providing a key basis for subsequent image optimization.

[0115] The specific implementation of step S70 is as follows:

[0116] Using the image enhancement coefficient matrix obtained in step S60 , for the original weld image Optimize the process to get the optimized weld image :

[0117] ;

[0118] in, is the error term introduced in the optimization process.

[0119] This is based on The matrix image optimization method can effectively eliminate the image quality degradation caused by oil mist interference and process parameter fluctuations, and improve the clarity and reliability of the weld image.

[0120] The specific implementation of step S80 is as follows:

[0121] Based on the steady-state component of the posture obtained in step S30 and the process parameter steady-state matrix obtained in step S50 , calculate the position compensation of the welding robot :

[0122] ;

[0123] in, represents the compensation cycle, is the error term introduced in the compensation process.

[0124] The posture compensation The pose fluctuation component Steady-state characteristics of process parameters The combination of dynamic characteristics and steady-state characteristics can effectively reduce the influence of posture fluctuation and process parameter fluctuation on welding quality.

[0125] The specific implementation of step S90 is as follows:

[0126] First, the optimized weld image data As the input of the machine vision module, image processing algorithms are used to detect and track welds in real time, providing reliable visual feedback for welding quality control.

[0127] At the same time, the corrected welding robot posture data calculated in step S80 is fed back to the motion control module to achieve precise adjustment of the welding trajectory.

[0128] This integrated feedback control scheme based on vision and posture can significantly improve the stability and repeatability of welding quality, and lays a solid technical foundation for the stable control of the automatic welding process of oil tank pontoon plates.

[0129] In general, the visual optimization method of the oil tank pontoon plate welding robot proposed in this invention makes full use of the welding process data collected by multiple sensors, adopts mathematical modeling methods such as multi-scale decomposition, time series decomposition, and multivariate analysis, and constructs a comprehensive feature model including weld image features, robot posture features, process parameter features, and oil mist density features. Based on these feature models, a dynamic compensation image optimization algorithm and posture correction algorithm are proposed to achieve accurate modeling and effective compensation of factors affecting welding quality. This method not only improves the clarity and reliability of the weld image, but also improves the robot's motion accuracy, laying a solid technical foundation for the stable control of the automatic welding process of the oil tank pontoon plate.

[0130] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned oil tank pontoon plate welding robot visual optimization method.

[0131] A third aspect of the present invention provides a petroleum storage tank pontoon plate welding robot visual optimization system, which includes the above-mentioned computer-readable storage medium.

[0132] Specifically, the principle of the present invention is: the visual optimization method proposed in the present invention fully considers various factors affecting welding quality and constructs a multi-dimensional feature model. Based on these models, a dynamic compensation mechanism is designed to effectively suppress visual interference and posture errors, greatly improving the stability and reliability of welding quality.

[0133] The following is an example of a specific application scenario of the present invention: A certain oil company is building a new 500,000 cubic meter large storage tank, which adopts a floating roof structure design. The pontoon plate is a key component of the floating roof, and its welding quality directly affects the safety of the entire tank system. Therefore, the company decided to adopt the oil tank pontoon plate welding robot vision optimization method proposed in the present invention to improve the stability and repeatability of the welding quality.

[0134] First, a six-axis industrial robot was installed at the pontoon plate welding workstation, equipped with various sensor devices. The high-definition camera collects weld image data in real time, with an acquisition frequency of 80Hz. The robot body sensor measures the position coordinates and attitude angle of the robot end effector, and the synchronous acquisition frequency is also 80Hz. The welding machine power supply monitors the welding current and voltage, the vibration sensor detects the vibration frequency of the pontoon plate, and the smoke detector measures the oil mist concentration in the welding area. The acquisition frequency of these three parameters is 50Hz. All collected data is transmitted to the industrial computer for real-time processing and analysis.

