Welding arc image collecting and processing system and method

Abstract

The invention relates to a welding arc image collecting and processing system and method, and relates to the technical field of welding arc image processing. Comprising an underwater CCD camera module used for carrying out image acquisition on the arc form, a video acquisition card used for obtaining video images and carrying out preprocessing, and an image processing module used for carrying out binarization processing on the images and carrying out analysis processing on characteristic parameters. According to the method, the video image is sliced, it is guaranteed that the frame-by-frame image in each slice is in a one-time continuous welding process, the data reference value in feature parameter extraction and analysis processing is guaranteed, meanwhile, the interest window image is obtained through segmentation, and the subsequent data operation and storage pressure is reduced; and meanwhile, a characteristic parameter-time sequence changing curve and a time sequence moving window are utilized, parameter stability in the corresponding window is analyzed, welding state parameters of a welding gun are associated after time delay is compensated, the mapping relation between the welding state parameters and corresponding arc images is established through machine learning, and data reference is provided for subsequent adjustment of the welding state parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of welding arc image processing, and in particular to a welding arc image acquisition and processing system and method. Background Art

[0002] Traditional arc welding is generally performed on land, using shielding gas to provide a stable gas medium to the arc space, which helps the arc burn stably. Although the environment is relatively good and the operation is more convenient, the assembly and maintenance of marine engineering structures cannot meet the conditions for onshore arc welding, and wet underwater arc welding is required.

[0003] The wet underwater welding arc burns in the arc bubble, which is one of the special phenomena that distinguishes the wet underwater welding process from the onshore welding process. After the arc is ignited underwater, the water molecules are decomposed into hydrogen and oxygen by the arc heat. The oxygen is consumed in the combustion reaction of the flux core and the burning of the metal during the droplet transition process; therefore, the arc is separated from the water by a mixture of hot gases, water vapor, hydrogen decomposed from water, and other gases caused by the arc. Affected by the burning process of the welding arc, these gases interact in the bubble, causing the gas composition to change continuously with time and the position of the bubble. The arc bubble starts with a small bubble around the arc column, and then gradually grows and floats up. In this process, the bubble is only partially broken, leaving a core bubble. The dynamic process of the bubble is the result of the combined effects of bubble buoyancy, gravity, pressure inside the bubble, and hydrostatic pressure, and plays an important role in the stability of the underwater welding arc; Wet underwater welding generally uses a manipulator to enter the underwater area for welding. Welding is a continuous process, which requires real-time acquisition of welding arc image data to monitor the welding environment status, welding equipment status and weld quality, so that compensatory measures can be taken in time after problems occur to ensure the normal and orderly progress of welding work. All of this depends on clear welding arc images and fast image analysis and processing. To this end, we provide a welding arc image acquisition and processing system and method. Summary of the invention

[0004] The object of the present invention is to provide a welding arc image acquisition and processing system and method.

[0005] The technical problems solved by the present invention are: (1) Existing welding image acquisition and processing cannot quickly analyze and provide guidance for subsequent welding in a short time; (2) During a welding process, the arc is not always lit. In a discontinuous welding process, the welding environment and parameters change greatly, making it impossible to extract and process effective parameters.

[0006] The present invention can be implemented by the following technical solutions: A welding arc image acquisition and processing system, comprising: an underwater CCD camera module, used for acquiring images of arc morphology during underwater welding; A filter module is installed at the front end of the lens of the underwater CCD camera module to protect the underwater CCD camera module and filter the arc light to ensure the imaging quality. The filter module includes a neutral color filter, a narrow-band color filter and a heat shield; The video acquisition card obtains the video images collected by the underwater CCD camera module for preprocessing, obtains the interest window image in each frame-by-frame image of multiple complete slices, and performs noise reduction filtering, bad pixel compensation and contrast enhancement on the interest window image; The image processing module obtains the interest window image preprocessed by the video acquisition card, performs binarization processing to obtain the arc image area, and obtains the arc characteristic parameters, and then performs correlation analysis processing on the working state of the welding gun based on the arc characteristic parameters.

[0007] A further technical improvement of the present invention is that the lens of the underwater CCD camera module is perpendicular to the arc column plane of the welding arc, and the distance between the two is maintained between 50 and 80 mm.

