3D binocular line laser weld contour data acquisition method and system

Abstract

The invention relates to the technical field of 3D vision measurement, and discloses a 3D binocular line laser weld contour data acquisition method and system. Two core steps of obtaining surface reflection characteristic information of a to-be-scanned area and adjusting output intensity of laser rays projected by a line laser device at different positions of the to-be-scanned area according to the surface reflection characteristic information are introduced, so that the problem that when a high-reflection welding seam is processed by a traditional method, the laser rays cannot be projected by the line laser device is effectively solved; the problem of image overexposure or underexposure caused by dramatic change of illumination characteristics is solved.

Description

Technical Field

[0001] This invention relates to the field of 3D vision measurement technology, and in particular to a 3D binocular laser weld contour data acquisition method and system. Background Technology

[0002] In automated welding production lines, acquiring high-precision, real-time, and non-contact weld contour data is a crucial step in ensuring product quality and achieving intelligent manufacturing. Typically, an advanced 3D binocular line laser acquisition system is cleverly integrated into the end effector of a robotic arm, with the binocular camera and line laser securely fixed together via a carefully designed mechanical base. This system undergoes rigorous calibration during initial deployment to accurately acquire the three-dimensional contour data of the weld. This data is then used for automated quality assessment, such as detecting weld geometry, bevel shape, and the presence of defects. Under ideal operating conditions, the system can stably output high-quality point cloud data, fully meeting the requirements of production cycle time and quality control.

[0003] However, traditional data acquisition methods face significant challenges when dealing with special materials and surface treatment processes, such as highly reflective metal welds. For example, in the production of new stainless steel pressure vessels, this material undergoes meticulous grinding before welding to meet specific corrosion resistance requirements, resulting in a near-mirror finish and high reflectivity. When a 3D scanning system scans a weld with such a highly reflective surface, the laser line, when projected onto the smooth base material areas on both sides of the weld, experiences strong specular reflection. Most of the laser energy is reflected at a specific angle, rather than forming uniform diffuse reflection as on a rough surface. This specular reflection acts like a beam of intense light directly into the lenses of one or two cameras, causing large areas of white overexposure in the captured image. The original laser lines are completely submerged in glaring spots. Faced with such an image, the image processing unit cannot identify the location of the laser lines, resulting in significant loss of 3D data in these highly reflective areas, creating data gaps.

[0004] To address the overexposure issue, on-site technicians attempted to reduce the camera's exposure time, hoping to suppress the strong reflections. This did indeed reduce the specular reflection spot in the base material area, making the laser line outline barely visible. However, a new problem arose. Due to the high temperatures during welding, the weld area has a relatively rough surface, making it a diffuse reflection zone. The light energy reflected back to the camera from this area is inherently much weaker than the specular reflection from the base material. After reducing the overall exposure time, the already weak laser line signal in the weld area became even dimmer, almost invisible in the image. This resulted in the overexposure problem in the base material area being solved, but the crucial weld outline data being lost. This created a dilemma: a high exposure time allows the weld to be clearly seen, but the base material area becomes overexposed; a low exposure time allows the base material to be clearly seen, but the weld area becomes too dark. Using a single, fixed exposure parameter is completely unsuitable for scanning scenarios with drastic changes in lighting characteristics, such as from highly reflective base materials to diffusely reflective welds. Summary of the Invention

[0005] This invention provides a 3D binocular laser weld contour data acquisition method and system, aiming to solve the technical problem in the prior art that when facing highly reflective metal welds, the use of a single fixed exposure parameter cannot adapt to the scanning scene with drastic changes in illumination characteristics from highly reflective base material to diffuse reflection welds, resulting in overexposed or underexposed images, and thus loss of three-dimensional data.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a 3D binocular laser weld contour data acquisition method, comprising: Obtain surface reflection characteristics information of the area to be scanned; Based on the surface reflection characteristics information, adjust the output intensity of the laser line projected by the line laser at different positions in the area to be scanned; With fixed single exposure parameters, images of laser lines with adjustable intensity are captured simultaneously.

[0007] Preferably, adjusting the output intensity of the laser line projected by the line laser at different positions in the area to be scanned based on the surface reflection characteristic information includes: A probe laser line and a measurement laser line are projected, wherein the probe laser line is located in front of the measurement laser line in the scanning direction; The intensity data of reflected light from the probe laser line on the surface of the area to be scanned is captured in real time; the reflected intensity data is converted into an intensity modulation command required by the measuring laser line at the same spatial position; Before the measuring laser line illuminates the surface area scanned by the probe laser line, the output intensity of the measuring laser line is adjusted; By adjusting the deflection angle of the micromirror, the intensity gradient modulation of the measurement laser line can be achieved in space.

[0008] Preferably, the real-time capture of the reflected light intensity data of the probe laser line on the surface of the area to be scanned includes: Multiple frames of probe laser line reflection images are continuously captured using an ultra-high-speed photosensitive array; Identify high-intensity reflection regions and low-intensity diffuse reflection regions in the multi-frame probe laser line reflection images; Based on the high-intensity reflection region, adjust the gain or exposure time of the high-intensity reflection region corresponding to the ultra-high-speed photosensitive array; Based on the low-intensity diffuse reflection region, adjust the gain or exposure time of the ultra-high-speed photosensitive array corresponding to the low-intensity diffuse reflection region. By fusing the multi-frame images of the probe laser line reflection, the intensity distribution of the reflected light from the probe laser line on the surface of the area to be scanned is obtained.

[0009] Preferably, the step of fusing the multi-frame probe laser line reflection images to obtain the reflected light intensity distribution of the probe laser line on the surface of the area to be scanned includes: Feature point extraction and feature point matching are performed on the multi-frame detection laser line reflection images; Based on the feature point matching results, the spatial transformation parameters between the multi-frame probe laser line reflection images are calculated; Based on the spatial transformation parameters, the multi-frame probe laser line reflection images are spatially aligned; By fusing the spatially aligned multi-frame probe laser line reflection images, the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned is obtained.

[0010] Preferably, adjusting the gain or exposure time of the high-intensity reflective region corresponding to the high-intensity reflective region of the ultra-high-speed photosensitive array includes: Identify extremely high-intensity reflection points in an image whose local brightness far exceeds a preset threshold; For the pixel units corresponding to the extremely high-intensity reflection points in the ultra-high-speed photosensitive array, perform ultra-short pulse exposure; For the low-intensity diffuse reflection region surrounding the extremely high-intensity reflection point, capture it while maintaining normal gain or exposure time; By fusing images captured under different exposure conditions, the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned is obtained.

[0011] Preferably, performing ultrashort pulse exposure on the pixel unit corresponding to the extremely high-intensity reflection point in the ultra-high-speed photosensitive array includes: The pixel units of the ultra-high-speed photosensitive array are configured such that each pixel unit is independently configured with an exposure control circuit and a gain adjustment circuit; The output signal strength of each pixel unit is monitored in real time using the ultra-high-speed photosensitive array. The ultra-high-speed photosensitive array identifies pixel units whose output signal strength exceeds a preset threshold. The control system of the ultra-high speed photosensitive array predicts the spatial position of the extreme high intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high intensity reflection point in the image. When capturing the next frame of the image, the exposure control circuit of the ultra-high-speed photosensitive array performs an ultra-short pulse exposure in advance on the pixel unit at the predicted position. The exposure pulse width of the corresponding pixel unit is dynamically adjusted by the exposure control circuit of the ultra-high-speed photosensitive array.

[0012] Preferably, the control system of the ultra-high-speed photosensitive array predicts the spatial position of the extreme high-intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high-intensity reflection point in the image, including: The machine scanning speed and the historical position sequence of the extreme high-intensity reflection point are obtained through the control system of the ultra-high speed photosensitive array. The control system identifies the geometric features and abrupt change areas on the weld surface. The control system segments the historical position sequence of the extreme high-intensity reflection point according to the geometric features and the abrupt change region, and performs nonlinear fitting in combination with the robot's current scanning speed to obtain the segmented motion trajectory. The control system extrapolates and predicts the spatial position of the extreme high-intensity reflection point in the next frame based on the segmented motion trajectory.

[0013] Preferably, the step of segmenting the historical position sequence of the extremely high-intensity reflection points and performing nonlinear fitting in conjunction with the robot's current scanning speed to obtain the segmented motion trajectory includes: The historical position sequence of each segment is preprocessed, including identifying and removing outliers and performing local smoothing on the remaining data. The segmented motion trajectory is obtained by nonlinearly fitting the historical position sequence after data preprocessing to the robot's current scanning speed.

[0014] Preferably, the data preprocessing for the historical location sequence of each segment includes: The distance distribution between data points within a local area and other data points in the neighboring area is statistically analyzed. Based on the distance distribution, clusters of data points that deviate significantly from other data points are identified as outlier clusters; Remove data points from the outlier clusters; Within a local region, the width and weight of the smoothing kernel function are dynamically adjusted based on the density of the data points and the intensity difference between adjacent data points to perform local smoothing.

