A laser vision system optimal imaging bandwidth and welding noise suppression method

Abstract

The present application belongs to the field of welding intelligent perception and machine vision technology, and discloses a kind of laser vision system optimal imaging bandwidth and welding noise suppression method, solve the existing laser vision system is disturbed by welding arc light, smoke, spatter, and the problem of poor imaging quality, low weld detection precision.The present application measures the spectral distribution of welding arc under multiple working conditions, selects the low-intensity spectral candidate bandwidth, and determines 660.0±10.0nm as the optimal imaging bandwidth through experiments;Based on the bandwidth matching narrowband filter device, laser light source and optical structure, a noise suppression system is built.The present application can effectively suppress welding visual noise, reduce the average image noise level to below 0.0332, significantly improve the accuracy and real-time performance of weld detection, and has better performance than similar products, providing technical support for robot welding intelligence, and has high engineering application value.

Description

Technical Field

[0001] This invention belongs to the field of laser vision sensing technology for robotic welding, specifically relating to a method for determining the optimal imaging bandwidth of a laser vision system and a welding noise suppression method based on the optimal bandwidth. It is applicable to various gas-shielded welding processes such as MAG, MIG, and TIG, and can effectively solve the problem of interference from arc light, smoke, and spatter on the imaging of the laser vision system during the welding process, thereby improving the accuracy of weld detection and tracking in robotic welding. Background Technology

[0002] Welding is an indispensable basic manufacturing method in the manufacturing industry. About 70% of steel products worldwide are welded products, and currently more than 90% of welding work is still done manually. Manual welding has problems such as high cost and harsh working environment. The strong heat radiation, fumes and arc light in the welding process can cause damage to the health of workers. Therefore, robotic welding has become an important development direction in the welding field.

[0003] Visual sensing technology is a core technology in robotic welding. With its non-contact, high-precision, fast detection, and strong adaptability, it has become the mainstream sensing method in robotic welding. It is divided into passive vision systems (PVS) and active vision systems (laser vision systems, LVS). Passive vision systems use the welding arc light as a light source and cannot acquire depth information, making them only suitable for inspecting planar welds. Laser vision systems, on the other hand, use laser light as an active light source and can acquire weld depth information, adapting to the inspection of complex curved surface welds. This makes it the most promising sensing method for robotic welding.

[0004] However, laser vision systems are subject to strong visual noise interference during welding, mainly including arc light, smoke, and spatter. Arc light and smoke can obscure laser stripes and reduce image contrast, while the brightness of spatter is similar to that of laser stripes, which can disrupt the shape of the laser stripes and lead to a significant decrease in image quality. In existing technologies, software algorithms are typically used for noise reduction, but due to the uncertainty of welding noise, the performance of these algorithms is still not ideal.

[0005] Hardware improvements are another approach to eliminating welding noise interference. Existing research based on arc spectral analysis attempts to use spectral filtering methods to eliminate noise. However, these studies primarily target passive vision systems and are not applicable to active vision systems. Due to the monochromatic nature of lasers, the multi-bandwidth filtering methods used in passive vision systems are unsuitable for active vision systems. Furthermore, the selection of imaging bandwidth in existing laser vision systems does not consider the diversity of welding conditions. Different welding currents, welding materials, and welding methods can cause changes in the arc spectral distribution, making single-bandwidth filters unsuitable for various welding conditions. Moreover, there is a lack of objective quantitative evaluation indicators for visual noise, and the judgment of image quality relies on subjective human judgment, making it difficult to scientifically select the filtering bandwidth. Therefore, there is an urgent need for a method for determining the optimal imaging bandwidth of a laser vision system based on arc spectral analysis, and to design a corresponding noise suppression system based on this bandwidth to solve the noise interference problem of laser vision systems in robotic welding. Summary of the Invention

[0006] To address the aforementioned deficiencies in existing technologies, this invention provides an optimal imaging bandwidth and welding noise suppression method for laser vision systems. The aim is to solve the problem of poor imaging quality caused by interference from arc light, smoke, and spatter during the welding process. By determining the optimal imaging bandwidth suitable for multiple welding conditions and simultaneously constructing a corresponding noise suppression system, efficient and stable suppression of welding noise is achieved, thereby improving the imaging quality and weld detection accuracy of the laser vision system.

