Bubble three-dimensional reconstruction method based on point cloud technology

Abstract

The invention discloses a bubble three-dimensional reconstruction method based on a point cloud technology, and relates to the technical field of fluid mechanics and computer vision. The invention provides a bubble three-dimensional reconstruction method based on a point cloud technology in order to solve the problem that in the prior art, a dynamic bubble three-dimensional structure is difficult to accurately and rapidly reconstruct in a complex geometric environment, and consequently the interface area concentration and flow characteristic analysis precision is insufficient. According to the method, bubble two-dimensional images are synchronously collected through a multi-angle high-speed camera, after histogram equalization and clear contour extraction, point cloud data are generated layer by layer in the bubble height direction, and multi-view point cloud registration is carried out through an iterative nearest point algorithm. And then a continuous closed three-dimensional model is generated, and model optimization is realized in combination with Laplacian smoothing and normal filtering. According to the method, the real form of the bubbles can be recovered with high precision, and reliable technical support is provided for gas-liquid two-phase flow interface area concentration calculation and dynamic behavior analysis in complex environments such as nuclear reactor rod bundle sub-channels.

Description

Technical Field

[0001] It involves the fields of fluid mechanics and computer vision technology, specifically bubble 3D reconstruction based on point cloud technology. Background Technology

[0002] In the study of gas-liquid two-phase flow, the morphology, size distribution, and dynamic behavior of bubbles in the flow field are key parameters for understanding heat transfer, mass transfer, and flow characteristics. Especially in complex and confined spaces such as nuclear reactor rod bundle channels and microchannel cooling systems, the three-dimensional structure of bubbles directly affects the local hydrodynamic behavior and the accuracy of the calculated interfacial area concentration (ai), thus affecting the reliability and safety analysis results of the two-phase flow model.

[0003] Currently, various methods are commonly used in academia and engineering to measure the morphology and size of bubbles, mainly including two-dimensional imaging, laser tomography, CT reconstruction, and stereo vision. For example, traditional two-dimensional imaging relies on high-speed cameras to acquire the planar projection of bubbles and calculates the bubble diameter or shape parameters through image segmentation and edge recognition. However, it can only provide two-dimensional information in a single direction and cannot restore the true three-dimensional structure of the bubble. Although laser tomography and CT reconstruction methods can achieve three-dimensional information acquisition, the equipment is expensive, the system is complex, and it is easily affected by refraction and scattering in water, making it difficult to accurately reconstruct high-speed dynamic bubbles. Stereo vision reconstruction methods attempt to infer the three-dimensional position and morphology of bubbles using binocular or multi-view camera systems. However, due to the transparency, significant reflection and refraction characteristics of the bubble surface, the image matching process is easily interfered with, resulting in unstable point correspondences and affecting reconstruction accuracy. In addition, some studies have attempted to use structured light and particle image velocimetry (PIV) technology to obtain the motion information of bubbles in the flow field, but the results are mainly used for velocity field and trajectory analysis and have not yet solved the problem of three-dimensional geometric reconstruction.

[0004] In summary, existing technologies suffer from the drawback of being unable to accurately and quickly reconstruct the three-dimensional structure of dynamic bubbles in complex geometric environments, resulting in insufficient accuracy in the analysis of interfacial area concentration and flow characteristics. Summary of the Invention