[0015] Secondly, the present invention provides a 3D binocular laser weld contour data acquisition system, comprising: The detection end is used to acquire information about the surface reflection characteristics of the area to be scanned; The adjustment end is used to adjust the output intensity of the laser line projected by the line laser at different positions in the area to be scanned, based on the surface reflection characteristic information. The capture end is used to fix a single exposure parameter and simultaneously capture images of laser lines projected with adjustable intensity.

[0016] This application discloses a 3D binocular line laser weld contour data acquisition method and system. By acquiring the surface reflection characteristics of the area to be scanned, and dynamically adjusting the output intensity of the line laser projected at different positions within the scanned area based on this information, the system simultaneously captures images with a fixed exposure parameter. This method effectively solves the technical problem in existing technologies where, when facing highly reflective metal welds, a single fixed exposure parameter cannot adapt to the drastic changes in illumination characteristics, such as from highly reflective base material to diffuse reflection welds, leading to overexposure or underexposure of images and consequently, loss of 3D data. Through the technical solution of this application, the system can intelligently adjust the laser intensity according to the reflection characteristics of different areas. This allows the camera, with fixed exposure parameters, to avoid overexposure in highly reflective areas while ensuring clear visibility of laser lines in diffuse reflection areas. This ensures the acquisition of high-quality, data-free 3D weld contour data even under complex surface conditions, significantly improving the quality control and intelligent manufacturing level of automated welding production lines. Attached Figure Description

[0017] Figure 1 This is a flowchart of a 3D binocular laser weld contour data acquisition method provided in an embodiment of the present invention; Figure 2 This is a flowchart of another 3D binocular laser weld contour data acquisition method provided in this embodiment of the invention; Figure 3 This is a flowchart of another 3D binocular laser weld contour data acquisition method provided in this embodiment of the invention; Figure 4 This is a schematic diagram of a 3D binocular laser weld contour data acquisition system provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Reference Figure 1 The present invention provides a flowchart of a 3D binocular laser weld contour data acquisition method, which includes the following steps: S1, Obtain surface reflection characteristics information of the area to be scanned; S2, Based on the surface reflection characteristic information, adjust the output intensity of the laser line projected by the line laser at different positions in the area to be scanned; S3, with fixed single exposure parameters, simultaneously captures images of laser lines projected with adjustable intensity.

[0020] This application provides a 3D binocular line laser weld contour data acquisition method, aiming to solve the challenges encountered by traditional methods in processing highly reflective welds. The core of this method lies in dynamically adjusting the output intensity of the line laser to adapt to changes in the surface reflection characteristics of the area to be scanned, thereby enabling the simultaneous capture of high-quality laser line images even with fixed single exposure parameters.

[0021] Specifically, the "area to be scanned" typically refers to the workpiece surface where weld contour data needs to be acquired, which may include the weld itself and the base material areas on both sides of the weld. "Surface reflection characteristics information" refers to the reflection behavior of laser light at different locations within the area to be scanned, such as specular or diffuse reflection, and the magnitude of the reflection intensity. This information is crucial for subsequent adjustments to the laser output intensity. A "line laser" is a device capable of projecting a line-shaped laser beam, commonly used in 3D vision measurement to obtain contour information through the deformation of the laser line on the object's surface. "Single exposure parameters" refers to the fixed exposure time, gain, and other parameters used by the camera when capturing images, contrasting with the traditional method of manually or segmentally adjusting exposure parameters for different areas.

[0022] When implementing the method of this application, it is first necessary to obtain the surface reflectivity information of the area to be scanned. This step can be achieved in several ways. For example, a typical workpiece can be pre-scanned and analyzed offline to establish a database containing the reflectivity characteristics of different materials and surface treatment processes. Before the actual scan, the corresponding reflectivity information is retrieved from the database according to the workpiece type. Another approach is to pre-scan the area using a low-power probe laser line before the formal acquisition. By analyzing the reflected light intensity at different locations of the probe laser line, the surface reflectivity information of the area can be obtained in real time. For example, when the probe laser line illuminates a highly reflective area, the reflected light intensity will be significantly higher than that of a diffuse reflection area.

[0023] Next, based on the acquired surface reflection characteristics, the output intensity of the line laser projected at different locations within the scanned area is adjusted. This adjustment is one of the key innovations of this application. For example, if an area is identified as a highly reflective region, the output intensity of the line laser in that area can be appropriately reduced to avoid camera overexposure. Conversely, if an area is a diffuse reflection region, the output intensity of the line laser can be appropriately increased to ensure that the camera can capture a sufficiently bright laser line. This intensity adjustment can be achieved through various techniques. For instance, the line laser can integrate a programmable optical modulator, such as a digital micromirror device (DMD) or a liquid crystal spatial light modulator (LCOS), to achieve continuous intensity gradient modulation of the laser line in space by changing the deflection angle of its internal micromirrors or the transmittance of the liquid crystal cells. Another approach is to include multiple independent laser emitting units within the line laser, each with independently controllable output power, thereby creating segmented intensity adjustments in space.

[0024] Finally, under fixed single exposure parameters, images of the projected laser line with adjusted intensity are captured simultaneously. This means that the camera's exposure time, gain, and other parameters remain constant throughout the scanning process. Because the line laser has been pre-adjusted for the output intensity of different areas based on surface reflection characteristics, the camera can capture uniformly bright laser line images without overexposure or underexposure, even under single exposure parameters. For example, when the laser line illuminates a highly reflective area, the reflected light intensity is within the camera's acceptable range due to the reduced laser intensity, avoiding overexposure. When the laser line illuminates a diffuse reflection area, the reflected light intensity is sufficient for the camera to capture clearly due to the increased laser intensity, avoiding underexposure. Image capture is typically performed by a high-speed camera to ensure continuous weld contour data can be acquired in real time during robot movement.

[0025] The 3D binocular line laser weld contour data acquisition method of this application effectively solves the problem of image overexposure or underexposure caused by drastic changes in illumination characteristics when processing highly reflective welds by introducing two core steps: "acquiring surface reflection characteristic information of the area to be scanned" and "adjusting the output intensity of the line laser projected by the line laser at different positions in the area to be scanned according to the surface reflection characteristic information".

[0026] Specifically, in traditional methods, the camera uses a single, fixed exposure parameter. When the laser beam illuminates a highly reflective base material area, it produces strong specular reflection, causing localized overexposure of the camera image. The laser beam is then submerged in the light spot and cannot be identified. Conversely, when the exposure time is reduced to avoid overexposure, the diffuse reflection light in the weld area becomes too weak, resulting in the loss of weld contour data. This dilemma severely impacts the accuracy and completeness of data acquisition.

[0027] This application addresses this problem at its source by intelligently adjusting the intensity at the laser projection end. First, by acquiring information about the surface reflectivity of the area to be scanned, the system can "predict" the differences in reflectivity between different areas. For example, through pre-scanning or database queries, it identifies which areas are highly reflective and which are diffusely reflective. Subsequently, based on this information, the line laser can selectively adjust its output intensity at different locations. In highly reflective areas, the laser intensity is actively reduced to avoid excessive reflection that could overexpose the camera; in diffusely reflective areas, the laser intensity is appropriately increased to ensure that the reflected light is sufficiently captured by the camera.

[0028] Therefore, even with a single, fixed exposure parameter, the camera can simultaneously capture uniformly bright and clearly visible laser line images. This method avoids the drawbacks of traditional methods that require frequent adjustments to camera exposure parameters or complex image fusion, simplifying system operation and improving acquisition efficiency and data quality. In this way, this application ensures high-quality laser line images are obtained across the entire area from highly reflective substrate to diffuse-reflective weld seams, thereby achieving high-precision, hole-free 3D weld contour data acquisition. Compared with existing technologies, the core innovation of this application lies in transferring illumination adaptability from the camera end to the laser projection end. By actively controlling the laser intensity to adapt to surface reflection characteristics, robust acquisition of complex surface environments is achieved while maintaining the simplicity of a single camera exposure parameter.

[0029] In some embodiments described above, the output intensity of the line laser projected by the line laser at different positions in the area to be scanned is adjusted based on surface reflection characteristics. However, in actual weld contour data acquisition, the weld surface often exhibits complex geometry and variable material properties, resulting in significant spatial non-uniformity of reflection characteristics. If only a simple, unpredictable intensity adjustment strategy is used, it may be difficult to achieve accurate, continuous, and real-time modulation of the output intensity of the measurement laser line in scenarios involving high-speed scanning or rapid changes in reflection characteristics. This could lead to local overexposure or underexposure of the image, affecting the quality and accuracy of the final acquired data.

