CN103506756B - Laser lap welding gap detection system and method based on molten pool image vision sensing - Google Patents

Abstract

The invention discloses a laser lap welding gap detecting system and a laser lap welding gap detecting method based on molten pool image visual sensing. The detecting system is a visual sensing system and comprises a CMOS camera, a light filtering system, a secondary light source, an image capture card, a computer, a display and the like, wherein the CMOS camera has a LinLog photosensory technology. According to the detecting method, the secondary light source and emitted light of molten pools are used as light sources; plasma is filtered out by utilizing the light filtering system, and light intensity is adjusted; the computer is used for collecting and displaying molten pool images in real time, wherein the molten pool images are obtained by the CMOS camera; molten pool image edges and areas and orifice image edges and areas are extracted and calculated by utilizing a Labview image processing platform, and the quantitative relations between molten pool sizes and orifice areas and between the molten pool sizes and orifice gaps are obtained. The detecting system is simple in structure, clear in monitoring image, high in light signal detecting accuracy, strong in anti-jamming capacity, good in engineering practicability and capable of monitoring laser welding of T-type overlap joint gaps in real time.

Description

Laser lap welding gap detection system and method based on molten pool image vision sensing

technical field

The invention relates to the field of welding technology, in particular to a laser lap welding gap detection system and method based on molten pool image visual sensing.

Background technique

Laser welding has the advantages of large aspect ratio, small heat-affected zone and thermal deformation, and high production efficiency. In recent years, with the development of laser devices, the quality and power of laser beams have been greatly improved. Therefore, laser welding has been increasingly used used in industrial production. The laser welding process is accompanied by a variety of light, sound, heat, and electric radiation phenomena, and these signals contain a lot of information related to the welding process and welding quality. Among them, the dynamic change of the laser welding molten pool has a considerable relationship with the stability of the welding process and the occurrence of defects. At the same time, the dynamic changes of the molten pool and the small hole are closely related and affect each other, which jointly determine the stability of the welding process and welding quality. Real-time observation of the dynamic changes and physical process characteristics of the molten pool during the welding process can reveal the physical mechanism of the welding process, guide welding production and improve welding quality.

Laser welding T-shaped lap joint is a new type of lap joint structure. The welding of this type of structure needs to penetrate the upper plate and connect the lower plate. Laser welding has unique advantages, and there are few related researches on it at present. In the deep penetration lap welding, it is difficult to avoid the gap between the upper and lower plates due to the influence of workpiece machining accuracy, fixture and welding deformation. The existence of gaps will definitely affect the stability of the welding process and cause welding defects. Therefore, if the gap can be monitored in real time, the stability of the laser lap welding process can be controlled. The stability of the molten pool during the welding process is directly related to the gap, and its dynamic change characteristics are closely related to the gap change. Therefore, the development of laser lap welding gap detection technology based on molten pool image vision sensing has become the key to ensure the quality of laser welded T-shaped lap joints.

Vision sensing monitoring systems have been widely used in different industrial fields for a long time. The welding process stability monitoring and tracking sensing methods generally include contact type, arc type, electromagnetic type, photoelectric type, visual type and so on. Among them, the visual sensor is considered to be the most promising sensor due to its advantages of no contact with the workpiece, large amount of information (can also be used as monitoring), strong anti-electromagnetic interference ability, high sensitivity, and many applicable groove forms. method. At present, commonly used visual sensors are divided into charge-coupled device (CCD) type and complementary metal-oxide-semiconductor (CMOS) type, which can convert light signals of different intensities into image information of different amplitudes. Compared with the CCD type, the CMOS sensor has the advantages of high chip integration, low power consumption, fast response speed, and wide dynamic range. Logarithmic amplification is performed on high-brightness signals to extend the dynamic range as much as possible while ensuring the image contrast in low-brightness areas, so it is especially suitable for working in environments with high contrast between light and dark like welding. LinLog technology is a special photosensitive technology, which is often used to shoot high-contrast images. The basic principle of this technology is to use logarithmic compression technology to compress the super bright areas in the image and reduce the image contrast. LinLog technology enables the sensor to compress the response logarithmically only in areas close to saturation and saturation, while retaining its linear response and sensitivity in low-illumination areas, so it not only expands the dynamic range of the entire sensor, but also ensures the quality of imaging in low-illumination areas. Therefore, it has become an important research direction to use machine vision to directly observe the interaction area between the laser and the workpiece, obtain the characteristic information of the molten pool and small holes through image processing, establish a real-time monitoring system for the welding process, and realize the closed-loop control of welding quality.