[0005] To address the shortcomings of existing technologies, which struggle to accurately and rapidly reconstruct the three-dimensional structure of dynamic bubbles in complex geometric environments, leading to insufficient accuracy in interfacial area concentration and flow characteristic analysis, the present invention provides the following technical solution: A bubble 3D reconstruction method based on point cloud technology, comprising: Acquire bubble images from multiple angles and capture them simultaneously to obtain a multi-faceted two-dimensional image sequence containing frontal and side views, which will be used for subsequent bubble contour extraction steps. The image preprocessing of the multifaceted two-dimensional image sequence includes the steps of using histogram equalization to enhance the image and improve the contrast, using median filtering to remove noise, and using the Canny operator to perform edge detection to extract bubble contour information, and outputting contour data for point cloud construction. Based on the contour data, the steps are as follows: extract the cross-sectional width layer by layer according to the bubble height direction and calculate the radial distance; form a point cloud set containing three-dimensional coordinates through rotation mapping; and output point cloud data for multi-view fusion. The steps involve spatial registration of point cloud data from different perspectives, point cloud alignment using an iterative nearest point algorithm to obtain a comprehensive point cloud model in a unified coordinate system, and outputting the registration results for surface reconstruction. The registered point cloud model is input into the 3D reconstruction module, and the Poisson surface reconstruction method is used to generate a continuous and closed bubble 3D model. The steps of the initial 3D structure for optimization processing are output. The steps involve optimizing the initial three-dimensional structure, using the Laplacian smoothing method to eliminate surface noise, and enhancing model details through normal filtering to obtain a three-dimensional reconstruction result with a smooth bubble surface and complete shape.

[0006] Furthermore, a preferred embodiment is provided in which the multi-angle bubble image is acquired by simultaneously shooting from the front and side views using two high-speed cameras, with a shooting frequency of 250 frames per second and a shutter speed range of 1 / 6000 to 1 / 9000 seconds.

[0007] Furthermore, a preferred embodiment is provided in which, in the image preprocessing step, image enhancement uses histogram equalization to improve the overall contrast of the image, denoising uses median filtering to suppress random noise, and edge detection uses the Canny operator method to accurately extract bubble boundaries.

[0008] Furthermore, a preferred implementation is provided in which, during the generation of the point cloud set, the contour width is analyzed layer by layer along the bubble height direction and the radial distance is calculated, and a point cloud data set containing three-dimensional coordinates is obtained through rotation mapping.

[0009] Furthermore, a preferred implementation method is provided, in which the point cloud registration step adopts the iterative nearest point algorithm, which achieves point cloud alignment by minimizing the Euclidean distance error between point clouds from different viewpoints, and obtains a comprehensive point cloud model under a unified coordinate system.

[0010] Furthermore, a preferred embodiment is provided in which the Poisson surface reconstruction method is used to perform global smooth fitting on the point cloud data in the 3D reconstruction step to generate a continuous, closed and topologically complete bubble 3D model.

[0011] Based on the same inventive concept, the present invention also provides a bubble 3D reconstruction device based on point cloud technology, comprising: A module that acquires bubble images from multiple angles and captures them simultaneously to obtain a multi-faceted two-dimensional image sequence containing frontal and side views is used for subsequent bubble contour extraction. The multi-faceted two-dimensional image sequence is preprocessed, including image enhancement by histogram equalization to improve contrast, noise removal by median filtering, edge detection by the Canny operator to extract bubble contour information, and outputting contour data for point cloud construction. Based on the contour data, the cross-sectional width is extracted layer by layer along the bubble height direction and the radial distance is calculated. A point cloud set containing three-dimensional coordinates is formed through rotation mapping, and a module for outputting point cloud data for multi-view fusion is generated. This module performs spatial registration on point cloud data from different perspectives, uses the iterative nearest point algorithm to align the point clouds, obtains a comprehensive point cloud model in a unified coordinate system, and outputs the registration results for surface reconstruction. The registered point cloud model is input into the 3D reconstruction module, and the Poisson surface reconstruction method is used to generate a continuous and closed bubble 3D model. The module outputs the initial 3D structure for optimization processing. The module optimizes the initial three-dimensional structure by using the Laplacian smoothing method to eliminate surface noise and enhancing model details through normal filtering, resulting in a three-dimensional reconstruction of the bubble with a smooth surface and complete shape.

[0012] Based on the same inventive concept, the present invention also provides a computer storage medium for storing a computer program, wherein when the computer program is read by a computer, the computer executes the method described thereon.