[0030] In this regard, refer to Figure 2 S2 includes: S21, Project a detection laser line and a measurement laser line, wherein the detection laser line is located in front of the measurement laser line in the scanning direction; S22, Real-time capture of the reflected light intensity data of the probe laser line on the surface of the area to be scanned; Convert the reflected intensity data into an intensity modulation command required by the measuring laser line at the same spatial position; S23, Before the measuring laser line illuminates the surface area scanned by the probe laser line, adjust the output intensity of the measuring laser line; S24, by adjusting the deflection angle of the micromirror, the intensity gradient modulation of the measurement laser line is achieved in space.

[0031] Specifically, this application introduces two types of laser lines: a probe laser line and a measurement laser line. The probe laser line is configured to be positioned in front of the measurement laser line in the scanning direction, and its main function is to scan the area to be scanned in advance to obtain information on the surface reflection characteristics in front. The measurement laser line follows immediately after, used for the actual acquisition of weld contour data. Through this front-to-back separation, the system can achieve advance perception of the reflection characteristics of the area in front.

[0032] Real-time capture of the intensity of reflected light from a probe laser line on the surface of the area to be scanned refers to the continuous monitoring and acquisition of the intensity of light reflected back from the surface by a photosensitive device. This reflected light intensity data reflects the differences in reflectivity at different locations within the area to be scanned.

[0033] The captured reflection intensity data is then converted into intensity modulation instructions required for the measurement laser line to be at the same spatial location. This conversion process typically involves a preset mapping relationship or algorithm that calculates, based on the reflection intensity of the probe laser line, the intensity value that the measurement laser line should output when it reaches the same location to obtain the best image effect (e.g., to avoid overexposure or underexposure).

[0034] Before the measuring laser line illuminates the surface area scanned by the probe laser line, the output intensity of the measuring laser line is adjusted according to the generated intensity modulation command. This means that the intensity adjustment of the measuring laser line is based on the reflection information of the area in front, rather than real-time feedback, thus achieving predictive adjustment.

[0035] This method achieves continuous intensity gradient modulation of the measurement laser line in space by adjusting the deflection angle of the micromirrors. A micromirror can be understood as a controllable optical element, such as a digital micromirror device (DMD) or a microelectromechanical system (MEMS) scanning mirror. By precisely controlling the deflection angle of the micromirrors, the energy distribution of the measurement laser line in space can be dynamically changed, thereby achieving continuous and fine-tuned adjustment of the laser line intensity at different locations, rather than simple on / off or segmented adjustments. This modulation method ensures that the laser line is projected with the most suitable intensity throughout the scanning path to accommodate minute changes in surface reflection characteristics.

[0036] This application's solution effectively solves the problem of achieving precise and continuous intensity modulation under complex surface reflection characteristics in traditional methods by introducing a separation mechanism between the probe laser line and the measurement laser line, combined with the precise modulation capability of the micromirror. Specifically, the probe laser line scans in front of the measurement laser line, enabling the acquisition of surface reflection characteristics information of the area to be scanned in advance. This pre-acquired reflected light intensity data is processed in real time and converted into intensity modulation commands required by the measurement laser line. Thus, before the measurement laser line actually illuminates a certain area, its output intensity has been pre-adjusted according to the predicted reflection characteristics of that area. This predictive adjustment mechanism avoids the lag problem that may exist in traditional real-time feedback adjustment. Furthermore, by adjusting the deflection angle of the micromirror, continuous spatial intensity gradient modulation of the measurement laser line can be achieved. This means that the intensity of the laser line can be smoothly and steplessly adjusted according to subtle changes in surface reflectivity, rather than discrete step-by-step adjustments. It is precisely because of this forward-looking information acquisition and refined intensity modulation method that this application's solution can ensure that the measurement laser line is always projected with optimal intensity under conditions of high-speed scanning or rapid changes in surface reflection characteristics, thereby guaranteeing the quality of subsequent image capture.

[0037] Through the above technical solution, this application can significantly improve the robustness and accuracy of 3D binocular laser weld contour data acquisition. By employing a probe laser line for forward scanning, the system can anticipate and respond to complex changes in the reflectivity of the weld surface, effectively avoiding localized overexposure or underexposure of the image due to drastic changes in reflectivity. Furthermore, by using a micromirror to perform spatially continuous intensity gradient modulation of the measurement laser line, the laser line intensity can be more precisely matched to the surface reflectivity, thus capturing laser line images with high signal-to-noise ratio and rich detail even with fixed single exposure parameters. This precise and dynamic intensity adjustment capability greatly improves the success rate of data acquisition and the accuracy of subsequent 3D reconstruction, making it particularly suitable for complex weld environments where high-reflectivity or low-reflectivity areas coexist.

[0038] In some preferred embodiments, a specific example is given below. Suppose it is necessary to acquire weld contour data of a workpiece welded from different metal materials, where one area is highly reflective polished stainless steel and the other is low-reflective anodized carbon steel. In conventional acquisition methods, if a single laser line of fixed intensity is used, the highly reflective areas may be severely overexposed, resulting in blurred or lost laser line contours; while the low-reflective areas may be underexposed, making it difficult to effectively identify the laser lines.

[0039] The solution proposed in this application effectively addresses this problem. Specifically, when the robot carrying the acquisition system moves along the weld direction, the probe laser line located in front of the measuring laser line first scans areas with different reflection characteristics. For example, when the probe laser line scans a polished stainless steel area, its reflected light intensity data will be significantly higher than that when scanning an oxidized carbon steel area. The system captures this intensity data in real time and, according to a preset algorithm, converts high-intensity reflection data into instructions to reduce the output intensity of the measuring laser line, and converts low-intensity reflection data into instructions to maintain or slightly increase the output intensity of the measuring laser line.

[0040] Subsequently, before the measuring laser line reaches the polished stainless steel area, the system, based on these instructions, pre-reduces the output intensity of the measuring laser line in that area by adjusting the deflection angle of the micromirrors. When the measuring laser line reaches the oxidized carbon steel area, its output intensity is adjusted to an appropriate level. Thus, even with fixed single exposure parameters, the capture end can simultaneously capture a laser line image with uniform brightness and clear contours throughout the entire weld area. This predictive and continuous intensity modulation ensures high-quality 3D weld contour data can be obtained even under complex and variable surface conditions.

[0041] In some of the above-described embodiments, traditional methods often face challenges in real-time capture of the reflected light intensity data of the probe laser line on the surface of the area to be scanned. Specifically, weld surfaces typically exhibit highly complex and non-uniform reflection characteristics, including both strong specular reflection and weak diffuse reflection. If a single, fixed gain or exposure parameter is used for image capture, it is highly likely that high-reflectivity areas will be severely overexposed, resulting in information loss, while low-reflectivity or diffuse reflection areas may be underexposed, making it difficult to extract effective intensity data. This limitation significantly affects the accuracy and reliability of the acquired surface reflection characteristic information, thereby restricting the precise adjustment of the subsequent measurement laser line output intensity.

[0042] In this regard, refer to Figure 3 S22 includes: S221 continuously captures multiple frames of probe laser line reflection images using an ultra-high-speed photosensitive array; S222, identify the high-intensity reflection region and the low-intensity diffuse reflection region in the multi-frame probe laser line reflection image; S223, adjust the gain or exposure time of the high-intensity reflection region corresponding to the high-intensity reflection region of the ultra-high-speed photosensitive array according to the high-intensity reflection region; S224, adjust the gain or exposure time of the low-intensity diffuse reflection region corresponding to the ultra-high speed photosensitive array according to the low-intensity diffuse reflection region; S225, fuse the multi-frame probe laser line reflection images to obtain the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned.

[0043] Specifically, the ultra-high-speed photosensitive array is configured to continuously capture images at an extremely high frame rate, thereby acquiring multiple reflection images of the same area within a very short time interval during the probe laser line scanning process. This continuous acquisition of multiple frames of probe laser line reflection images aims to provide a sufficient data foundation for subsequent region identification and parameter adjustment.

[0044] The identification of high-intensity reflection regions and low-intensity diffuse reflection regions in the multi-frame probe laser line reflection images can be understood as analyzing the image content to distinguish between regions with significantly higher brightness than the average level (usually corresponding to specular reflection or strong reflection) and regions with lower brightness and uniform light scattering (usually corresponding to diffuse reflection). This identification process can be achieved through image processing techniques such as setting brightness thresholds, gradient analysis, or local contrast analysis.

[0045] In practical applications, adjusting the gain or exposure time of the ultra-high-speed photosensitive array corresponding to the high-intensity reflection area, based on the high-intensity reflection area, refers to dynamically reducing the gain of the photosensitive array or shortening its exposure time in the high-brightness area identified in the image. The purpose is to avoid pixel saturation in the high-intensity reflection area, ensuring that the reflection intensity information in this area can be accurately recorded and preventing "overflow" phenomena.