At present, there are many researches on the visual sensing technology of laser welding, but there is no research on the use of visual sensing technology to monitor and track the laser lap welding gap. After literature search, in the article titled "Detection of gaps in laser lap welding process" (Isamu Miyamoto et al., Laser Engineering, 1996, Vol. 24, No. 9, pp. 67-69), Isamu MIYAMOTO et al. used The photodiode monitors the relationship between the signal intensity of the plasma in the hole and the plasma outside the hole and the gap during the laser lap welding process. The study found that the frequency of the AC signal emitted by the plasma light reaches about 10kHz, and when the gap is greater than 0.3mm and the frequency is 4-6 kHz, the mean square value of the AC changes suddenly. This method finds that the signal intensity and peak frequency of the plasma in the hole have a good correspondence with the gap through the calculation of the signal intensity and Fourier transform. In this paper, the plasma signal collected by the photodiode cannot feed back the change of the gap in real time, and the author has not done further research on the relationship between the image characteristics of the plasma and the molten pool and the gap.

The patent application number is 03116161.8, and the patent titled "Visual Sensing Method for Dynamic Features of Arc Welding Pool" discloses a visual sensing method for dynamic features of arc welding pool. The method uses welding arc light to illuminate the welding pool, and uses The filter intercepts the light of a specific wavelength range in the reflected arc light as the imaging light source, and adjusts the light intensity with a light-reducing film, changes the direction of light propagation through the reflection of the plane mirror, and uses a CCD camera and an ordinary optical lens to capture the front of the welding pool at the same time. It is imaged on the same CCD camera target surface as the information on the reverse side. This method can obtain clear images of the front and back sides of the welding pool during the welding process, and realize effective monitoring and control of the welding process. This patented method is dynamic in the arc welding pool. Good results have been achieved in feature monitoring, but not for laser welding.

The patent application number is 201210325926.9, and the title of the invention is "Narrow Gap Welding Monitoring and Weld Deviation Detection Method Based on Infrared Vision Sensing", which discloses a narrow gap welding monitoring and welding seam deviation detection method based on infrared vision sensing. The infrared vision sensing system used includes an infrared CMOS camera, a filter system, an image acquisition card, a computer, and a display. The signal is interfered and the light intensity is adjusted, and the computer collects and displays the welding images captured by the infrared CMOS camera in real time. The image on the side away from the arc is intercepted through the image interception window for processing, and the weld deviation information is obtained by extracting the single-side edge of the groove, which can effectively avoid the interference of the moving arc and improve the real-time performance of the weld deviation detection. The patent application It is aimed at the detection of the narrow gap arc welding molten pool, but does not consider the laser welding situation.

Contents of the invention

In view of the defect that the detection system and method briefly analyzed above do not consider the laser welding situation, the present invention proposes a detection system and method based on CMOS visual sensing of laser welding T-shaped lap joints with LinLog photosensitive technology, which can monitor welding in real time The dynamic behavior of the molten pool and the orifice under different gaps in the process can extract the edge and area of the molten pool and the orifice and the periodic change information. This method can achieve the purpose of monitoring and tracking the gap in the welding process.