[0013] Based on the same inventive concept, the present invention also provides a computer, including a processor and a storage medium, wherein when the processor reads a computer program stored in the storage medium, the computer executes the method described thereon.

[0014] Based on the same inventive concept, the present invention also provides a computer program product, which, when executed, implements the method described.

[0015] Compared with the prior art, the advantages of the technical solution provided by the present invention are as follows: This invention introduces a multi-angle high-speed camera synchronous imaging method in the image acquisition stage, which enables the complete recording of the morphological features of bubbles from different perspectives. Compared with the traditional single-view two-dimensional imaging method, it significantly reduces the problem of geometric information loss caused by bubble occlusion, reflection and refraction, thus providing a more comprehensive data foundation for subsequent three-dimensional reconstruction.

[0016] This invention employs a combination of histogram equalization to enhance contrast, median filtering to remove noise, and the Canny operator to extract edges during the image preprocessing stage, effectively improving the accuracy of bubble contour extraction. Compared to conventional threshold segmentation or the Sobel operator, this combined processing can reduce false edge interference while maintaining contour integrity, thereby improving the stability and accuracy of contour information.

[0017] In the point cloud generation stage, this invention analyzes and calculates the radial distance layer by layer according to the height direction of the bubble based on the two-dimensional contour information of the bubble, and constructs a point cloud coordinate set through rotational symmetry. This method enables high-fidelity reproduction of the geometric features of the bubble surface, and compared with the traditional method of directly inferring the three-dimensional shape from pixel grayscale, it can more realistically reflect the changes in bubble volume and surface curvature.

[0018] In the point cloud registration stage, this invention employs the Iterative Closest Point (ICP) algorithm to align point clouds generated from multiple perspectives with high precision, significantly improving the consistency of point cloud fusion from different angles and the continuity of the overall model. Compared with existing feature-matching-based registration algorithms, this method can achieve stable registration even when the bubble surface texture is not obvious, reducing mismatch points and structural distortion.

[0019] In the 3D reconstruction stage, this invention introduces the Poisson surface reconstruction method to perform continuous surface fitting on the registered point cloud, resulting in a smooth and closed 3D bubble model. Compared with traditional Delaunay triangulation or Alpha reconstruction algorithms, this method can better address the problem of uneven point cloud density, enabling the generated bubble model to have higher accuracy and visual consistency in complex curved surface regions.

[0020] In the model optimization stage, this invention combines Laplacian smoothing and normal filtering techniques to suppress noise and enhance details in the 3D model. This approach ensures the realism of the model structure while eliminating surface burrs caused by lighting variations or image noise during reconstruction, resulting in a smoother, more continuous 3D bubble model with excellent structural visualization. This provides high-quality input data for subsequent analysis of interface area concentration and fluid dynamics parameters.

[0021] It is applicable to the reconstruction of the three-dimensional structure and dynamic characteristic analysis of bubbles in gas-liquid two-phase flow in complex flow environments such as nuclear reactor rod bundle channels. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method. Figure 2 Example of a two-dimensional outline image of a bubble. Figure 3 This is an example of bubble point cloud data. Figure 4 This is an example of a 3D reconstruction model of a bubble. Detailed Implementation

[0023] To make the advantages and benefits of the technical solution provided by the present invention clearer, the technical solution provided by the present invention will now be described in further detail with reference to the accompanying drawings, specifically: Implementation Method 1: This implementation method provides a bubble 3D reconstruction method based on point cloud technology, including: Acquire bubble images from multiple angles and capture them simultaneously to obtain a multi-faceted two-dimensional image sequence containing frontal and side views, which will be used for subsequent bubble contour extraction steps. The image preprocessing of the multifaceted two-dimensional image sequence includes the steps of using histogram equalization to enhance the image and improve the contrast, using median filtering to remove noise, and using the Canny operator to perform edge detection to extract bubble contour information, and outputting contour data for point cloud construction. Based on the contour data, the steps are as follows: extract the cross-sectional width layer by layer according to the bubble height direction and calculate the radial distance; form a point cloud set containing three-dimensional coordinates through rotation mapping; and output point cloud data for multi-view fusion. The steps involve spatial registration of point cloud data from different perspectives, point cloud alignment using an iterative nearest point algorithm to obtain a comprehensive point cloud model in a unified coordinate system, and outputting the registration results for surface reconstruction. The registered point cloud model is input into the 3D reconstruction module, and the Poisson surface reconstruction method is used to generate a continuous and closed bubble 3D model. The steps of the initial 3D structure for optimization processing are output. The steps involve optimizing the initial three-dimensional structure, using the Laplacian smoothing method to eliminate surface noise, and enhancing model details through normal filtering to obtain a three-dimensional reconstruction result with a smooth bubble surface and complete shape.