[0046] Simultaneously, adjusting the gain or exposure time of the ultra-high-speed photosensitive array corresponding to the low-intensity diffuse reflection region, based on the low-intensity diffuse reflection region, refers to dynamically increasing the gain of the photosensitive array or extending its exposure time in the low-brightness region identified in the image. The purpose is to enhance the signal in the low-intensity diffuse reflection region, enabling it to be clearly captured, thereby obtaining more complete information on the surface reflection characteristics.

[0047] Therefore, fusing the multi-frame images of the probe laser line reflection to obtain the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned refers to synthesizing multiple frames of images captured after adjustments to different gains or exposure times. This fusion process aims to integrate effective information obtained from different regions under optimal exposure conditions to form an image that is neither overexposed nor underexposed and can comprehensively reflect the intensity distribution of reflected light from the surface of the area to be scanned.

[0048] This application's solution effectively solves the problem of accurately capturing reflected light intensity data under complex weld surface reflection characteristics by introducing an ultra-high-speed photosensitive array combined with a regional gain or exposure time adjustment strategy. First, the continuous capture capability of the ultra-high-speed photosensitive array ensures sufficient image data density during rapid laser line scanning. Second, by intelligently identifying high-intensity reflection areas and low-intensity diffuse reflection areas in the image, the system can dynamically adjust the photosensitive parameters for different areas. Specifically, for high-intensity reflection areas, reducing gain or shortening exposure time effectively avoids pixel saturation and preserves detail information in bright areas; while for low-intensity diffuse reflection areas, increasing gain or extending exposure time enhances signal strength, allowing details in dark areas to be clearly presented. Finally, by fusing these multiple frames captured under different optimized parameters, a reflected light intensity distribution map with good detail and accurate intensity across the entire dynamic range can be generated. This regional, dynamically adjusted strategy enables the system to adapt to the drastically changing reflection characteristics of the weld surface, thereby obtaining more accurate and comprehensive surface reflection characteristic information.

[0049] Through the above technical solution, this application can significantly improve the accuracy and robustness of surface reflection characteristic information acquisition during 3D binocular laser weld contour data acquisition. Compared with the traditional method using a single exposure parameter, this application effectively avoids local overexposure or underexposure of the image by adjusting the gain or exposure time differently for different reflection areas. This ensures that the reflected light intensity data can be accurately captured under complex and variable weld surface conditions, whether in areas of strong reflection or weak diffuse reflection. Therefore, the obtained surface reflection characteristic information is more comprehensive and realistic, providing a reliable basis for the subsequent precise modulation of laser line output intensity measurement, ultimately contributing to improving the accuracy and stability of weld contour data acquisition.

[0050] In some preferred embodiments, it is assumed that the weld surface to be scanned contains a portion of polished mirror-like areas and a portion of untreated rough diffuse reflection areas. As the probe laser line scans these areas, the ultra-high-speed photosensitive array continuously captures multiple frames of images. The system first analyzes these images to identify the extremely high-intensity reflection areas corresponding to the mirror-like areas and the low-intensity diffuse reflection areas corresponding to the rough areas. For the mirror-like areas, the photosensitive array's control system immediately adjusts the exposure time of the corresponding pixel unit, for example, by performing ultra-short pulse exposure to prevent image saturation. Simultaneously, for the rough diffuse reflection areas, the system maintains a normal gain or exposure time to ensure sufficient diffuse light signal is captured. Subsequently, these images captured under different exposure conditions are fused, for example, using HDR (High Dynamic Range) synthesis technology to integrate the details of the high-intensity areas and the low-intensity areas into a single image.

[0051] Ultimately, a complete reflected light intensity distribution map is obtained that clearly shows both the specular reflection intensity and the diffuse reflection intensity, thus providing precise input for subsequent measurement and adjustment of the laser line intensity.

[0052] Specifically, the above-mentioned fusion of multiple frames of probe laser line reflection images to obtain the reflected light intensity distribution of the probe laser line on the surface of the area to be scanned can be achieved according to the following steps.

[0053] The fusion process includes: Feature point extraction and feature point matching are performed on the multi-frame detection laser line reflection images; Based on the feature point matching results, the spatial transformation parameters between the multi-frame probe laser line reflection images are calculated; Based on the spatial transformation parameters, the multi-frame probe laser line reflection images are spatially aligned; By fusing the spatially aligned multi-frame probe laser line reflection images, the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned is obtained.

[0054] Feature point extraction refers to identifying pixels or regions with unique textures, corners, or edges from each frame of the detected laser line reflection image. These feature points should have high repeatability and robustness across different image frames. For example, algorithms such as SIFT (Scale Invariant Feature Transform), SURF (Speed-Up Robust Feature Transform), or ORB (Oriented Fast and Rotationally Shortened Feature Transform) can be used for feature point extraction. Feature point matching refers to finding correspondences between different image frames, i.e., determining which feature points represent the same physical location in different images. This is usually accomplished by calculating the similarity between feature descriptors (e.g., Euclidean distance or Hamming distance) and combining it with geometric constraints (such as the RANSAC algorithm) to eliminate incorrect matches.

[0055] Furthermore, based on the feature point matching results, spatial transformation parameters between the multi-frame probe laser line reflection images can be calculated. These parameters describe the geometric deformations existing between different image frames, such as translation, rotation, scaling, affine transformation, or perspective transformation. These transformation parameters, such as the homography matrix or fundamental matrix, can be solved using the least squares method or other optimization algorithms through the matched feature point pairs.

[0056] Based on this, the multi-frame probe laser line reflection images are spatially aligned according to the spatial transformation parameters. The purpose of spatial alignment is to transform all image frames to the same coordinate system, so that the pixel positions of the same physical point in different image frames can correspond precisely. This is usually achieved through image resampling (such as bilinear interpolation or bicubic interpolation), mapping the original image pixels to new positions in the target coordinate system.

[0057] Finally, the spatially aligned multi-frame probe laser line reflection images are fused to obtain the reflected light intensity distribution of the probe laser line on the surface of the area to be scanned. The fusion method may include pixel-level weighted averaging, median filtering, or image quality-based fusion strategies on the aligned images. For example, different weights can be assigned based on the brightness value or signal-to-noise ratio of each pixel to generate a comprehensive, high dynamic range reflected light intensity distribution image.

[0058] This application's solution effectively solves the geometric deformation and misalignment problems that may exist between images when continuously capturing multiple frames of probe laser line reflection images. This is achieved through refined feature point extraction, matching, spatial transformation parameter calculation, and spatial alignment. By precisely aligning images captured at different times to a unified spatial reference, it ensures that the reflected light intensity information from the same surface region can be accurately superimposed and integrated, thus avoiding blurring or distortion of intensity information caused by image misalignment. It is precisely because of this precise spatial alignment that the subsequent fusion step can effectively combine image information captured under different exposure or gain settings to generate a complete and accurate reflected light intensity distribution that includes both details of high-intensity reflection areas and information of low-intensity diffuse reflection areas.

[0059] The above technical solution significantly improves the accuracy and robustness of the intensity distribution of reflected light on the surface of the scanned area by the probe laser line. By precisely aligning and fusing multiple frames of images, image distortion and misalignment caused by motion or changes in viewing angle can be effectively eliminated, resulting in a more stable and detailed information on surface reflection characteristics. This high-precision distribution of reflected light intensity is crucial for subsequent accurate adjustment of the output intensity of the measurement laser line, ensuring superior weld contour data acquisition under complex surface conditions and avoiding overexposure or underexposure problems caused by inaccurate reflection intensity information.

[0060] In some embodiments described above, this application proposes adjusting the gain or exposure time of high-intensity reflection regions in the probe laser line reflection image corresponding to the ultra-high-speed photosensitive array to obtain a more accurate distribution of reflected light intensity. However, in actual weld contour data acquisition, the weld surface may have extremely high-intensity reflection points with local brightness far exceeding that of conventional high-intensity regions. These extreme reflection points may cause local saturation of the photosensitive element, resulting in image information loss or blurred details. If only a uniform gain or exposure time adjustment strategy is adopted, it is difficult to simultaneously ensure the imaging quality of both the extremely high-intensity reflection points and their surrounding low-intensity diffuse reflection regions. Therefore, this application further proposes a scheme for more refined adjustment of the gain or exposure time of the aforementioned high-intensity reflection regions to effectively address the challenges posed by extreme high-intensity reflections.

[0061] In this regard, this application further proposes the above-mentioned adjustment of the gain or exposure time of the high-intensity reflection region corresponding to the ultra-high-speed photosensitive array based on the high-intensity reflection region, including: Identify extremely high-intensity reflection points in an image whose local brightness far exceeds a preset threshold; For the pixel units corresponding to the extremely high-intensity reflection points in the ultra-high-speed photosensitive array, perform ultra-short pulse exposure; For the low-intensity diffuse reflection region surrounding the extremely high-intensity reflection point, capture it while maintaining normal gain or exposure time; By fusing images captured under different exposure conditions, the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned is obtained.