[0007] The present invention is achieved through the following technical solution.

[0008] A method for determining the optimal imaging bandwidth of a laser vision system includes the following steps: S1. Build an arc spectral measurement system to collect spectral distribution data of the welding arc under different welding conditions; the welding conditions include different welding currents, different welding materials (including carbon steel / stainless steel / aluminum alloy welding materials), and different gas shielded welding methods, etc. S2. Five candidate imaging bands with low arc spectral intensity were identified from the spectral distribution data collected from a wide range of experiments. These bands are: ultraviolet rising band, infrared attenuation band, 335.0±3.0nm, 451.0±7.0nm, and 660.0±10.0nm. S3. Combining the quantum efficiency characteristics of industrial cameras, candidate bands with wavelengths below 350nm are excluded, and the effective candidate imaging bands are the infrared attenuation band, 451.0±7.0nm, and 660.0±10.0nm. S4. Establish a quantitative evaluation index for visual noise of the laser vision system. Use this index to verify the imaging performance of each effective candidate imaging band. Determine 660.0±10.0 nm as the optimal imaging bandwidth of the laser vision system. The laser vision system working under a bandwidth of 660.0±10.0 nm can effectively suppress welding noise.

[0009] Furthermore, the arc spectral measurement system described in step S1 includes a welding power source, a welding torch, a welding workpiece, a spectrometer, an optical probe, an optical fiber, and a laser vision system. The spectrometer is an Ocean Optics HR2000+CG-UV-NIR with a spectral acquisition range of 186.7~1100.8nm, a resolution of 0.473nm, a minimum integration time of 1ms, and the measurement distance between the optical probe and the welding workpiece is set to 180mm.

[0010] Furthermore, the different welding currents mentioned in step S1 are gradient currents of 120~300A, the different welding materials include Q195 carbon steel, Q255 carbon steel, Q275 carbon steel, and aluminum alloy, and the different gas shielded welding methods include MAG welding and MIG welding.

[0011] Furthermore, the method for establishing the visual noise quantification evaluation index in step S4 is as follows: define the foreground grayscale ratio. , grayscale k pixel ratio, grayscale k The number of pixels, n Total number of pixels; defines the image noise level. ,in For a clean image foreground grayscale ratio free from arc interference, The foreground grayscale ratio of a noisy image.

[0012] Furthermore, a welding noise suppression method for a laser vision system based on the aforementioned optimal imaging bandwidth, which matches the laser source, narrowband filter devices, and optical structure with an optimal imaging bandwidth of 660.0±10.0nm, is used to build a laser vision noise suppression system to achieve noise suppression of welding arc light, smoke, and spatter. The method specifically includes the following steps: T1. A narrowband filter with a center wavelength of 660.0nm, a bandwidth of 20.0nm, and a cutoff depth of OD4 is selected as the core filtering device. T2. A semiconductor laser with a wavelength of 660.0nm is selected as the laser source, and a Powell prism is used to shape the laser spot into a linear laser stripe with uniform brightness. T3. Install a neutral density filter with a transmittance of 30% at the front end of the narrowband filter to avoid the impact of high-intensity electric arc light on the performance of the filter. T4. Select an industrial camera, and then integrate the laser light source, narrowband filter, neutral density filter and industrial camera to ensure that the laser optical axis and the camera optical axis are tilted at a pre-set angle, and build a laser vision noise suppression system. T5. The laser vision noise suppression system is mounted on the end effector of the welding robot to achieve real-time imaging and weld detection during the welding process.

[0013] The noise suppression method proposed in this application enables the sensor to capture clear images during the welding process. Through hand-eye calibration between the vision sensor and the robot tool, the sensor can guide the robot to achieve real-time tracking. Since the optimal imaging bandwidth proposed in this patent is a common law discovered under different welding conditions, its application in the laser vision system can also adapt to different welding processes, which ensures that the noise suppression method proposed in this application has stability and universality.