[0008] A welding arc image acquisition and processing method comprises the following steps: Step 1: After adjusting the positions of the welding gun and the underwater CCD camera module, start the underwater CCD camera module to capture images of the welding arc and output them to a video capture card; Step 2: The video capture card slices and segments the video image to obtain an interest window image, and performs noise reduction and enhancement processing on the interest window image; Step 3: The image processing module performs image binarization processing on the interest window image sequence one by one to obtain the arc image area, and obtains the arc characteristic parameters based on the arc image area; Step 4: construct a characteristic parameter-timing change curve based on the arc characteristic parameters, set a timing moving window to divide the corresponding characteristic parameter-timing change curve in the complete slice, and then analyze the parameter stability in the timing moving window; Step 5: record the arc characteristic parameters in the timing moving window lacking stability and associate them with the welding state parameters of the welding gun under timing synchronization, compare the welding state parameters under the unstable state with the welding state parameters under the stable state, and obtain the abnormal state parameters; Step 6: Map the abnormal state parameters to the arc image under synchronous timing based on machine learning, and quickly obtain the arc image in the subsequent welding process to match the abnormal state parameters of the welding gun to provide a reference for correction.

[0009] A further technical improvement of the present invention is that the process of the video capture card slicing the video image includes: The slice start position is defined as the K1 specific frame images before the specific frame image where the arc light first appears in the image sequence, and the end position is defined as the K2 specific frame images after the specific frame image where the arc light disappears, thereby forming a complete slice.

[0010] A further technical improvement of the present invention is that the interest window image is the image region where the arc light is located in the frame-by-frame image, and the pixel size of the region is set to a specific size.

[0011] A further technical improvement of the present invention is that the steps of performing binary image processing on the interest window image are: Perform grayscale processing on the interest window image; Obtain the global binary segmentation threshold of the image corresponding to the window of interest; The global binary segmentation threshold is used to distinguish the pixel set in the interest window image, and the part with gray value lower than the global binary segmentation threshold is regarded as the background, otherwise it is regarded as the foreground, and the boundary markers between the foreground and the background are extracted to obtain multiple closed connected domains; The closed connected domain with the largest area is taken as the arc image region.

[0012] A further technical improvement of the present invention is that the process of obtaining the global binary segmentation threshold comprises: Assume the total number of pixels in the image is , the gray level of the image is , then the grayscale range of the image is , the gray value is The number of pixels is ,but The gray value of the pixel is The probability is: ; Arbitrarily set a temporary separation threshold , the grayscale range is divided into two intervals, namely and , then calculate the probability value of the gray value being in the corresponding interval and ; Then we have: The grayscale means in the two intervals are and ; The global grayscale arithmetic contour mean is ; Will As an independent variable, and , construct the variance function of the grayscale mean relative to the global grayscale arithmetic profile mean in the two intervals, solve the maximum value of the variance function, and when the variance reaches the maximum value, the corresponding As the global binarization segmentation threshold.

[0013] The further technical improvement of the present invention is that: the width of the time sequence moving window is dynamically set in a positive correlation with the number of image frames in the complete slice, and the moving step length is one third of the time sequence moving window length; When calculating the stability of arc characteristic parameters in the timing moving window, the variance value of each characteristic parameter in the current window is calculated and compared with the set threshold. If it does not exceed the set threshold, the arc characteristic parameter is unstable, and vice versa.

[0014] A further technical improvement of the present invention is that when the arc characteristic parameters are synchronously associated with the welding state parameters of the welding gun, it is necessary to compensate for the shooting delay of the underwater CCD camera module and then determine the corresponding associated time of the welding gun.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses a video capture card to slice the video images collected by the underwater CCD camera module, thereby ensuring that each frame of the image in each complete slice is in a continuous welding process, ensuring that the data reference value is improved in the subsequent arc characteristic parameter extraction and stability analysis, and thus associated with the welding state parameters, providing data support for the subsequent welding state parameter correction; 2. Use the video acquisition card to define the region of interest for each frame to obtain the window image of interest, which greatly reduces the order of magnitude of useless interference information, reduces the subsequent data calculation pressure and data storage pressure, and thus improves the accuracy and efficiency of subsequent image processing.