[0024] Multi-angle bubble images were acquired by simultaneously shooting from the front and side views using two high-speed cameras at a shooting frequency of 250 frames per second and a shutter speed range of 1 / 6000 to 1 / 9000 seconds.

[0025] In the image preprocessing steps, image enhancement uses histogram equalization to improve the overall image contrast, denoising uses median filtering to suppress random noise, and edge detection uses the Canny operator to accurately extract bubble boundaries.

[0026] During the generation of the point cloud set, the contour width is analyzed layer by layer along the height direction of the bubble and the radial distance is calculated. A point cloud data set containing three-dimensional coordinates is obtained through rotation mapping.

[0027] The point cloud registration step uses an iterative nearest point algorithm to align the point clouds by minimizing the Euclidean distance error between point clouds from different viewpoints, thus obtaining a comprehensive point cloud model in a unified coordinate system.

[0028] In the 3D reconstruction step, the Poisson surface reconstruction method is used to perform global smooth fitting on the point cloud data to generate a continuous, closed and topologically complete bubble 3D model.

[0029] Implementation Method Two: This implementation method is a further detailed description of the technical solution provided in Implementation Method One, specifically: A bubble 3D reconstruction method based on point cloud technology is applicable to the identification and structural reconstruction of bubble morphology in complex flow environments such as nuclear reactor rod bundle channels. This method achieves high-precision reconstruction of the bubble's 3D morphology through steps including multi-view image acquisition, image preprocessing, point cloud generation, point cloud registration, 3D reconstruction, and model optimization.

[0030] First, image acquisition was performed. High-speed cameras were placed on both sides of the experimental setup, and the movement of the bubble in the channel was simultaneously captured from both frontal and side views at a sampling frequency of 250 frames per second and a shutter speed of 1 / 6000 to 1 / 9000 of a second, obtaining a multi-faceted two-dimensional image sequence. The experimental system used a rectangular transparent test section simulating a nuclear reactor rod bundle channel. The bubble was formed by injection from a bottom air injector, and the water flow was provided with a stable flow rate by a constant flow pump. Through synchronous trigger control, the temporal correspondence of images from different angles was achieved, providing a foundation for subsequent data registration.

[0031] Next, image preprocessing is performed. The acquired raw image is input into the image processing module, where image enhancement, denoising, and edge detection are executed sequentially. Image enhancement uses histogram equalization to improve overall contrast, making the bubbles more distinct from the background; denoising uses median filtering to effectively eliminate random noise caused by lighting or water flow disturbances; edge detection uses the Canny operator algorithm to identify areas with significant grayscale changes, thereby extracting the bubble's contour boundaries. The preprocessed image retains the main morphological features of the bubble, providing support for generating stable contour data.

[0032] Next, the point cloud generation step is performed. Based on the preprocessed bubble contour information, the contour width of each layer is extracted layer by layer along the height direction, and the corresponding radial distance is calculated. By rotating and mapping the cross-sectional width information at different heights, a point cloud dataset containing three-dimensional coordinates is generated. Each point cloud point corresponds to a spatial position on the bubble surface, thus forming a discretized geometric representation of the bubble surface. The point cloud data output by this step can completely describe the spatial shape of the bubble, providing data input for subsequent multi-angle registration and 3D reconstruction.