[0014] Furthermore, the power of the semiconductor laser in step T2 is 200mW.

[0015] Furthermore, the industrial camera mentioned in step T4 is a monochrome industrial camera, which has no significant attenuation in quantum efficiency in the 660.0nm band and has adjustable exposure function.

[0016] Furthermore, the laser vision noise suppression system is adapted to welding conditions including welding current of 100~300A, MAG / MIG / TIG welding methods, and carbon steel / stainless steel / aluminum alloy welding materials, with average image noise levels under each condition. ≤0.0332.

[0017] To further explain, the laser vision system obtained in step T4 will be tested under different welding conditions to verify its noise suppression performance. The process involves using a welding robot equipped with sensors to capture a sequence of images under a specific welding condition. After storing the image data, the noise level δ of each image is calculated using the method for establishing the visual noise quantification evaluation index. Then, the average noise level in the image sequence is calculated. After evaluation and testing, the average image noise level under various operating conditions was determined. ≤0.0332.

[0018] To further explain, the laser vision noise suppression system described in step T5, when mounted on the end effector of the welding robot, can realize 3D point cloud reconstruction of the weld pool and real-time autonomous tracking of weld seams such as V-grooves. The laser stripes are not significantly obstructed, and the interference of spatter on imaging is a minor and recoverable disturbance.

[0019] The core innovation of this invention lies in: (1) Five candidate regions for low-intensity arc spectral bandwidth were determined by measuring the arc spectrum under different welding conditions; (2) Experiments show that the average value of the image noise level δ of the laser vision system operating at a bandwidth of 660.0±10.0 nm is 0.0332, and the maximum value is 0.03256, which is significantly lower than other bandwidths (such as 335.0±3.0 nm, 451.0±7.0 nm, and 940.0 nm). (3) Compared with existing commercial systems on the market (such as the META Vision system), the laser vision system of the present invention can provide clearer images during the welding process, which significantly improves the accuracy and real-time performance of weld inspection; (4) This invention provides technical support for the automation and intelligence of robotic welding, and can realize high-precision real-time detection of the welding process, thereby improving welding efficiency and quality.

[0020] The optimal imaging bandwidth of the laser vision system and the welding noise suppression method of the present invention have the following beneficial effects: 1. The optimal imaging bandwidth for adapting to multiple working conditions was determined: Through arc spectrum analysis under multiple welding conditions, combined with the quantum efficiency characteristics of industrial cameras and visual noise quantification index, 660.0±10.0nm was determined for the first time as the optimal imaging bandwidth of the laser vision system. This bandwidth maintains low arc spectrum intensity under different welding currents, welding materials and welding methods, and has strong adaptability. 2. Achieved highly efficient suppression of welding noise: The laser vision noise suppression system, built based on optimal imaging bandwidth, specifically addresses the interference problems caused by arc light, smoke, and spatter. Spatter is the main noise source in this band; compared to arc light and smoke, spatter causes minimal and recoverable interference with laser stripes. The system's average image noise level under various welding conditions is [not specified]. ≤0.0332, far lower than existing technologies; 3. It solves the core defects of existing hardware filtering: avoids the center wavelength drift and bandwidth expansion problems of infrared attenuation section filters, and overcomes the problem of short-wavelength laser being blocked by Mie scattering of welding smoke. The filter works stably and the laser stripe imaging is clear and complete. 4. An objective evaluation index for visual noise quantification was established: For the first time, the average and maximum values ​​of image contamination degree δ were proposed as evaluation indexes for the imaging quality of laser vision systems, realizing the objective determination of the degree of welding noise interference and providing a scientific basis for the selection of imaging bandwidth; 5. High engineering application value: The laser vision noise suppression system of this invention has a compact structure and can be directly mounted on the end effector of existing welding robots without large-scale modification of the robots. Moreover, the noise suppression effect of the system is significantly better than that of existing commercial laser vision systems. It can realize 3D point cloud reconstruction of the weld pool and real-time tracking of complex groove welds, thereby improving the automation and intelligence level of robot welding. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the arc spectrum measurement principle constructed for this invention; Figure 2 The arc light spectrum under different currents was measured by the arc measurement system built for this invention. Figure 3 The schematic diagram of the laser vision system designed for this invention; Figure 4 This is a schematic diagram illustrating the application of the laser vision system designed in this invention in robotic welding. Detailed Implementation

[0022] The invention is further described below with reference to the accompanying drawings.