[0016] 3. The present invention performs binary image processing on the interest window image through the image analysis module, and adopts the global segmentation threshold to improve the processing accuracy, so as to obtain the arc image area, thereby facilitating the subsequent extraction of arc characteristic parameters, and uses the time-series moving window to slide on the characteristic parameter-time-series change curve constructed based on the arc characteristic parameters, determine the volatility of the arc characteristic parameters, associate the welding state parameters of the welding gun after compensating for the time delay, compare the welding state parameters associated with the stable state arc characteristic parameters with the welding state parameters associated with the arc characteristic parameters in the fluctuating state, obtain the abnormal state parameters, and use machine learning to establish a mapping relationship between it and the corresponding arc image. When encountering an arc image in a similar state in the future, the abnormal state parameters can be quickly matched and adjusted, so that the welding level and quality are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0018] Figure 1 is a system block diagram of the present invention; Figure 2 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0019] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0020] See also Figure 1 As shown, a welding arc image acquisition and processing system includes a filter module, an underwater CCD camera module, a video acquisition card and an image processing module; The filter module is set up because the arc light intensity of the welding arc greatly exceeds the normal working range of the CCD device. Direct video shooting will cause image distortion or even damage the CCD device. The filter module uses a neutral filter to reduce the incident welding arc light to an acceptable range, and a narrow-band filter is used to ensure that detailed information such as the molten pool image will not be lost in large quantities. In addition, a heat shield is set up to prevent interference from infrared light. The underwater CCD camera module uses FM105-1 / 3" black and white CCD camera. The glass hemispherical waterproof sealed window and sealed optical aluminum alloy shell of the entire module are protected to prevent corrosion and damage in the underwater environment and welding process. During the video acquisition process, in order to obtain the actual state of the welding arc as much as possible and avoid the need for subsequent geometric correction of the image, the lens axis of the camera is always kept perpendicular to the arc column plane of the welding arc during the video image acquisition process, and the distance is kept between 50mm and 80mm.

[0021] The video capture card is used to pre-process the video images collected by the underwater CCD camera module, and convert the pre-processed image signals into digital signals and store them in the computer. The specific steps of the video capture card pre-processing the video images include: Automatically cut the video image length into multiple slices; The video images of each slice are arranged and numbered continuously in a frame-by-frame manner; Perform interest window segmentation on each frame of each slice to obtain an interest window image; The image in the window of interest is subjected to preprocessing operations such as noise reduction filtering, bad pixel compensation and contrast enhancement.

[0022] Furthermore, the above process of generating multiple slices is specifically as follows: The starting position of the slice is the K1 specific frame images before the specific frame image where the arc light appears for the first time in the image sequence, and the end position is the K2 specific frame images after the specific frame image where the arc light disappears, thereby forming a complete slice. A complete slice is a continuous welding process; The window of interest image mainly presents the effective rectangular area where the arc light is located. The size of the effective rectangular area is determined by a preset pixel size, thereby isolating most of the interference information outside the effective rectangular area, reducing the computing power consumption in the subsequent processing process and improving the image processing efficiency.