[0033] Next, point cloud registration is performed. Point cloud data from different perspectives (such as frontal and side views) are spatially aligned. An iterative nearest-point algorithm is used to minimize the distance error between the two sets of point clouds by continuously adjusting the translation and rotation parameters of the point clouds. The registered point cloud set forms a unified geometric representation in the same coordinate system, ensuring the continuity of the bubble surface structure and accurate contour matching after the fusion of multi-view information. The output of this step is a comprehensive point cloud model in a unified coordinate system.

[0034] Next, 3D reconstruction is performed. The registered integrated point cloud data is input into the reconstruction module, and the Poisson surface reconstruction method is used for surface fitting and closed surface generation. This method generates a continuous, closed, and topologically complete 3D bubble model through overall smooth interpolation of the point cloud distribution. The reconstruction result can realistically reflect the volume, surface area, and surface curvature distribution of the bubble, providing basic geometric data for subsequent physical analysis.

[0035] Finally, model optimization is performed. The generated 3D bubble model undergoes post-processing to improve its geometric accuracy and visual quality. The optimization process includes two parts: smoothing and detail enhancement. Smoothing uses the Laplacian smoothing algorithm to remove surface fluctuations caused by point cloud noise during reconstruction; detail enhancement employs normal filtering technology to enhance subtle surface undulations and edge features of the bubble. The optimized model has a smooth surface, coherent structure, and realistically visible details, accurately reflecting the spatial morphology of the bubble.

[0036] By executing the above steps sequentially, this invention achieves the complete reconstruction process from multi-view two-dimensional images to a three-dimensional bubble model. The data flow between each step is clear: image acquisition provides raw information for preprocessing; preprocessing outputs stable contour data for point cloud generation; point cloud generation forms a registerable three-dimensional data foundation; the registration result serves as input for three-dimensional reconstruction; and the final optimized model can be directly used for interface area concentration calculation and two-phase flow dynamic analysis. Compared with traditional two-dimensional analysis methods, this invention significantly improves the accuracy and reliability of bubble three-dimensional morphology reconstruction, especially under complex optical and geometric conditions, while still consistently obtaining high-quality reconstruction results.

[0037] Implementation Method 3: Combination Figure 1-4 This embodiment describes the technical solution provided above in further detail through specific examples. Specifically: This embodiment provides a bubble 3D reconstruction method based on point cloud technology. It aims to generate and integrate the 3D structure of the bubble by acquiring multi-faceted 2D contour images and using point cloud technology, thereby more accurately reflecting the true morphology and dynamic characteristics of the bubble and providing reliable data support for fluid mechanics research and nuclear reactor thermal-hydraulic analysis. The method of this embodiment includes the following steps: Image acquisition: Use a high-speed camera to capture two-dimensional images of the bubble from different angles to obtain multi-faceted two-dimensional contour images. Image preprocessing: The acquired two-dimensional image is preprocessed, including image enhancement, denoising and edge detection, to extract the contour information of the bubble. Point cloud generation: Based on the extracted contour information, point cloud technology is used to generate point cloud data of the bubble, representing the three-dimensional discrete point set on the bubble surface. Point cloud registration: Register point cloud data from different angles and integrate them into a unified bubble point cloud model. 3D Reconstruction: Perform 3D reconstruction on the registered point cloud data to generate a 3D model of the bubble. Model optimization: The generated 3D model is optimized, including smoothing and detail enhancement, to improve the accuracy and realism of the model.

[0038] The core innovation of this implementation method lies in: Two-dimensional contours are reconstructed using point cloud technology: By converting two-dimensional contour information into three-dimensional point cloud data, the geometric properties of the bubble surface can be accurately represented. Integrating multi-faceted 2D contours: Multiple sets of point cloud data generated from multi-view images are integrated into a unified 3D model through registration technology, overcoming the problem of insufficient information from a single viewpoint. Reconstructing 3D Bubble Structures: Combining point cloud generation and 3D reconstruction techniques, high-fidelity 3D bubble models are generated, suitable for reconstructing dynamic, non-rigid bubbles.