[0062] Specifically, identifying extremely high-intensity reflection points in an image where the local brightness far exceeds a preset threshold refers to using image processing algorithms to analyze the captured image of the probe laser line's reflection and detect pixels or pixel regions in the image whose brightness values ​​are significantly higher than those of typical high-intensity reflection areas. These extremely high-intensity reflection points are usually caused by the specular reflection characteristics of the laser line on the weld surface or specific geometric structures, and their brightness may far exceed the linear response range of the photosensitive array.

[0063] Specifically, for pixel units corresponding to extremely high-intensity reflection points in the ultra-high-speed photosensitive array, ultrashort pulse exposure is performed. This can be understood as precisely controlling the exposure time of the corresponding pixel units in the photosensitive array for these locally overbright areas, making it much shorter than the conventional exposure time. The purpose of ultrashort pulse exposure is to capture the effective light signal of extremely high-intensity reflection points in an extremely short time, avoiding pixel saturation and thus preserving their brightness information and details.

[0064] In practical applications, capturing the low-intensity diffuse reflection area surrounding an extremely high-intensity reflection point while maintaining a conventional gain or exposure time means that while the extremely high-intensity reflection point is exposed with an ultra-short pulse, the surrounding relatively low-brightness diffuse reflection area is captured using a gain or exposure time consistent with conventional imaging conditions. This ensures that these diffuse reflection areas receive sufficient exposure, thus clearly revealing their texture and contour information, and avoiding underexposure due to an overall exposure time that is too short.

[0065] Therefore, fusing images captured under different exposures to obtain the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned refers to image fusion processing of an ultrashort pulse exposure image captured for extremely high-intensity reflection points and a conventional exposure image captured for surrounding low-intensity diffuse reflection areas. The fusion process can employ multi-exposure fusion algorithms, such as weighted fusion and HDR (High Dynamic Range) synthesis, to integrate the effective information from different exposure images and generate a reflected light intensity distribution image with a wide dynamic range. This image can simultaneously and clearly display the details of extremely high-intensity reflection areas and the texture of low-intensity diffuse reflection areas, thus providing more comprehensive and accurate information on surface reflectivity characteristics.

[0066] Through the above technical solution, this application effectively solves the problem in 3D binocular line laser weld contour data acquisition that, due to the complex reflection characteristics of the weld surface, especially when there are extremely high-intensity reflection points, traditional single exposure or simple gain adjustment is insufficient to simultaneously capture highlights and shadow details. This solution significantly improves the dynamic range and signal-to-noise ratio of the image by using ultra-short pulse exposure on the extremely high-intensity reflection points while maintaining normal exposure on the surrounding areas, followed by image fusion. This avoids image saturation at extremely high-intensity reflection points, effectively preserving detail information in highlight areas while ensuring clarity in low-intensity diffuse reflection areas. This results in a more accurate and complete distribution of reflected light intensity on the surface of the area to be scanned by the acquired probe laser line, providing a more reliable basis for subsequent line laser output intensity adjustment, thereby improving the accuracy and robustness of the final weld contour data acquisition.

[0067] In some preferred embodiments, a specific example is given below. Suppose that during contour data acquisition of a stainless steel weld, due to the high specular reflectivity of the stainless steel surface, when the laser line projected by the line laser illuminates a specific angle or area of ​​the weld, it may produce locally extremely bright specular reflection points. These extremely high-intensity reflection points easily lead to pixel saturation of the photosensitive array under normal exposure, manifesting as large areas of white "dead zones" in the image, thus losing the geometric and reflection information of that area.

[0068] The proposed solution first continuously captures multiple frames of probe laser line reflection images using an ultra-high-speed photosensitive array and analyzes these images in real time. When an extremely high-intensity reflection point is identified in the image with local brightness far exceeding a preset threshold (e.g., exceeding 90% of the 255 grayscale value), the control system immediately performs an extremely short pulse exposure (e.g., the exposure time is shortened to 1 / 1000 of the conventional exposure time) on the pixel units corresponding to these extremely high-intensity reflection points in the ultra-high-speed photosensitive array. Simultaneously, for the relatively low-brightness diffuse reflection areas surrounding these extremely high-intensity reflection points, the photosensitive array maintains a conventional gain or exposure time for capture. Subsequently, the system fuses the captured ultra-short pulse exposure image (containing effective information of the highlight areas) with the conventional exposure image (containing effective information of the shadow and diffuse reflection areas). For example, a weighted average or gradient-based fusion algorithm can be used to combine the effective pixel values ​​of the two images to generate a high dynamic range reflected light intensity distribution image. In this way, even on complex weld surfaces with extremely high-intensity reflection points, it is possible to obtain unsaturated yet detailed reflected light intensity distribution data, providing precise input for subsequent laser intensity modulation.

[0069] In some embodiments described above, this application proposes to perform ultrashort pulse exposure on pixel units corresponding to extremely high-intensity reflection points in an ultra-high-speed photosensitive array to avoid local oversaturation of the image. However, in actual weld contour data acquisition, extremely high-intensity reflection points often move rapidly as the robot scans. If exposure adjustment is only performed after a high-intensity reflection point is detected, it may lead to response lag, causing the moving high-intensity area to still be oversaturated in the next frame, or failing to accurately align with the target pixel for effective exposure, thus affecting the accuracy and efficiency of data acquisition. Therefore, this application further proposes to optimize the execution method of the aforementioned ultrashort pulse exposure by introducing pixel-level independent control, real-time monitoring, motion prediction, and pre-exposure mechanisms to more accurately and promptly respond to dynamically changing extremely high-intensity reflection areas.

[0070] The step of performing ultrashort pulse exposure on the pixel units corresponding to the extremely high-intensity reflection points in the ultra-high-speed photosensitive array includes: The pixel units of the ultra-high-speed photosensitive array are configured such that each pixel unit is independently configured with an exposure control circuit and a gain adjustment circuit; The output signal strength of each pixel unit is monitored in real time using the ultra-high-speed photosensitive array. The ultra-high-speed photosensitive array identifies pixel units whose output signal strength exceeds a preset threshold. The control system of the ultra-high speed photosensitive array predicts the spatial position of the extreme high intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high intensity reflection point in the image. When capturing the next frame image, the exposure control circuit of the ultra-high-speed photosensitive array performs an ultra-short pulse exposure in advance on the pixel unit at the predicted position. The exposure pulse width of the corresponding pixel unit is dynamically adjusted by the exposure control circuit of the ultra-high-speed photosensitive array.

[0071] Specifically, to achieve precise and dynamic exposure control for extremely high-intensity reflection points, each pixel unit of the ultra-high-speed photosensitive array is designed with an independently configured exposure control circuit and gain adjustment circuit. This means that each pixel unit can independently adjust its own exposure time and gain according to the light intensity it receives, rather than the entire photosensitive array adjusting uniformly. The purpose of this independent configuration is to achieve fine-grained local exposure control, avoiding oversaturation in high-intensity areas from affecting other normal areas.

[0072] The ultra-high-speed photosensitive array is used to monitor the output signal intensity of each pixel unit in real time. This monitoring process is continuous and high-speed to ensure that minute changes in light intensity can be captured in a timely manner. When the output signal intensity of a certain pixel unit or a group of pixel units exceeds a preset threshold, the ultra-high-speed photosensitive array can identify these pixel unit groups, which typically correspond to extremely high-intensity reflection points in the image. The preset threshold is set to distinguish between normal reflection areas and extremely high-intensity reflection areas that may lead to oversaturation.

[0073] Furthermore, to address the dynamic movement of extremely high-intensity reflective points, the control system of the ultra-high-speed photosensitive array is configured to predict the spatial position of the extremely high-intensity reflective point in the next frame based on the robot's current scanning speed and the movement trend of the extremely high-intensity reflective point in the image. This prediction mechanism aims to anticipate the position of the high-intensity reflective point in the future, providing a basis for subsequent early exposure. The robot's current scanning speed can be provided by the robot's sensors or control system, while the movement trend of the extremely high-intensity reflective point in the image can be obtained by analyzing historical image data.

[0074] Therefore, when capturing the next frame of the image, the exposure control circuit of the ultra-high-speed photosensitive array can perform ultra-short pulse exposure on the pixel unit at the predicted position in advance. This "advance execution" strategy ensures that when the extremely high-intensity reflection point actually reaches the predicted position, the corresponding pixel unit is already in the ultra-short pulse exposure state, thereby effectively avoiding image oversaturation. The purpose of ultra-short pulse exposure is to complete the acquisition of light signals in an extremely short time to prevent pixel saturation caused by prolonged exposure to high-intensity light.

[0075] Furthermore, the exposure control circuit of the ultra-high-speed photosensitive array can dynamically adjust the exposure pulse width of the corresponding pixel unit. This means that the duration of the ultrashort pulse is not fixed, but can be adjusted in real time according to factors such as the actual reflected light intensity and the predicted moving speed to achieve the best exposure effect, avoiding oversaturation while ensuring that enough light signal is collected, thereby obtaining high-quality image data.