[0023] Example 1: Method for Determining the Optimal Imaging Bandwidth of a Laser Vision System This invention establishes an arc spectral measurement system to collect arc spectral distribution data under different welding conditions, identifies candidate bands for low-intensity arc spectra, filters effective bands based on the quantum efficiency characteristics of industrial cameras, verifies the imaging performance of each band using established visual noise quantification evaluation indicators, and finally determines the optimal imaging bandwidth of the laser vision system. The specific steps are as follows: 1. Setup of the arc spectroscopy measurement system: such as Figure 1 As shown, the arc spectroscopy measurement system mainly consists of a welding power source, a welding torch, a workpiece, a spectrometer, and a laser vision system. The spectrometer is an Ocean Optics HR2000+CG-UV-NIR, with a spectral acquisition range of 186.7–1100.8 nm, a resolution of 0.473 nm, and a minimum integration time of 1 ms. The measurement distance d is set to 180 mm. 2. Multi-condition arc spectrum acquisition: Spectral distribution data of welding arcs are acquired under different welding conditions. The welding current is set to a gradient current of 120A, 180A, 240A, and 300A. The welding materials used are Q195 carbon steel, Q255 carbon steel, Q275 carbon steel, and aluminum alloy. The welding methods used are mainstream gas shielded welding methods such as MAG and MIG, covering common working conditions of industrial robot welding. 3. Low-intensity candidate band identification: as shown in the attached document. Figure 2 As shown, analysis of the collected arc spectral data under various operating conditions identified five candidate imaging bands where the arc spectral intensity remained consistently low: the ultraviolet rising band (200~320nm), the infrared attenuation band (800~1100nm), 335.0±3.0nm, 451.0±7.0nm, and 660.0±10.0nm. 4. Screening of effective candidate wavelength bands: Considering the quantum efficiency characteristics of industrial cameras, industrial cameras have a weak response to light signals with wavelengths below 350nm. Therefore, the ultraviolet rising band and the 335.0±3.0nm band are excluded, resulting in three effective candidate imaging bands: infrared attenuation band (800~1100nm), 451.0±7.0nm, and 660.0±10.0nm. 5. Establishment of Visual Noise Quantification Indicators: Establish objective visual noise quantification evaluation indicators to verify the imaging performance of each effective candidate band: Define grayscale percentage ,in k ∈[0,255] represents the image gray level. grayscale k Pixel count, n This represents the total number of pixels in the image. Define foreground grayscale ratio This represents the percentage of pixels in an image that are not against a black background. Define image noise level ,in For a clean image foreground grayscale ratio free from arc interference, The foreground grayscale ratio of a noisy image; In the image sequence δ average and maximum value δ max As a core evaluation indicator, the smaller the indicator value, the less the image is affected by welding noise and the better the imaging performance. 6. Optimal Imaging Bandwidth Verification and Determination: Based on the above visual noise quantification indicators, imaging experiments of the laser vision system were conducted on the three effective candidate bands to verify the noise suppression effect and imaging performance of each band. Infrared attenuation section: The filter exhibits center wavelength drift and bandwidth expansion issues, with an average drift amplitude of 213.0nm and a bandwidth doubling to 42.2nm. Laser stripes are easily and completely blocked by arc light, resulting in high values. 451.0±7.0nm band: Short-wavelength lasers are easily affected by Mie scattering from welding fumes, which severely obscure the laser stripes and result in poor imaging quality. 660.0±10.0nm band: The main noise in this band is spatter. Compared with arc light and smoke, spatter has the least interference with laser stripes, and the filter does not show significant wavelength drift or bandwidth change. and δ max All values ​​are the minimum values ​​in the three bands, resulting in clear imaging and complete laser stripes; Ultimately, 660.0 ± 10.0 nm was determined to be the optimal imaging bandwidth for the laser vision system.