The invention provides a method for detecting the gap of a laser welded T-shaped lap joint based on visual sensing of molten pool images, comprising the following steps:

(1) Place the workpiece (3) on the workbench and move with the workbench, while the laser head (1) is fixed; it adopts a CMOS camera (5) with LinLog photosensitive technology, a filter system (11), and an auxiliary light source (9), image acquisition card (6), computer (7), visual sensing system of display (8), the welding image signal is sent to computer (7) through image acquisition card (6), and displayed by display (8); Place the camera (5) on the side of the workbench at 500-1000mm and form an included angle of 75-85 ^° with the surface of the workpiece (3), adjust the auxiliary light source (9) to be 300-500mm in front of the laser head (1) and be in line with the workpiece ( 3) The surface is 25-35 ^° ; adjust the image size and focal length of the LinLog photosensitive technology CMOS camera (5);

(2) Adjust the CMOS camera with LinLog photosensitive technology (5) so that it can collect the image of the entire molten pool on the surface of the workpiece (3). If the image of the molten pool is not centered or the complete molten pool image cannot be collected, adjust the CMOS camera with LinLog photosensitive technology (5) position; according to the welding speed and weld length, properly adjust the exposure and control the acquisition time and frame number, and record the acquisition frequency;

(3) Observe the dynamic changes of the molten pool body during laser welding, and analyze the changes in the size, shape, brightness and slump defects of the molten pool and orifice in real time;

(4) Import the collected melt pool image into another computer equipped with the Labview image processing platform, carry out edge extraction and area calculation of the orifice, and analyze the corresponding relationship between the orifice area and the gap; use Photoshop software to Measure the width and length of the pool, and analyze the corresponding relationship with the gap;

(5) According to the corresponding relationship between the above-mentioned orifice area and the gap amount and the corresponding relationship between the width and length of the molten pool and the gap amount, through the real-time observation of the molten pool image during the welding process and the orifice area and the molten pool width, length and The calculation of the area detects the amount of gap in the welding process in real time.

The present invention also provides a visual sensing system for laser welding T-shaped lap joint gaps based on molten pool image visual sensing, including a CMOS camera with LinLog photosensitive technology, a filter system, an auxiliary light source, an image acquisition card, a computer, and a display , the filter system is coaxially connected with the camera; the image acquisition card is placed in the card slot of the computer and connected with the LinLog photosensitive technology CMOS camera through the video line, and the welding image signal collected by the LinLog photosensitive technology CMOS camera is passed through the image acquisition card Send it to the computer; the monitor is connected to the computer to display the collected images of the welding area in real time.

Using this visual sensing system to monitor the welding process gap can have three working modes: the first working mode is to realize the laser welding T-shaped lap joint molten pool image monitoring function, and the second working mode is to realize the edge and area extraction of the orifice and the molten pool Function, working mode three is to realize the dual functions of melt pool and orifice image monitoring and edge and area extraction; for working mode one, the computer is a computer; for working mode two, the computer is a computer and Labview image processing platform; For working mode 3, the computer is a dual-computer system composed of the image acquisition of the main computer and the image processing of the slave computer. The main computer collects the image information of the welding area and sends it to the monitor for real-time display. Image comparative analysis, to find out the corresponding relationship between the change of molten pool and orifice image and the amount of gap. The above-mentioned method for detecting the gap of the laser welded T-shaped lap joint based on the visual sensing of the molten pool image of the present invention mainly uses the third working mode of the visual sensing system.

Compared with the prior art, the beneficial effects of the present invention are as follows:

First. The invention establishes a monitoring and detection method for detecting laser lap welding gaps, proposes a quantitative characterization method corresponding to the dynamic change of molten pool and the gap amount, and has carried out multiple experiments to verify that the detection accuracy can fully meet the requirements of weld seam tracking ;

second. The present invention adopts the CMOS camera with LinLog photosensitive technology to monitor the dynamic changes of the molten pool and the orifice in the laser lap welding T-joint process in real time, and obtains the size, shape and fluctuation information of the molten pool and the orifice under different gaps; the image of Labview is used The processing platform and Photoshop software carry out follow-up processing on the molten pool image, quantitatively analyze the corresponding relationship between the gap and the size of the molten pool and the area of the orifice, and provide a basis for real-time monitoring; through on-site welding and monitoring using the detection system and method of the present invention, the gap The volume tracking effect has reached the expected goal, can accurately determine the existence of the gap, and can quantitatively analyze the size of the gap, the accuracy can reach 0.1mm, which improves the welding quality and welding efficiency of laser welding T-shaped lap joints;

third. The CMOS camera used in the detection system and method is small in size, low in power consumption, fast in response, and wide in dynamic range, and is equipped with a filter system. The image quality of the molten pool in the welding area collected is high and the edges are easy to extract, and the real-time monitoring effect is good. , Strong engineering practicability.