[0076] Through the above technical solution, this application overcomes the problems of local image oversaturation, data loss, or decreased acquisition accuracy caused by the dynamic movement of extremely high-intensity reflection points in existing technologies. Specifically, pixel-level independent exposure control and gain adjustment, combined with real-time monitoring and identification of high-intensity reflection points, enable the system to perform refined management of local areas. Crucially, by predicting the spatial position of the extremely high-intensity reflection point in the next frame and performing ultra-short pulse exposure at this predicted position, the system response time is significantly shortened, achieving precise and forward-looking exposure control of dynamically high-intensity areas. Furthermore, dynamically adjusting the exposure pulse width further optimizes the exposure effect, ensuring clear, unsaturated image data even in extremely high-intensity reflection areas. Therefore, this application significantly improves the adaptability, accuracy, and data quality of the 3D binocular laser weld contour data acquisition method under complex and dynamic conditions, providing more reliable raw data for subsequent weld identification and 3D reconstruction.

[0077] In some preferred embodiments, a specific example is given below. Suppose a 3D binocular laser weld contour data acquisition system is scanning a section of stainless steel weld. The weld surface has locally polished areas that produce extremely high-intensity specular reflections under laser illumination. As a robot moves along the weld at a preset speed, these high-intensity reflection points also move along with it in the image of the photosensitive array.

[0078] Specifically, each pixel unit of the ultra-high-speed photosensitive array is designed to integrate a miniature photodiode, an analog-to-digital converter, and independent exposure gating circuitry and gain amplifier. The photosensitive array captures images in real time as a laser line sweeps across the weld surface. The control system continuously monitors the output signal strength of each pixel unit. For example, when the output signal strength of a pixel unit group exceeds 80% of its saturation threshold for multiple consecutive frames, that pixel unit group is identified as an extremely high-intensity reflection point.

[0079] Simultaneously, the control system acquires the robot's current scanning speed, for example, 50 mm / s. By analyzing the position sequence of the extremely high-intensity reflection point in the past 10 frames of images, the control system uses algorithms such as Kalman filtering or polynomial fitting to predict the precise pixel coordinates of the reflection point in the next frame of the image. For example, it predicts that it will move from the current position (X,Y) to (X+ΔX, Y+ΔY).

[0080] Before capturing the next frame, the control system sends a command to the pixel unit at the predicted location (X+ΔX, Y+ΔY) via the exposure control circuit of the photosensitive array, instructing it to enter the ultra-short pulse exposure mode in advance, for example, setting the exposure pulse width to 1 / 100 of the conventional exposure time. When the laser line actually illuminates the predicted location, the corresponding pixel unit is already prepared for exposure with an ultra-short pulse, effectively avoiding oversaturation. Furthermore, if the predicted reflection intensity is particularly high, the exposure pulse width can be dynamically adjusted to be even shorter to ensure optimal image quality. In this way, even under high-speed scanning and complex reflection conditions, the system can stably acquire clear, unsaturated weld contour data.

[0081] In some embodiments described above, the control system of the ultra-high-speed photosensitive array predicts the spatial position of the extreme high-intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high-intensity reflection point in the image. However, in actual weld scanning, the geometry of the weld surface is often complex and variable, including bevels, weld toes, weld crowns, and various abrupt change areas. If only a simple linear model or a prediction method that does not consider these complex geometric features is used, the prediction of the trajectory of the extreme high-intensity reflection point may be inaccurate, especially when the robot's scanning speed changes or it passes through areas of geometric abrupt change. This prediction error may cause the ultrashort pulse exposure to fail to accurately target the high-intensity reflection area, thereby reducing the suppression effect of image oversaturation and affecting the quality of the final acquired data.

[0082] In response, this application further proposes a method for predicting the spatial position of an extremely high-intensity reflection point in the next frame using a control system of an ultra-high-speed photosensitive array, based on the robot's current scanning speed and the movement trend of the extremely high-intensity reflection point in the image. Specifically, this method includes: The robot's current scanning speed and the historical position sequence of extremely high-intensity reflection points are obtained through the control system of the ultra-high-speed photosensitive array. The control system identifies the geometric features and abrupt change areas on the weld surface. The control system segments the historical position sequence of extremely high-intensity reflection points based on geometric features and abrupt change regions, and performs nonlinear fitting in combination with the robot's current scanning speed to obtain the segmented motion trajectory. By controlling the system based on segmented motion trajectories, the spatial location of the next frame of the extreme high-intensity reflection point is extrapolated and predicted.

[0083] Specifically, acquiring the robot's current scanning speed and the historical position sequence of extremely high-intensity reflection points through the control system of the ultra-high-speed photosensitive array means that the control system continuously receives speed information from the robot body and records the pixel coordinates or their corresponding spatial coordinates of the extremely high-intensity reflection points in continuously captured image frames, thereby forming a time-series historical position data. The robot's current scanning speed can be directly provided by the robot controller, while the historical position sequence of the extremely high-intensity reflection points is identified and tracked in each image frame using image processing algorithms.

[0084] Furthermore, identifying the geometric features and abrupt change regions of the weld surface through the control system involves analyzing the acquired weld contour data, preset weld models, or real-time images to identify typical geometric features of the weld, such as weld bevels, weld toes, and weld crowns, as well as potential surface discontinuities, depressions, protrusions, or sharp angles. These geometric features and abrupt change regions are crucial for understanding the movement patterns of extremely high-intensity reflection points on the weld surface, as the trajectory of these reflection points is significantly influenced by these surface morphologies.

[0085] The system segmentes the historical position sequence of extreme high-intensity reflection points based on geometric features and abrupt change regions, and performs nonlinear fitting in conjunction with the robot's current scanning speed to obtain segmented motion trajectories. Given the complexity of the weld surface geometry, the motion trajectory of extreme high-intensity reflection points is not always linear. Therefore, based on the identified geometric features and abrupt change regions, the historical position sequence is divided into several logical segments. For example, different segments can be defined near the straight section of the weld bevel, the arc section of the weld toe, or abrupt change points. For each segment, nonlinear fitting algorithms such as polynomial fitting, spline fitting, Kalman filtering, or Gaussian process regression are used, combined with the robot's current scanning speed, to more accurately describe the motion trend of the extreme high-intensity reflection points within that segment, thus obtaining a series of segmented motion trajectories. This segmented processing and nonlinear fitting method can better adapt to changes in the local geometric features of the weld surface, improving the accuracy of trajectory modeling.

[0086] Therefore, the control system extrapolates and predicts the spatial location of the extreme high-intensity reflection point in the next frame based on the segmented motion trajectory. Once the precise segmented motion trajectory is obtained, the control system can use this trajectory data to calculate the precise spatial location of the extreme high-intensity reflection point in the next frame image using an extrapolation algorithm (e.g., prediction based on the trend of the last segment of the trajectory, or iterative calculation combined with a prediction model). This predicted location will guide the exposure control circuit of the ultra-high-speed photosensitive array to perform ultrashort pulse exposure on the corresponding pixel unit in advance when capturing the next frame image.

[0087] This application addresses the inaccuracy issues of traditional prediction methods caused by the complex geometry of the weld and variations in robot scanning speed by introducing the identification of geometric features and abrupt change regions on the weld surface, and by segmenting and nonlinearly fitting the historical position sequence of extremely high-intensity reflection points. Specifically, acquiring the robot's current scanning speed and historical position sequence provides the foundational data for prediction. Identifying the geometric features and abrupt change regions on the weld surface enables the system to understand the inherent laws governing the motion of reflection points, especially when the geometry changes. Therefore, segmenting the historical position sequence and combining it with nonlinear fitting based on the robot's scanning speed allows for more accurate modeling of the complex motion trajectory of reflection points, avoiding the limitations of a single model in complex scenarios. Finally, extrapolation prediction based on these precise segmented motion trajectories ensures higher accuracy in predicting the spatial position of extremely high-intensity reflection points in the next frame.

[0088] The above technical solution significantly improves the prediction accuracy of the spatial position of extreme high-intensity reflection points in the next frame. By fully considering the complex geometric features of the weld surface and the dynamic changes in robot scanning speed, the prediction method of this application can adapt to various complex weld morphologies and scanning conditions, thereby ensuring that ultrashort pulse exposure can more accurately target high-intensity reflection areas and effectively suppress oversaturation in the image. Therefore, under fixed single exposure parameters, the quality of the captured weld contour image is significantly improved, providing a more reliable data foundation for subsequent 3D reconstruction and weld inspection.