[0023] The image processing module performs image processing and parameter extraction analysis on the interest window image sequence one by one. The specific steps include: S1, determining the image interference items existing in the interest window image: the reflection of the arc light from the welding nozzle, the reflection of the arc light from the welded workpiece, and the splash of the molten pool; S2. Perform grayscale processing on the interest window image and perform image binarization processing on it: S21, obtaining the global binarization segmentation threshold of the interest window image: Assume the total number of pixels in the image is , the gray level of the image is , then the grayscale range of the image is , the gray value is The number of pixels is ,but The gray value of the pixel is The probability is: ; Arbitrarily set a temporary separation threshold , the grayscale range is divided into two intervals, namely and , then calculate the probability value of the gray value being in the corresponding interval and ; The grayscale means in the two intervals are and ; Then the global grayscale arithmetic contour mean is ; Further calculate the variance of the grayscale mean in the two intervals relative to the global grayscale arithmetic profile mean, that is, ; Will As an independent variable, and , solve the maximum value of the above variance function value, and when the variance reaches the maximum value, the corresponding As the global binary segmentation threshold; S22, taking the pixel point set whose gray value in the interest window image is lower than the global binary segmentation threshold as the background part and marking it as black, taking the pixel point set whose gray value in the interest window image is higher than the global binary segmentation threshold as the foreground part and marking it as white, thereby determining the boundary between the foreground part and the background part, extracting the boundary marker, and then obtaining multiple closed connected domains; taking the closed connected domain with the largest area as the arc image area, and defining the remaining closed connected areas as the splash image area; S3, extracting arc characteristic parameters such as arc shape, arc length, arc column area and circumference of the arc based on the arc image area, wherein the arc shape includes regular, curved, bifurcated, etc.; The arc morphology is identified based on the recognition and judgment model of the BP neural network. The recognition and judgment model uses a large number of manually calibrated arc images to construct a training set. The recognition and judgment model is trained with arc images as input and arc morphology as output. S4, constructing multiple characteristic parameter-time series change curves using arc characteristic parameters of an image sequence in a complete slice, wherein a change curve can also be generated after setting different state values ​​for each arc form; S5, setting a dynamic temporal moving window according to the number of image frames in the image sequence in the complete slice, wherein the width of the temporal moving window is positively correlated with the number of image frames; and the moving step length is one third of the temporal moving window length; The time series moving window is slid on the characteristic parameter-time series change curve according to the moving step. In each sliding process, the variance value of each characteristic parameter in the current window is calculated, that is, the data fluctuation. When the variance value of a characteristic parameter exceeds the set threshold, the characteristic parameter value of the current window is recorded and marked as the fluctuation parameter value; S6. According to the occurrence time period of the fluctuation parameter value recorded in step S5 and the shooting delay of the underwater CCD camera module, the corresponding associated time period of the welding gun working corresponding to the fluctuation parameter value is calculated, and the welding state parameters of the welding gun in the associated time period are obtained, including welding current, arc voltage, welding speed, length of the tungsten electrode extending out of the nozzle, and distance between the tungsten electrode and the weldment. The welding state parameters in the associated time period are compared with the welding state parameters when the characteristic parameter fluctuation is at a set threshold, so as to obtain abnormal state parameters, and establish a mapping relationship between the abnormal state parameters and the arc image based on machine learning; S7. During the welding process, the arc image is quickly acquired according to step S2 and step S3 to obtain the arc shape and associate the abnormal state parameters of the welding gun under the shape and make targeted adjustments.

[0024] like Figure 2 As shown, the arc image acquisition and processing method based on the above-mentioned welding arc image acquisition and processing system includes: After the underwater manipulator reaches the welding working position, the lens axis of the underwater CCD camera module is adjusted to be perpendicular to the arc column plane of the welding arc, and the distance between the lens and the arc column plane is adjusted; When the robot drives the welding gun to weld, the underwater CCD camera module is started to collect images of the welding arc; The collected image data is pre-processed by the video capture card and transmitted to the image processing module in the computer for analysis and processing. The pre-processing of the video capture card includes: Automatically slice the video image to generate multiple complete slices, and arrange and number each complete slice in the form of frame-by-frame images, and then segment them to obtain interest window images; Preprocessing operations also include noise reduction filtering, bad pixel step size, and contrast enhancement; The image processing module performs image binarization processing on the image sequence in the complete slice one by one to obtain the arc image area and obtain the arc characteristic parameters; The image processing module also constructs a characteristic parameter-time series change curve based on the arc characteristic parameters obtained from the complete slice, and uses a time series moving window with a certain width and step length to slide on the curve, and records the parameter values ​​in the window where the characteristic parameter fluctuates greatly, marking them as fluctuating parameter values; according to the time period when the fluctuating parameter value occurs, the welding state parameters in the corresponding associated time period are obtained after time delay correction, and the mapping between abnormal state parameters and arc images is established; In the subsequent actual welding process, the arc image is quickly acquired and associated with the abnormal state parameters of the welding gun under the image state, providing a reference for the subsequent adjustment of the welding gun working parameters.

[0025] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A welding arc image acquisition and processing system, characterized in that: include: Underwater CCD camera module, used to capture images of arc morphology during underwater welding; A filter module is installed at the front end of the lens of the underwater CCD camera module to protect the underwater CCD camera module and filter the arc light to ensure the imaging quality. The filter module includes a neutral color filter, a narrow-band color filter and a heat shield; The video acquisition card obtains the video images collected by the underwater CCD camera module for preprocessing, obtains the interest window image in each frame-by-frame image of multiple complete slices, and performs noise reduction filtering, bad pixel compensation and contrast enhancement on the interest window image; The image processing module obtains the interest window image preprocessed by the video acquisition card, performs binarization processing to obtain the arc image area, and obtains the arc characteristic parameters, and then performs correlation analysis processing on the working state of the welding gun based on the arc characteristic parameters.