Of course, any product implementing the present invention does not necessarily need to achieve all the above-mentioned advantages at the same time.

Description of drawings

Fig. 1 is the schematic diagram of the visual sensing system for laser welding of T-shaped lap joints in Embodiment 1 and Embodiment 2 of the present invention;

Figure 2 is a schematic diagram of the optical imaging structure of Embodiment 1 and Embodiment 2 of the present invention;

Fig. 3 is an example diagram of the real-time monitoring effect of laser welding T-shaped lap joint molten pool and orifice images under different gaps in Example 1 of the present invention;

Fig. 4 is an example diagram of the real-time monitoring effect of the complete image of the laser welded T-shaped lap joint molten pool under different gaps in Example 1 of the present invention;

Fig. 5 is a schematic diagram of the method for measuring the length and width of the molten pool in Embodiment 1 of the present invention;

Fig. 6 is the photograph of the T-shaped lap joint under the condition of no gap and the presence of gap in Embodiment 1 of the present invention;

Fig. 7 is a flow chart of extracting molten pool and orifice characteristic parameters in Embodiment 1 of the present invention;

Fig. 8 is an example of image processing of molten pool and orifice in Embodiment 1 of the present invention;

Fig. 9 is a schematic diagram before and after extraction of molten pool and orifice image contours in Embodiment 1 of the present invention;

Fig. 10 is the change of the length and the width of the molten pool with the amount of the gap in Example 1 of the present invention;

Fig. 11 is the variation curve of the molten pool area under different gap amounts in Example 1 of the present invention;

Fig. 12 is the corresponding relationship between the average area of the orifice and the gap amount in Example 1 of the present invention;

Fig. 13 is an image of the molten pool during the actual detection process of the gap amount in Embodiment 2 of the present invention.

Detailed ways

The invention is applicable to the detection of laser lap joint non-penetrating welding gaps. Before carrying out gap detection in actual production, it is necessary to test one or several groups of samples with exact gap values to determine the corresponding relationship between the gap and the molten pool parameters, and then according to these corresponding relationship diagrams, the gap value is unknown. The samples were subjected to real-time monitoring of the welding gap. Below in conjunction with the accompanying drawings, the embodiment and implementation process of the present invention will be described in detail.

Embodiment 1 The detection of the relationship between the gap amount and the molten pool parameters

The detection method of the laser welded T-shaped lap joint gap based on molten pool image visual sensing in this embodiment and embodiment 2 adopts the visual sensing system as shown in Figure 1 and Figure 2, and the system includes: LinLog photosensitive technology CMOS camera (5), filter system (11), image acquisition card (6), computer (7), display (8), wherein, computer (7) is a slave computer equipped with a Labview image processing platform for image acquisition by the main computer A dual-machine system composed of computer image processing, the main computer collects the image information of the welding area and sends it to the monitor (8) for real-time display, the computer performs image processing and calculation, and feeds back to the main computer for image comparison and analysis in time to find out the molten pool and holes The corresponding relationship between the mouth image change and the gap amount.