[0135] Based on the multi-sensor data collected above, the weld image I(x, y) is first decomposed by wavelet transform. The decomposition level n=5 is used to obtain the approximate components to And detail component D. As shown in Table 1, the low-frequency approximation component It reflects the basic morphological characteristics of the weld, the medium and high frequency components and Contains some detailed texture information, while the high-frequency components , and D are mainly noise components.

[0136] Table 1 Wavelet transform decomposition results

[0137]

[0138] Based on this, the steady-state feature matrix of the weld image is constructed. and the dynamic feature matrix .in, Each element of It represents the steady-state eigenvalue of the image block in the i-th row and j-th column, reflecting the basic morphology of the weld; Each element of It represents the dynamic feature value of the corresponding image block, which includes the slight texture changes on the weld surface.

[0139] At the same time, the time series data of the robot posture P(t) is analyzed. P(t) is divided into steady-state components and the fluctuation component ,in Can be further broken down into:

[0140] ;

[0141] Among them, the amplitude coefficient , angular frequency rad / s,Phase ;Attenuation coefficient , time constant ;Acceleration coefficient This reflects the dynamic fluctuation characteristics of the robot posture during the actual welding process.

[0142] In addition, oil mist concentration sampling points were set up at 10 different locations in the welding area, and the measured data are shown in Table 2. Using multiple regression analysis, the oil mist density characteristic matrix was established. :

[0143] ;

[0144] The oil mist diffusion coefficient is , is the measurement error term. This matrix reflects the distribution characteristics of oil mist concentration at different locations.

[0145] Table 2 Oil mist concentration sampling data

[0146]

[0147] At the same time, the three process parameters of welding current I(t), voltage U(t) and floating plate vibration frequency F(t) were also decomposed to obtain steady-state components and fluctuating components:

[0148] , ;

[0149] , ;

[0150] , ;

[0151] Based on the above various feature data, the image enhancement coefficient matrix is ​​first established :

[0152] ;

[0153] in, , , enhancement factor , compensation factor .this The matrix reflects the combined impact of oil mist concentration dynamics and process parameter fluctuations on image quality.

[0154] Yes After the matrix is ​​obtained, the original weld image I(x, y) can be optimized to obtain the optimized weld image :

[0155] ;

[0156] in, is the error term introduced in the optimization process. This image optimization method based on dynamic features effectively eliminates the image quality degradation caused by oil mist and process parameter fluctuations.

[0157] At the same time, the present invention also compensates for the dynamic characteristics of the robot's posture. and Matrix, calculate the pose compensation :

[0158] ;

[0159] The compensation cycle , is the error term in the compensation process. This posture compensation method based on dynamic characteristics and steady-state characteristics effectively reduces the impact of robot posture fluctuation on welding quality.

[0160] Based on the above visual optimization and posture compensation measures, the specific operation process in the automatic welding process of the oil tank pontoon plate is as follows:

[0161] 1. Start the welding robot and various sensor devices to collect weld images, robot posture, welding process parameters and oil mist concentration data in real time.

[0162] 2. Process and analyze the collected data, including wavelet transform decomposition, time series decomposition, multivariate analysis, etc., to construct steady-state feature and dynamic feature matrices.

[0163] 3. Based on the feature matrix, calculate the image enhancement coefficient matrix and posture compensation .

[0164] 4. Utilize The original weld image I(x,y) is optimized to obtain a clear and reliable At the same time, Feedback is sent to the robot motion control module to achieve precise adjustment of the welding trajectory.

[0165] 5. The optimized weld image and corrected robot posture data are used as the input of the visual servo control system to track and correct the welding trajectory in real time to ensure the welding quality.

[0166] In practical applications, this visual optimization method has been successfully applied in a 500,000 cubic meter storage tank construction project of a certain oil company. After long-term operation tests, the clarity and reliability of weld images have been significantly improved, and the stability and repeatability of welding quality have also been greatly improved. At the same time, the robot's motion accuracy has also been significantly improved, and the welding efficiency and production efficiency have also been greatly improved.