[0039] Through the above technical solution, this embodiment can significantly improve the accuracy of bubble three-dimensional reconstruction, providing more reliable basic data for the calculation of interface area concentration ($a_i$) and dynamic analysis of two-phase flow, and is particularly suitable for complex environments such as nuclear reactor rod bundle channels.

[0040] In a specific embodiment: like Figure 1 As shown, this embodiment proposes a bubble 3D reconstruction method based on point cloud technology, applicable to bubble morphology analysis in nuclear reactor rod bundle channels. The method is implemented through the following steps: Image Acquisition Two high-speed cameras (shooting frequency 250 frames / second, shutter speed range 1 / 6000 to 1 / 9000 second) were used to simultaneously capture two-dimensional images of the bubbles from different angles (e.g., frontal and side views). The experimental setup was a 2×2 rectangular test section simulating a nuclear reactor rod bundle channel, 500 mm long, 32 mm wide, and with a hydraulic diameter of 26 mm. The bubbles were generated by injecting air into the water through a bottom-mounted air injector. Figure 2 The image shown is an example of a two-dimensional outline image of a bubble.

[0041] Image preprocessing The acquired 2D images are preprocessed to extract the bubble's contour information. Specifically, this includes: Image enhancement: Histogram equalization is used to improve image contrast. Denoising: Use median filtering to remove image noise. Edge detection: The Canny operator is used to extract the contour boundaries of the bubbles and generate pixel-level contour data.

[0042] Point cloud generation Based on the preprocessed contour information, point cloud technology is used to generate point cloud data for the bubble. The contour width is analyzed layer by layer along the bubble height (z-axis), the radial distance $r_i$ at each height is calculated, and three-dimensional point cloud coordinates are generated through rotational symmetry. For example... Figure 3 The image shows an example of the generated bubble point cloud data. The point cloud is represented as a discrete set of points containing three-dimensional coordinates $(x, y, z)$.

[0043] Point cloud registration Point cloud data from different perspectives (front view and side view) are registered and integrated into a unified point cloud model. The registration process uses the Iterative Closest Point (ICP) algorithm to achieve alignment of point clouds from multiple perspectives by minimizing the distance error between point sets.

[0044] 3D Reconstruction The registered point cloud data is used to perform 3D reconstruction to generate a 3D model of the bubble. The reconstruction process uses the Pissn surface reconstruction method, which generates a continuous surface mesh by fitting the point cloud data to fully represent the geometry of the bubble.

[0045] Model optimization The generated 3D model is optimized to improve accuracy and realism: Smoothing process: The Laplace smoothing method is used to eliminate surface noise. Detail enhancement: Normal filtering is used to enhance model details. For example... Figure 4 The image shows an example of an optimized 3D bubble reconstruction model.

[0046] Implementation effect Through the steps described above, this implementation successfully reconstructed the three-dimensional structure of a bubble from a multi-faceted two-dimensional contour image. Compared to single-view reconstruction, multi-view point cloud integration reduces volume and surface area estimation errors by approximately 15%-25%. The reconstructed model exhibits high visual consistency with the experimental images and accurately reflects the irregular shape and dynamic characteristics of the bubble. This method provides reliable support for calculating the interfacial area concentration in nuclear reactor rod bundle channels.