[0089] In some preferred embodiments, it is assumed that a robot is scanning a weld with a V-groove and irregular weld toe. The control system of the ultra-high-speed photosensitive array continuously acquires the robot's current scanning speed, for example, 50 mm / s, and records the pixel coordinate sequence of extremely high-intensity reflection points (e.g., specular reflections from the weld groove edge) over the past 100 frames. The control system first analyzes existing weld contour data or real-time images to identify the straight segments of the V-groove and the arc-shaped abrupt change region at the weld toe. Subsequently, the control system divides the historical position sequence into different segments based on these geometric features. For example, linear or low-order polynomial fitting can be used for the straight segments of the V-groove; while for the arc-shaped region of the weld toe, high-order polynomial or spline fitting can be used, optimized in conjunction with the robot's current scanning speed. Through this piecewise nonlinear fitting, a trajectory that accurately reflects the movement of the reflection point on the weld surface is obtained. Finally, based on this piecewise motion trajectory, the control system extrapolates and predicts the precise pixel position of the extremely high-intensity reflection point in the next frame, for example, (X+ΔX, Y+ΔY). When capturing the next frame of the image, the exposure control circuit of the ultra-high-speed photosensitive array can perform an ultra-short pulse exposure on the pixel unit at the predicted position in advance, thereby effectively suppressing oversaturation in that area and ensuring image quality.

[0090] In some embodiments described above, the control system predicts the spatial position of the extreme high-intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high-intensity reflection point in the image. However, in practical applications, due to factors such as the complex and variable weld surface, fluctuating ambient lighting, and sensor noise, the acquired historical position sequence data of the extreme high-intensity reflection point may contain outliers or noise points. Directly segmenting and nonlinearly fitting these raw data may result in an inaccurate fitted motion trajectory, thereby affecting the accuracy of the next frame position prediction and reducing the control precision of ultrashort pulse exposure.

[0091] To address this, this application further proposes segmenting the historical position sequence of the extremely high-intensity reflection points and performing nonlinear fitting in conjunction with the robot's current scanning speed to obtain segmented motion trajectories, including: The historical position sequence of each segment is preprocessed, including identifying and removing outliers and performing local smoothing on the remaining data. The segmented motion trajectory is obtained by nonlinearly fitting the historical position sequence after data preprocessing to the robot's current scanning speed.

[0092] Specifically, "data preprocessing" refers to a series of cleaning and optimization operations performed on the raw data before performing nonlinear fitting on the historical location sequence, aiming to improve data quality and fitting accuracy. Its purpose is to eliminate noise and outliers in the data, ensuring that subsequent fitting processes are based on a more reliable dataset.

[0093] "Identifying and removing outliers" refers to using specific algorithms or statistical methods to detect and remove points in a dataset that significantly deviate from other data points. These outliers may originate from measurement errors, transient interference, or abrupt changes in the weld surface, and if left untreated, they will severely distort the accuracy of the fitted model. Removing them can effectively avoid their negative impact on the overall fitting results.

[0094] "Local smoothing of the remaining data" refers to smoothing the cleaned data sequence within a local region after outliers have been removed. This helps reduce random fluctuations in the data, making the data trend clearer and more continuous, thus providing a smoother input for nonlinear fitting. Local smoothing can be implemented using various techniques, such as moving averages, Gaussian smoothing, or Savitzky-Golay filters, and the choice can be adjusted according to the specific data characteristics and the required degree of smoothing.

[0095] In practical applications, "nonlinear fitting of the preprocessed historical position sequence with the robot's current scanning speed" refers to using the cleaned and smoothed historical position sequence and the robot's current scanning speed as input parameters after data preprocessing, and constructing a motion model using a nonlinear fitting algorithm. Compared to linear fitting, nonlinear fitting can more accurately capture the motion patterns of extremely high-intensity reflection points under the complex geometry of the weld surface, especially when the robot's scanning speed changes or the weld trajectory is curved, providing a more refined and accurate trajectory description.

[0096] This application's solution effectively addresses the adverse effects of noise and outliers in the original historical position sequences on fitting accuracy by introducing a data preprocessing step before nonlinear fitting. Specifically, identifying and removing outliers eliminates interference from abnormal data in the construction of the motion trajectory model, ensuring the fitting process is based on a more reliable dataset. Subsequently, local smoothing is applied to the remaining data to further eliminate random fluctuations, making the data trend clearer and more continuous, providing high-quality input for nonlinear fitting. It is precisely because of this thorough data preprocessing that the subsequent nonlinear fitting, combined with the robot's current scanning speed, can more accurately capture the actual motion patterns of extremely high-intensity reflection points, thus obtaining a more precise segmented motion trajectory.

[0097] Through the above technical solutions, this application significantly improves the accuracy and robustness of predicting the motion trajectory of extreme high-intensity reflection points. By identifying and eliminating outliers, interference from abnormal data on the fitting results is avoided, ensuring the reliability of the motion trajectory model. Local smoothing further reduces data noise, making the fitted trajectory smoother and more realistic. These improvements enable more accurate prediction of the spatial position of the next frame, thereby allowing for more precise control of the timing and position of ultrashort pulse exposure, effectively avoiding overexposure or underexposure problems caused by inaccurate prediction, and improving the overall quality and efficiency of weld contour data acquisition.

[0098] In some preferred embodiments, the following steps can be used when preprocessing the historical position sequence of each segment: First, we statistically analyze the distance distribution between each data point in the local area and other data points in the neighborhood. For example, we can calculate the average distance from each data point to its K nearest neighbors.

[0099] Secondly, based on the distance distribution, clusters of data points that deviate significantly from other data points are identified as outlier clusters. For example, if the average distance of a data point to its K nearest neighbors is much greater than the average distance threshold of the entire dataset, it is marked as an outlier.

[0100] Subsequently, data points from the outlier clusters are removed. Finally, within a local region, the width and weight of the smoothing kernel function are dynamically adjusted based on the density of the data points and the intensity differences between adjacent data points to perform local smoothing. For example, in regions with high data point density, a narrower smoothing kernel function can be used to retain more details; while in regions with sparse data points or large intensity differences, a wider smoothing kernel function can be used to achieve a stronger smoothing effect.

[0101] After such preprocessing, and combined with the robot's current scanning speed for nonlinear fitting, such as using the least squares method or the Levenberg-Marquardt algorithm to perform polynomial fitting or spline curve fitting on the preprocessed data, a more accurate and stable segmented motion trajectory can be obtained.

[0102] In some embodiments described above, a data preprocessing method is proposed for the historical position sequence of each segment, including identifying and removing outliers and performing local smoothing on the remaining data, to perform nonlinear fitting in conjunction with the robot's current scanning speed to obtain the segmented motion trajectory. However, in practical applications, the historical position sequence data may contain various noises, transient interferences, or erroneous data caused by abnormal reflections. If the outlier identification is not accurate enough or the local smoothing is not adaptive enough, the data preprocessing effect may be poor, thereby affecting the accuracy of subsequent nonlinear fitting and the robustness of trajectory prediction.

[0103] In response, this application further proposes the above-mentioned data preprocessing for the historical position sequence of each segment, including: The distance distribution between data points within a local area and other data points in the neighboring area is statistically analyzed. Based on the distance distribution, clusters of data points that deviate significantly from other data points are identified as outlier clusters; Remove data points from the outlier clusters; Within a local region, the width and weight of the smoothing kernel function are dynamically adjusted based on the density of the data points and the intensity difference between adjacent data points to perform local smoothing.

[0104] Specifically, when preprocessing the historical location sequence for each segment, it is first necessary to statistically analyze the distance distribution between data points within a local region and other data points in their neighborhood. This local region can be a pre-defined window of fixed size or an adaptive region based on data point density. The distance distribution can be calculated by determining the Euclidean distance, Manhattan distance, or other suitable distance metric between each data point and all other data points in its neighborhood. For example, the average or median distance from each data point to its K nearest neighbors can be calculated.

[0105] Based on the distance distribution, clusters of data points that deviate significantly from other data points are identified as outlier clusters. For example, a statistical threshold can be set; if the distance distribution parameters (such as mean distance, standard deviation, or skewness) from a data point to other data points in its neighborhood significantly exceed this threshold, the data point is initially identified as an outlier. Furthermore, density-based clustering algorithms (such as DBSCAN) or methods based on statistical principles (such as box plots, Z-scores, or IQR criteria) can be used to identify data points that exhibit anomalous spatial aggregation or isolation, thereby forming outlier clusters.

[0106] Subsequently, data points in the outlier clusters are removed. This step aims to eliminate erroneous or inaccurate data points caused by sensor errors, environmental noise, transient reflection anomalies, or robot motion jitter, thereby purifying the data sequence and ensuring the quality and reliability of the dataset used for subsequent fitting.