[0167] It should be noted that the variables, subscripts and constants involved in the full text are explained as shown in Table 3 below:

[0168] Table 3 Variable explanation table

[0169]

[0170] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A visual optimization method for a welding robot for a petroleum tank pontoon plate, characterized in that: The following steps are involved: S10, collecting weld image data, welding robot posture data, welding current data, welding voltage data, oil mist concentration data, and floating plate vibration frequency data during the welding process of the oil storage tank floating plate; S20, performing multi-scale decomposition calculation on the weld image data to obtain a weld image steady-state feature matrix and a weld image dynamic feature matrix; S30, performing multivariate decomposition calculation on the posture data of the welding robot to obtain a posture steady-state component and a posture fluctuation component; S40, performing multivariate analysis and calculation on the oil mist concentration data to establish an oil mist density characteristic matrix; S50, performing multi-dimensional decomposition on the welding current data, welding voltage data, and floating plate vibration frequency data to obtain a welding process parameter steady-state matrix and a welding process parameter fluctuation matrix; S60, establishing an image interference compensation model based on the oil mist density characteristic matrix and the welding process parameter fluctuation matrix, and calculating an image enhancement coefficient matrix; S70, optimizing the weld image data using the image enhancement coefficient matrix to obtain optimized weld image data; S80, calculating the welding robot posture compensation amount based on the posture fluctuation component and the welding process parameter steady-state matrix to obtain corrected welding robot posture data: wherein, corrected welding robot posture data = original welding robot posture data + welding robot posture compensation amount; S90, outputting the optimized weld image data and the corrected welding robot posture data for use as input of a machine vision module of a welding robot control system.

2. The oil tank pontoon plate welding robot visual optimization method according to claim 1 is characterized in that: The multi-scale decomposition calculation process of the weld image in S20 is specifically expressed as follows: ; In the formula, is the original weld image; For the Layer approximation component; For the details; is the number of decomposition layers, ranging from 3 to 5; is the image noise error term; is the image coordinate.

3. The visual optimization method for the oil tank pontoon plate welding robot according to claim 2 is characterized in that: The specific calculation process of the pose multivariate decomposition in S30 is as follows: ; In the formula, is the robot posture time series data; is the steady-state component of posture; is the posture fluctuation component; is the pose measurement error term; is the time variable.

4. The oil tank pontoon plate welding robot visual optimization method according to claim 3 is characterized in that: The establishment process of the oil mist density characteristic matrix in S40 is specifically expressed as follows: ; In the formula, is the oil mist concentration sampling point data; is the oil mist diffusion coefficient; are the number of rows and columns of sampling points respectively; is the oil mist measurement error term.

5. The oil tank pontoon plate welding robot visual optimization method according to claim 4 is characterized in that: The specific decomposition process of welding process parameters in S50 is as follows: ; ; In the formula, is the steady-state matrix of process parameters; is the process parameter fluctuation matrix; are the steady-state components of current, voltage, and vibration frequency respectively; They are the fluctuation components of current, voltage and vibration frequency respectively.

6. The oil tank pontoon plate welding robot visual optimization method according to claim 5 is characterized in that: The calculation process of the image enhancement coefficient matrix in S60 is specifically expressed as follows: ; In the formula, is the enhancement coefficient, and its value range is 0.5-2; is the compensation factor.

7. The oil tank pontoon plate welding robot vision optimization method according to claim 6 is characterized in that: The image optimization process in S70 is specifically expressed as follows: ; In the formula, For the optimized image; is the error term of the optimization process.

8. The oil tank pontoon plate welding robot vision optimization method according to claim 7 is characterized in that: The calculation process of the posture compensation in S80 is specifically expressed as follows: ; In the formula, For the compensation cycle; To compensate for the error term.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the oil tank pontoon plate welding robot vision optimization method according to any one of claims 1 to 8.

10. The oil tank pontoon plate welding robot visual optimization system is characterized by: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images