2. A welding arc image acquisition and processing system according to claim 1, characterized in that: The lens of the underwater CCD camera module is perpendicular to the arc column plane of the welding arc, and the distance between the two is maintained between 50 and 80 mm.

3. A welding arc image acquisition and processing method according to any one of claims 1 to 2, characterized in that: The steps include: Step 1: After adjusting the positions of the welding gun and the underwater CCD camera module, start the underwater CCD camera module to capture images of the welding arc and output them to a video capture card; Step 2: The video capture card slices and segments the video image to obtain an interest window image, and performs noise reduction and enhancement processing on the interest window image; Step 3: The image processing module performs image binarization processing on the interest window image sequence one by one to obtain the arc image area, and obtains the arc characteristic parameters based on the arc image area; Step 4: construct a characteristic parameter-timing change curve based on the arc characteristic parameters, set a timing moving window to divide the corresponding characteristic parameter-timing change curve in the complete slice, and then analyze the parameter stability in the timing moving window; Step 5: record the arc characteristic parameters in the timing moving window lacking stability and associate them with the welding state parameters of the welding gun under timing synchronization, compare the welding state parameters under the unstable state with the welding state parameters under the stable state, and obtain the abnormal state parameters; Step 6: Map the abnormal state parameters to the arc image under synchronous timing based on machine learning, and quickly obtain the arc image in the subsequent welding process to match the abnormal state parameters of the welding gun to provide a reference for correction.

4. A welding arc image acquisition and processing method according to claim 3, characterized in that: The process of the video capture card slicing the video image includes: The slice start position is defined as the K1 specific frame images before the specific frame image where the arc light first appears in the image sequence, and the end position is defined as the K2 specific frame images after the specific frame image where the arc light disappears, thereby forming a complete slice.

5. A welding arc image acquisition and processing method according to claim 3, characterized in that: The interest window image is an image region where the arc light is located in the frame-by-frame image, and the region is set to a specific pixel size.

6. A welding arc image acquisition and processing method according to claim 3, characterized in that: The steps of performing binary image processing on the interest window image are as follows: Perform grayscale processing on the interest window image; Obtain the global binary segmentation threshold of the image corresponding to the window of interest; The global binary segmentation threshold is used to distinguish the pixel set in the interest window image, and the part with gray value lower than the global binary segmentation threshold is regarded as the background, otherwise it is regarded as the foreground, and the boundary markers between the foreground and the background are extracted to obtain multiple closed connected domains; The closed connected domain with the largest area is taken as the arc image region.

7. A welding arc image acquisition and processing method according to claim 6, characterized in that: The process of obtaining the global binary segmentation threshold includes: Assume the total number of pixels in the image is , the gray level of the image is , then the grayscale range of the image is , the gray value is The number of pixels is ,but The gray value of the pixel is The probability is: ; Arbitrarily set a temporary separation threshold , the grayscale range is divided into two intervals, namely and , then calculate the probability value of the gray value being in the corresponding interval and ; Then we have: The grayscale means in the two intervals are and ; The global grayscale arithmetic contour mean is ; Will as independent variables, and , construct the variance function of the grayscale mean relative to the global grayscale arithmetic profile mean in the two intervals, solve the maximum value of the variance function, and when the variance reaches the maximum value, the corresponding As the global binarization segmentation threshold.

8. A welding arc image acquisition and processing method according to claim 3, characterized in that: The width of the temporal moving window is dynamically set in a positive correlation with the number of image frames in the complete slice, and the moving step length is one-third of the temporal moving window length; When calculating the stability of arc characteristic parameters in the timing moving window, the variance value of each characteristic parameter in the current window is calculated and compared with the set threshold. If it does not exceed the set threshold, the arc characteristic parameter is unstable, and vice versa.

9. A welding arc image acquisition and processing method according to claim 3, characterized in that: When the arc characteristic parameters are synchronously associated with the welding state parameters of the welding gun, it is necessary to compensate for the shooting delay of the underwater CCD camera module and then determine the corresponding associated time of the welding gun.

Images