[0024] Example 2: Welding noise suppression method based on optimal imaging bandwidth Based on the optimal imaging bandwidth of 660.0±10.0nm determined above, this invention constructs a laser vision noise suppression system by matching the laser source, narrowband filter devices, and optical structure to achieve efficient suppression of welding noise. The specific steps are as follows: 1. Narrowband filter selection: Select a multilayer film structure narrowband filter based on Fabry-Perot theory, with core parameters matching the optimal imaging bandwidth: center wavelength λ0=660.0nm, transmission bandwidth Δλ0=20.0nm, cutoff depth OD=4, to ensure high transmittance for 660.0nm laser and high blocking rate for arc light of other wavelengths.

[0025] 2. Laser source selection and shaping: A semiconductor laser with a wavelength of 660.0nm and a power of 200mW was selected as the active source. A Powell prism was used instead of a traditional cylindrical prism to shape the circular laser spot into a linear laser stripe with uniform brightness, thereby improving the imaging quality of the laser stripe and the accuracy of weld inspection.

[0026] 3. Optical protection structure design: A neutral density filter with a transmittance of 30% is installed at the front end of the narrowband filter to avoid irreversible damage to the performance of the filter caused by the high-intensity electric arc light during welding, and to ensure the long-term stable operation of the filter.

[0027] 4. Laser vision system integration and assembly: The laser light source, Powell prism, neutral density filter, narrowband filter, and monochrome industrial camera are integrated and assembled to ensure that the laser optical axis is coaxial with the camera optical axis and that the laser stripe can be fully entered into the camera's field of view. The laser vision noise suppression system is built. The system has a compact structure and can be directly mounted on the end effector of the welding robot.

[0028] 5. On-site application of robotic welding: as shown in the attached document. Figure 4 As shown, a laser vision noise suppression system is mounted on the end effector of a welding robot to complete the linkage debugging between the system and the robot, realizing real-time imaging, weld feature extraction, 3D measurement of the molten pool and autonomous weld tracking during the welding process. It is compatible with welding methods such as MAG, MIG, and TIG, as well as welding materials such as carbon steel, stainless steel, and aluminum alloy.

[0029] Experimental Example 1 like Figure 3 As shown, the laser vision system of this invention is installed at the end effector of a welding robot to achieve real-time detection and tracking of the weld seam. During the welding process, the laser vision system continuously captures images for 3D reconstruction of the weld seam and robot motion control. Experiments show that the system can effectively guide the robot to complete automated welding, providing clear weld seam images even under high-current welding conditions.

[0030] Experimental Example 2 The laser vision noise suppression system of this invention was applied to different welding currents, welding materials, and welding methods. 400 frames of images were acquired under each condition, and the average image noise level was calculated. The results are as follows:

[0031] Experimental results show that the noise suppression system of the present invention achieves an average image noise level under various welding conditions. All values ​​are ≤0.0332, indicating stable noise suppression performance and strong adaptability.

Claims

1. A method for determining the optimal imaging bandwidth of a laser vision system, characterized in that, Includes the following steps: S1. Build an arc spectral measurement system to collect spectral distribution data of welding arc under different welding conditions, including different welding currents, different welding materials, and different gas shielded welding methods; S2. Five candidate imaging bands with low arc spectral intensity were identified from the collected spectral distribution data, namely the ultraviolet rising band, the infrared attenuation band, 335.0±3.0nm, 451.0±7.0nm, and 660.0±10.0nm. S3. Combining the quantum efficiency characteristics of industrial cameras, candidate bands with wavelengths below 350nm are excluded, and the effective candidate imaging bands are the infrared attenuation band, 451.0±7.0nm, and 660.0±10.0nm. S4. Establish a quantitative evaluation index for visual noise of the laser vision system. Use this index to verify the imaging performance of each effective candidate imaging band and determine 660.0±10.0nm as the optimal imaging bandwidth of the laser vision system.