When the above-mentioned visual sensing system is in use, the workpiece (3) is placed on the workbench and moves with the workbench, the laser head (1) is fixed, and the camera (5) is fixed and placed on the side of the workbench and aligned with the workpiece (3) The bottom surface is at an included angle of 75 ^° , and the auxiliary light source (9) is adjusted to be 300mm in front of the laser head (1) and at 30 ^° to the surface of the workpiece (3); adjust the camera position, image size and focal length to collect the weld pool and The dynamic image of the orifice, before welding, set the CMOS camera (5), adjust the focal length and angle of the lens, so that the image of the laser focus position on the upper part of the workpiece is clearly displayed on the monitor (8), and the collected welding image signals are passed through the image acquisition card (6) Send it to the computer (7) and display it on the monitor (8); according to the welding speed and weld length, adjust the exposure appropriately and control the acquisition time and frame number, and record the acquisition frequency; among them, the zoom range of the CMOS camera (5) is 18 -45mm, the aperture value is 5.6-32, and the exposure is 1-10ms; as shown in Figure 2, the filter system (11) is coaxially connected with the camera (5), and the filter system (11) includes a narrow-band filter (13), neutral light reduction film (12), protective glass (14) (i.e. UV mirror), the central wavelength of the narrow band filter (13) is 808nm, and the transmittance of the neutral light reduction film (12) is 25%, the protective glass is used for anti-splash, using the auxiliary light source (9) and the radiated light of the molten pool (10) as the light source, the filter system (11) can effectively eliminate interference such as arc light, smoke, and splash, and use neutral The light reduction film adjusts the light intensity so that a clear image of the welding area can be collected, including an aperture (15), a zoom lens (16), a neutral light reduction film (12), a narrow-band filter (13) and a UV film (14 ) (that is, the protective glass) optical imaging system is coaxially connected with the camera (5); the image acquisition card (6) is placed in the card slot of the computer (7) and connected with the camera (5) through a video cable, and the camera (5) ) and the welding image signal collected by the image acquisition card (6) is sent to the computer (7); the monitor (8) is connected to the computer (7) to display the collected welding area image in real time.

The above-mentioned visual image sensing system and method are used to collect images of the molten pool of laser welded T-shaped lap joints. The test conditions are as follows:

Welding method: high power _CO2 laser welding;

Test material: low-alloy high-strength steel;

The laser welding equipment model is TLF15000 turbo CO ₂ fast axial-flow laser produced by TRUMPF in Germany, with a maximum output power of 15 kW, a focal length of 357 mm, a spot radius of 0.43 mm, and continuous output;

The high-speed camera adopts the MV-D1024-TrackCam of Swiss Photonfocus Company, with a maximum sampling frequency of 10000 fps;

Image acquisition card: VIEDEO-PCI-XR; light reduction film parameters: light reduction 25%;

The plasma image is captured by adjusting the relevant parameters of the high-speed camera and its auxiliary components to optimize the shooting effect.

4mm thick low-alloy high-strength steel was used for welding experiments, and the selected welding process parameters were: laser power P = 8kW, scanning speed v = 1.5m/min, defocus -2, the shielding gas used in the welding process was pure He, and the flow rate was 30L/min.

During the laser welding process, observe the dynamic changes of the molten pool body, and analyze the changes in the size, shape, brightness and slump defects of the molten pool and orifice in real time; import the collected molten pool images into another computer equipped with the Labview image processing platform, Edge extraction and area calculation are carried out on the orifice. The edge extraction is specifically: first use the median method for smoothing and filtering, then perform grayscale analysis and threshold segmentation, and finally use the Log The edge operator performs smoothing and integral processing on the thresholded image to filter out the interference information and extract the edge line; analyze the corresponding relationship between the area of the orifice and the amount of the gap; use Photoshop software to measure the width and length of the molten pool, and analyze its relationship with Corresponding relation of clearance amount. The method of calculating the area of the molten pool and the orifice is: after the original molten pool image is obtained by high-speed photography, the molten pool image is preprocessed to improve the signal-to-noise ratio of the molten pool image; and then the image is binarized , get the number of pixels in the optical path direction and the entire melt pool and orifice, and obtain the area of the melt pool and orifice.

Figure 3 is the dynamic image of the molten pool under different gaps of the laser lap welded T-joint monitored by the above-mentioned visual sensing system, and Figure 4 is the real-time monitoring effect of the complete image of the molten pool of the laser welded T-shaped lap joint under different gaps Example diagram, Fig. 5 Schematic diagram of the method for measuring the length and width of the molten pool. From Figure 3, Figure 4 and Figure 5, it can be seen that the molten pool and the orifice (the white area at the front of the molten pool) at different times have fluctuations and certain periodicity. When there is no gap (the gap is 0mm), the molten metal on the surface of the molten pool is in a wave shape, and the orifice fluctuates with it, and there is no sinking phenomenon; when there is a gap, and with the increase of the gap, the molten pool gradually becomes smaller and lower. The collapse became more and more serious, the light and dark of the orifice alternated, and gradually tilted.