[0107] Within a local region, the width and weight of the smoothing kernel function are dynamically adjusted based on the density of the remaining data points and the intensity differences between adjacent data points to perform local smoothing. The smoothing kernel function can be a Gaussian kernel, mean kernel, median filter, or other nonlinear filter. Dynamically adjusting the kernel width and weight means that the smoothing degree is not fixed but adaptively adjusted according to local data characteristics. For example, in regions with high data point density and small intensity differences between adjacent data points (typically corresponding to straight sections of a weld), a wider kernel function and larger weights can be used for stronger smoothing to effectively suppress random noise; while in regions with low data point density or large intensity differences between adjacent data points (e.g., weld corners, edges, or geometric abrupt changes), a narrower kernel function and smaller weights are used to avoid over-smoothing that could lead to the loss or blurring of important geometric features. The intensity differences can refer to image pixel intensity, laser reflection intensity, or the height value of contour points, etc.

[0108] Through the above technical solutions, this application can significantly improve the accuracy and robustness of preprocessing historical location sequence data of extreme high-intensity reflection points. Specifically, through a refined outlier identification and removal mechanism, erroneous data caused by environmental interference, sensor noise, or instantaneous reflection anomalies can be effectively filtered out, avoiding the negative impact of these abnormal data on subsequent trajectory fitting. In addition, dynamically adjusting the width and weight of the smoothing kernel function allows the smoothing process to adaptively match local data characteristics, effectively suppressing noise in stable regions while accurately preserving key details in areas of abrupt changes in weld geometry (such as weld edges and corners), avoiding the feature blurring or over-smoothing problems that may occur with traditional fixed-parameter smoothing. This optimized data preprocessing method provides a higher quality and more reliable data foundation for subsequent nonlinear fitting, thereby ensuring the accuracy and stability of the spatial location prediction of extreme high-intensity reflection points in the next frame, and thus improving the overall performance and applicability of the entire 3D binocular laser weld contour data acquisition method.

[0109] Reference Figure 4 The first embodiment of the present invention provides a structural diagram of a 3D binocular laser weld seam contour data acquisition system, comprising: The detection end is used to acquire information about the surface reflection characteristics of the area to be scanned; The adjustment end is used to adjust the output intensity of the laser line projected by the line laser at different positions in the area to be scanned, based on the surface reflection characteristic information. The capture end is used to fix a single exposure parameter and simultaneously capture images of laser lines projected with adjustable intensity.

[0110] It should be noted that the 3D binocular laser weld contour data acquisition system provided in this embodiment of the invention is used to execute all the process steps of the 3D binocular laser weld contour data acquisition method in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0111] This invention also provides a terminal device. The terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps described in the various embodiments of the 3D binocular laser weld contour data acquisition method, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments.

[0112] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0113] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for acquiring 3D binocular laser weld contour data, characterized in that, The method includes: Obtain surface reflection characteristics information of the area to be scanned; Based on the surface reflection characteristics information, adjust the output intensity of the laser line projected by the line laser at different positions in the area to be scanned; With fixed single exposure parameters, images of laser lines with adjustable intensity are captured simultaneously.

2. The 3D binocular laser weld contour data acquisition method according to claim 1, characterized in that, The step of adjusting the output intensity of the laser line projected by the line laser at different positions in the area to be scanned based on the surface reflection characteristic information includes: A probe laser line and a measurement laser line are projected, wherein the probe laser line is located in front of the measurement laser line in the scanning direction; The intensity data of reflected light from the probe laser line on the surface of the area to be scanned is captured in real time; the reflected intensity data is converted into an intensity modulation command required by the measuring laser line at the same spatial position; Before the measuring laser line illuminates the surface area scanned by the probe laser line, the output intensity of the measuring laser line is adjusted; By adjusting the deflection angle of the micromirror, the intensity gradient modulation of the measurement laser line can be achieved in space.

3. The method for acquiring 3D binocular laser weld contour data according to claim 2, characterized in that, The real-time capture of the reflected light intensity data of the probe laser line on the surface of the area to be scanned includes: Multiple frames of probe laser line reflection images are continuously captured using an ultra-high-speed photosensitive array; Identify high-intensity reflection regions and low-intensity diffuse reflection regions in the multi-frame probe laser line reflection images; Based on the high-intensity reflection region, adjust the gain or exposure time of the high-intensity reflection region corresponding to the ultra-high-speed photosensitive array; Based on the low-intensity diffuse reflection region, adjust the gain or exposure time of the ultra-high-speed photosensitive array corresponding to the low-intensity diffuse reflection region. By fusing the multi-frame images of the probe laser line reflection, the intensity distribution of the reflected light from the probe laser line on the surface of the area to be scanned is obtained.

4. The 3D binocular laser weld contour data acquisition method according to claim 3, characterized in that, The process of fusing the multi-frame probe laser line reflection images to obtain the reflected light intensity distribution of the probe laser line on the surface of the area to be scanned includes: Feature point extraction and feature point matching are performed on the multi-frame detection laser line reflection images; Based on the feature point matching results, the spatial transformation parameters between the multi-frame probe laser line reflection images are calculated; Based on the spatial transformation parameters, the multi-frame probe laser line reflection images are spatially aligned; By fusing the spatially aligned multi-frame probe laser line reflection images, the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned is obtained.

5. The method for acquiring 3D binocular laser weld contour data according to claim 3, characterized in that, The step of adjusting the gain or exposure time of the high-intensity reflection region corresponding to the high-intensity reflection region of the ultra-high-speed photosensitive array includes: Identify extremely high-intensity reflection points in an image whose local brightness far exceeds a preset threshold; For the pixel units corresponding to the extremely high-intensity reflection points in the ultra-high-speed photosensitive array, perform ultra-short pulse exposure; For the low-intensity diffuse reflection region surrounding the extremely high-intensity reflection point, capture it while maintaining normal gain or exposure time; By fusing images captured under different exposure conditions, the intensity distribution of reflected light from the probe laser line on the surface of the area to be scanned is obtained.

6. The 3D binocular laser weld contour data acquisition method according to claim 5, characterized in that, The step of performing ultrashort pulse exposure on the pixel units corresponding to the extremely high-intensity reflection points in the ultra-high-speed photosensitive array includes: The pixel units of the ultra-high-speed photosensitive array are configured such that each pixel unit is independently configured with an exposure control circuit and a gain adjustment circuit; The output signal strength of each pixel unit is monitored in real time using the ultra-high-speed photosensitive array. The ultra-high-speed photosensitive array identifies pixel units whose output signal strength exceeds a preset threshold. The control system of the ultra-high speed photosensitive array predicts the spatial position of the extreme high intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high intensity reflection point in the image. When capturing the next frame of the image, the exposure control circuit of the ultra-high-speed photosensitive array performs an ultra-short pulse exposure in advance on the pixel unit at the predicted position. The exposure pulse width of the corresponding pixel unit is dynamically adjusted by the exposure control circuit of the ultra-high-speed photosensitive array.

7. The 3D binocular laser weld contour data acquisition method according to claim 6, characterized in that, The control system of the ultra-high-speed photosensitive array predicts the spatial position of the extreme high-intensity reflection point in the next frame based on the robot's current scanning speed and the movement trend of the extreme high-intensity reflection point in the image, including: The machine scanning speed and the historical position sequence of the extreme high-intensity reflection point are obtained through the control system of the ultra-high speed photosensitive array. The control system identifies the geometric features and abrupt change areas on the weld surface. The control system segments the historical position sequence of the extreme high-intensity reflection point according to the geometric features and the abrupt change region, and performs nonlinear fitting in combination with the robot's current scanning speed to obtain the segmented motion trajectory. The control system extrapolates and predicts the spatial position of the extreme high-intensity reflection point in the next frame based on the segmented motion trajectory.

8. The 3D binocular laser weld contour data acquisition method according to claim 7, characterized in that, The process of segmenting the historical position sequence of the extremely high-intensity reflection points and performing nonlinear fitting in conjunction with the robot's current scanning speed to obtain segmented motion trajectories includes: The historical position sequence of each segment is preprocessed, including identifying and removing outliers and performing local smoothing on the remaining data. The segmented motion trajectory is obtained by nonlinearly fitting the historical position sequence after data preprocessing to the robot's current scanning speed.

9. A method for acquiring 3D binocular laser weld contour data according to claim 8, characterized in that, The data preprocessing for the historical location sequence of each segment includes: The distance distribution between data points within a local area and other data points in the neighboring area is statistically analyzed. Based on the distance distribution, clusters of data points that deviate significantly from other data points are identified as outlier clusters; Remove data points from the outlier clusters; Within a local region, the width and weight of the smoothing kernel function are dynamically adjusted based on the density of the data points and the intensity difference between adjacent data points to perform local smoothing.

10. A 3D binocular laser weld seam contour data acquisition system, characterized in that, The system includes: The detection end is used to acquire information about the surface reflection characteristics of the area to be scanned; The adjustment end is used to adjust the output intensity of the laser line projected by the line laser at different positions in the area to be scanned, based on the surface reflection characteristic information. The capture end is used to fix a single exposure parameter and simultaneously capture images of laser lines projected with adjustable intensity.

Images