[0047] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for 3D reconstruction of bubbles based on point cloud technology, characterized in that, include: Acquire bubble images from multiple angles and capture them simultaneously to obtain a multi-faceted two-dimensional image sequence containing frontal and side views, which will be used for subsequent bubble contour extraction steps. The image preprocessing of the multifaceted two-dimensional image sequence includes the steps of using histogram equalization to enhance the image and improve the contrast, using median filtering to remove noise, and using the Canny operator to perform edge detection to extract bubble contour information, and outputting contour data for point cloud construction. Based on the contour data, the steps are as follows: extract the cross-sectional width layer by layer according to the bubble height direction and calculate the radial distance; form a point cloud set containing three-dimensional coordinates through rotation mapping; and output point cloud data for multi-view fusion. The steps involve spatial registration of point cloud data from different perspectives, point cloud alignment using an iterative nearest point algorithm to obtain a comprehensive point cloud model in a unified coordinate system, and outputting the registration results for surface reconstruction. The registered point cloud model is input into the 3D reconstruction module, and the Poisson surface reconstruction method is used to generate a continuous and closed bubble 3D model. The steps of the initial 3D structure for optimization processing are output. The steps involve optimizing the initial three-dimensional structure, using the Laplacian smoothing method to eliminate surface noise, and enhancing model details through normal filtering to obtain a three-dimensional reconstruction result with a smooth bubble surface and complete shape.

2. The bubble 3D reconstruction method based on point cloud technology according to claim 1, characterized in that, Multi-angle bubble images were acquired by simultaneously shooting from the front and side views using two high-speed cameras at a shooting frequency of 250 frames per second and a shutter speed range of 1 / 6000 to 1 / 9000 seconds.

3. The bubble 3D reconstruction method based on point cloud technology according to claim 1, characterized in that, In the image preprocessing steps, image enhancement uses histogram equalization to improve the overall image contrast, denoising uses median filtering to suppress random noise, and edge detection uses the Canny operator to accurately extract bubble boundaries.

4. The bubble 3D reconstruction method based on point cloud technology according to claim 1, characterized in that, During the generation of the point cloud set, the contour width is analyzed layer by layer along the height direction of the bubble and the radial distance is calculated. A point cloud data set containing three-dimensional coordinates is obtained through rotation mapping.

5. The bubble 3D reconstruction method based on point cloud technology according to claim 1, characterized in that, The point cloud registration step uses an iterative nearest point algorithm to align the point clouds by minimizing the Euclidean distance error between point clouds from different viewpoints, thus obtaining a comprehensive point cloud model in a unified coordinate system.

6. The bubble 3D reconstruction method based on point cloud technology according to claim 1, characterized in that, In the 3D reconstruction step, the Poisson surface reconstruction method is used to perform global smooth fitting on the point cloud data to generate a continuous, closed and topologically complete bubble 3D model.

7. A bubble 3D reconstruction device based on point cloud technology, characterized in that, include: A module that acquires bubble images from multiple angles and captures them simultaneously to obtain a multi-faceted two-dimensional image sequence containing frontal and side views is used for subsequent bubble contour extraction. The multi-faceted two-dimensional image sequence is preprocessed, including image enhancement by histogram equalization to improve contrast, noise removal by median filtering, edge detection by the Canny operator to extract bubble contour information, and outputting contour data for point cloud construction. Based on the contour data, the cross-sectional width is extracted layer by layer along the bubble height direction and the radial distance is calculated. A point cloud set containing three-dimensional coordinates is formed through rotation mapping, and a module for outputting point cloud data for multi-view fusion is generated. This module performs spatial registration on point cloud data from different perspectives, uses the iterative nearest point algorithm to align the point clouds, obtains a comprehensive point cloud model in a unified coordinate system, and outputs the registration results for surface reconstruction. The registered point cloud model is input into the 3D reconstruction module, and the Poisson surface reconstruction method is used to generate a continuous and closed bubble 3D model. The module outputs the initial 3D structure for optimization processing. The module optimizes the initial three-dimensional structure by using the Laplacian smoothing method to eliminate surface noise and enhancing model details through normal filtering, resulting in a three-dimensional reconstruction of the bubble with a smooth surface and complete shape.

8. A computer storage medium for storing computer programs, characterized in that, When the computer program is read by the computer, the computer executes the method of claim 1.

9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 1.

10. A computer program product, as a computer program, is characterized by: When the computer program is executed, it implements the method of claim 1.

Images