Figure 6 is a photo of the T-shaped lap joint with no gap and with a gap. From the appearance of the T-shaped lap joint in Figure 6, it can be seen that the gap has a great influence on the weld shape, and the collapse is serious when the gap is large.

The image processing platform technology based on Labview is used for image processing. The schematic diagram of the melt pool and orifice processing flow is shown in Figure 7, and the processing example is shown in Figure 8. The image processing platform based on Labview is used to extract the edge and area of the molten pool and the orifice from the captured molten pool image, specifically including: after the original image is obtained by high-speed photography, the image is preprocessed to improve the signal-to-noise ratio, and then the image is processed Perform binarization processing to obtain the number of pixels in the direction of the optical path and the entire molten pool or orifice, and calculate the area. Figure 9 is a comparison of the images of the molten pool and the orifice before and after extraction. It can be seen that the edge of the orifice and the molten pool can be better extracted through image processing, and the external interference factors are basically filtered out. The obtained image can reflect the molten pool and the to the greatest extent. The viewing area of the orifice.

Figure 10 shows the variation of the length and width of the molten pool with the amount of gap. It can be seen that with the increase of the gap, the length of the molten pool gradually decreases, while the width does not change much. When the gaps are 0mm, 0.2mm, 0.4mm, 0.6mm, and 0.8mm, the corresponding molten pool lengths are 12.4mm, 9.8mm, 9.6mm, 8.5mm, 8mm, and 7.5mm, respectively.

Figure 11 is the change curve of molten pool area under different gaps. It can be seen that with the increase of the gap, the area of the molten pool gradually decreases, and when the gap is greater than 0.2mm, the area of the molten pool decreases suddenly. When the gap is 0mm, 0.2mm, 0.4mm, 0.6mm, 0.8mm, the corresponding melt pool area is 50525, 43439, 29928, 25872, 23465, 22642 (Pixel).

Figure 12 shows the correspondence between the average area of the orifice and the amount of gap. It can be seen that with the increase of the gap, the average area of the orifice gradually decreases, and basically changes linearly. When the gaps are 0mm, 0.2mm, 0.4mm, 0.6mm, 0.8mm, and 1.0mm, the average area sizes of the corresponding orifices are 2859, 2521, 2053, 1962, 1824, and 1334 (unit pixel).

Example 2 The actual detection of the gap amount of the laser welded T-shaped lap joint

Adopt the detection system and method of embodiment 1, and utilize the relation between the laser welded T-type lap joint gap and molten pool parameter that it obtains, through to the real-time observation of molten pool image in welding process and orifice area and molten pool For the calculation of width, length and area, gap testing is performed on the following two sets of laser welded T-shaped lap joints. The testing results are shown in Figure 13.

The monitoring results of the first group a are as follows: the length of the molten pool is 9.6mm, the area of the molten pool is 44323 Pixel, and the area of the orifice is 2665. Through the comprehensive analysis of the above data and the dynamic characteristics of the molten pool and the orifice, it can be known that the gap in the welding process is about 0.2mm. It can be seen from the cross-section of the welded joint corresponding to Figure 13 that the gap is 0.21 mm, which is 0.01 mm different from the measured gap.

The monitoring results of the second group b are as follows: the length of the molten pool is 7.8mm, the area of the molten pool is 22053 Pixel, and the orifice area is 2063. Through the comprehensive analysis of the above data and the dynamic characteristics of the molten pool and the orifice, it can be known that the gap in the welding process is about 0.26mm. It can be seen from the cross-section of the welded joint corresponding to Figure 13 that the gap is 0.23 mm, which is 0.03 mm different from the measured gap.

It can be seen that by using the above detection method and detection system, a clear dynamic image of the molten pool and the orifice during the welding process can be obtained. By observing and calculating the area of the molten pool and the orifice and their fluctuations, the generation of the gap in the welding process can be quickly and accurately judged. And its size provides a criterion for the analysis and detection of subsequent welding quality.