2. A method for suppressing welding noise in a laser vision system based on the optimal imaging bandwidth described in claim 1, characterized in that, Based on the optimal imaging bandwidth of 660.0±10.0nm, a laser vision noise suppression system was built by matching the laser source, narrowband filter devices, and optical structure to achieve noise suppression of welding arc light, smoke, and spatter. The specific steps include: T1. A narrowband filter with a center wavelength of 660.0nm, a bandwidth of 20.0nm, and a cutoff depth of OD4 is selected as the core filtering device. T2. Select a semiconductor laser with a wavelength of 660.0nm as the laser source, and use a Powell prism to shape the laser spot into a linear laser stripe with uniform brightness. T3. Install a neutral density filter with a transmittance of 30% at the front end of the narrowband filter to avoid the impact of high-intensity electric arc light on the performance of the filter. T4. Select an industrial camera, and then integrate and assemble the laser light source, narrowband filter, neutral density filter, and industrial camera to ensure that the laser optical axis is coaxial with the camera optical axis, and build a laser vision noise suppression system. T5. The laser vision noise suppression system is mounted on the end effector of the welding robot to achieve real-time imaging and weld detection during the welding process.

3. The optimal imaging bandwidth determination method according to claim 1, characterized in that, The arc spectral measurement system described in step S1 includes a welding power source, a welding torch, a welding workpiece, a spectrometer, an optical probe, an optical fiber and a laser vision system. The spectrometer is Ocean Optics HR2000+CG-UV-NIR, with a spectral acquisition range of 186.7~1100.8nm, a resolution of 0.473nm, a minimum integration time of 1ms, and the measurement distance between the optical probe and the welding workpiece is set to 180mm.

4. The optimal imaging bandwidth determination method according to claim 1, characterized in that, The different welding currents mentioned in step S1 are gradient currents of 120~300A, the different welding materials include Q195 carbon steel, Q255 carbon steel, Q275 carbon steel, and aluminum alloy, and the different gas shielded welding methods include MAG welding and MIG welding.

5. The method for determining the optimal imaging bandwidth according to claim 1, characterized in that, The method for establishing the visual noise quantification evaluation index in step S4 is as follows: Define foreground grayscale ratio , grayscale k pixel ratio, grayscale k The number of pixels, n Total number of pixels; defines the image noise level. ,in For a clean image foreground grayscale ratio free from arc interference, The foreground grayscale ratio of a noisy image.

6. The welding noise suppression method according to claim 2, characterized in that, The narrowband filter described in step T1 employs a multilayer film structure based on the Fabry-Perot interferometer theory, with the core parameter being the center wavelength. λ 0=660.0nm, transmission bandwidth Δλ0=20.0nm, cutoff depth OD=4, peak transmittance T0 is adapted to the quantum efficiency of industrial cameras.

7. The welding noise suppression method according to claim 2, characterized in that, The power of the semiconductor laser mentioned in step T2 is 200mW.

8. The welding noise suppression method according to claim 2, characterized in that, The industrial camera mentioned in step T4 is a monochrome industrial camera, which has no significant attenuation in quantum efficiency in the 660.0nm band and has adjustable exposure function.

9. The welding noise suppression method according to claim 2, characterized in that, The laser vision noise suppression system is compatible with welding conditions including welding current of 100~300A, MAG / MIG / TIG welding methods, and carbon steel / stainless steel / aluminum alloy welding materials. The average image noise level under each condition is specified. ≤0.0332.

10. The welding noise suppression method according to claim 2, characterized in that, After the laser vision noise suppression system described in step T5 is mounted on the end effector of the welding robot, it can realize the 3D point cloud reconstruction of the weld pool and the real-time autonomous tracking of weld seams such as V-groove. The laser stripes are not significantly obstructed, and the interference of spatter on the imaging is a minor and recoverable interference.

Images