The preferred embodiments of the invention disclosed above are only to help illustrate the invention. The preferred embodiments are not exhaustive in all detail, nor are the inventions limited to specific embodiments described. Obviously, many modifications and variations can be made based on the contents of this specification. This description selects and specifically describes these embodiments in order to better explain the principle and practical application of the present invention, so that those skilled in the art can well understand and utilize the present invention. The invention is to be limited only by the claims, along with their full scope and equivalents.

Claims (4)

1. a detection method based on the laser welding T-type lap joint gap of molten pool image visual sensing, is characterized in that, comprises the steps: (1) Place the workpiece (3) on the workbench and move with the workbench, while the laser head (1) is fixed; a CMOS camera (5) with LinLog photosensitive technology, filter system (11), auxiliary Light source (9), image acquisition card (6), computer (7), visual sensing system of display (8), welding image signal is sent to computer (7) through image acquisition card (6), and displayed on display (8) ;Place the camera (5) on the side of the table at 500-1000mm and form an angle of 75-85 ^° with the surface of the workpiece (3), adjust the auxiliary light source (9) so that it is 300-500mm in front of the laser head (1) and Form an included angle of 25-35 ^° with the surface of the workpiece (3); adjust the image size and focal length of the CMOS camera (5) with LinLog photosensitive technology; (2) Adjust the CMOS camera (5) with LinLog photosensitive technology so that it can collect the image of the entire molten pool on the surface of the workpiece (3). When the image of the molten pool is not centered or the complete molten pool image cannot be collected, adjust the LinLog photosensitive technology The position of the CMOS camera (5); according to the welding speed and the length of the weld, properly adjust the exposure and control the acquisition time and frame number, and record the acquisition frequency; (3) Observe the dynamic changes of the molten pool during the laser welding process, and analyze the changes in the size, shape, brightness and slump defects of the molten pool and orifice in real time; (4) Import the collected melt pool image into another computer equipped with the Labview image processing platform, carry out edge extraction and area calculation for the orifice, and analyze the corresponding relationship between the orifice area and the gap; use Photoshop software to analyze the melt pool Measure the width and length of the gap, and analyze its corresponding relationship with the gap; (5) According to the corresponding relationship between the above-mentioned orifice area and the gap amount and the corresponding relationship between the width and length of the molten pool and the gap amount, through the real-time observation of the molten pool image during the welding process and the orifice area and the molten pool width, length and The calculation of the area monitors the gap amount during the welding process in real time. 2. The detection method for the gap of the laser welded T-shaped lap joint based on the visual sensing of the molten pool image according to claim 1, characterized in that, in the step (4), the acquired molten pool image is When processing, the median method is used to smooth filter first, then the gray level analysis and threshold segmentation are performed, and finally the Log edge operator combining Gaussian filter and Laplacian edge detection algorithm is used to smooth and integrate the thresholded image. Integral processing filters out interference information and extracts edge lines. 3. The method for detecting gaps of laser welded T-shaped lap joints based on molten pool image visual sensing according to claim 1, characterized in that, in the steps (4) and (5), the calculated molten pool and The method of the area of the orifice is: after obtaining the original molten pool image through high-speed photography, preprocessing the molten pool image to improve the signal-to-noise ratio of the molten pool image; and then binarizing the image to obtain the optical path direction Sum the number of pixels of the entire melt pool and orifice to obtain the area of the melt pool and orifice. 4. The method for detecting gaps in laser welded T-shaped lap joints based on molten pool image visual sensing according to claim 1, characterized in that the zoom range of the CMOS camera (5) with LinLog photosensitive technology is 18- 45mm, aperture value 5.6-32, exposure value 4-10ms, frame rate 800fps; the filter system (11) is coaxially connected with the camera (5); the filter system (11) includes a A filter (13) with a wavelength of 808nm, two neutral light-reducing filters (12) with a transmittance of 25% and a protective glass (14); the auxiliary light source (9) used is a power of 2W and a wavelength of 808 nm semiconductor laser.