A multi-view diamond growth temperature field 3D reconstruction method and system

Abstract

This application discloses a multi-view 3D reconstruction method and system for the temperature field of diamond growth, relating to the field of diamond growth. The method includes simultaneously acquiring multi-view two-dimensional infrared radiation images of the deposition stage and three-dimensional geometric information of the diamond crystal surface; performing coordinate system calibration based on the multi-view two-dimensional infrared radiation images and three-dimensional geometric information; processing the two-dimensional infrared radiation images from each viewpoint to obtain the corresponding two-dimensional temperature distribution map; reconstructing the three-dimensional surface model of the diamond crystal based on the three-dimensional geometric information; matching and mapping the calibrated coordinates of the three-dimensional surface model in the coordinate system with the calibrated coordinates of the corresponding two-dimensional temperature distribution map in the coordinate system to obtain fused multi-view temperature data, generating a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model. This application can obtain a high-precision three-dimensional temperature field of the diamond crystal growth surface.

Description

Technical Field

[0001] This application relates to the field of diamond growth, and in particular to a method and system for 3D reconstruction of the temperature field of diamond growth from multiple perspectives. Background Technology

[0002] In the process of preparing single-crystal diamond by microwave plasma chemical vapor deposition (MPCVD), the three-dimensional temperature distribution on and inside the crystal growth surface is one of the most crucial physical fields determining crystal quality, growth rate, defect formation, and internal stress. Precisely understanding the three-dimensional temperature field is invaluable for comprehending growth mechanisms, optimizing process windows, and suppressing crack and dislocation propagation. Currently, the sensing and monitoring technology for growth temperature in this field is mainly limited to two-dimensional or single-point levels, presenting fundamental limitations.

[0003] 1. Single-point infrared thermometry: This method involves installing a single-point infrared thermometer in the observation window of the crystal chamber and measuring a specific location (usually the top center). The main drawback of this method is its extremely low information dimensionality; it only provides temperature readings at a single discrete point, completely failing to reflect the spatial distribution of temperature on the crystal surface, let alone reveal the three-dimensional temperature gradient. Temperature information in critical areas such as crystal edges and flanks is completely missing, and these areas are often the origin of defects (such as polycrystalline structures or cracks).

[0004] 2. Two-dimensional infrared thermal imaging technology: This is currently a relatively advanced online monitoring method, which uses a single infrared thermal imager to acquire a two-dimensional temperature distribution image (temperature cloud map) of the entire observation field. While this technology can provide temperature distribution information on a plane, it has two inherent and serious drawbacks: Dimensional Miss and Information Confusion: Thermal imagers output the projection of the surface temperature of a three-dimensional object onto a two-dimensional imaging plane. The temperature value of each pixel in the image is the integral result of all radiation sources along the line of sight corresponding to that pixel, penetrating a certain plasma region, and cannot distinguish thermal radiation from the top, sides, or different depths of the crystal. For diamond crystals with three-dimensional morphology (such as growth mounds or steps), the temperatures of the top and sides completely overlap and interfere with each other in the two-dimensional image, resulting in the measured "temperature" not being an accurate mapping of the true surface temperature, but rather severely confused pseudo-information.

[0005] Limited perspective and incomplete data: Thermal imagers with a single fixed perspective inevitably have blind spots. For example, a top-down view cannot show the side of a crystal, and a side view cannot show the top surface. Therefore, it is impossible to obtain temperature data of the complete, closed surface of the crystal, and it is impossible to construct a complete three-dimensional temperature field model that can be used for scientific analysis and process feedback.

[0006] 3. Contact measurements such as thermocouples: This involves embedding the thermocouple in a base or attempting to contact the sample. This method is not only invasive, interfering with the local thermal field, but it can also only measure the temperature at the installation point, failing to obtain any spatial distribution information. It is completely unsuitable for non-intrusive in-situ, full-field monitoring of crystal growth surfaces.

[0007] 4. Limitations of General-Purpose 3D Temperature Field Measurement Techniques: Currently, there are some general-purpose object temperature field measurement methods based on the fusion of multi-view infrared thermal imaging and 3D models. However, these methods are mainly for temperature measurement of conventional industrial objects or static devices, and their application environment differs fundamentally from the MPCVD diamond growth process. The MPCVD deposition chamber contains intense microwave plasma, and the emitted visible light and near-infrared radiation severely interfere with the temperature measurement signals in the mid- and far-infrared bands. Furthermore, the surface morphology and emissivity of diamond crystals change dynamically during growth, requiring absolutely non-contact monitoring to avoid contaminating the thermal field. General-purpose 3D temperature field reconstruction methods do not consider and cannot resolve these MPCVD-specific interferences and constraints.

[0008] Related technologies, especially the two-dimensional infrared thermal imaging technology closest to this application, are limited by their inherent working principle (two-dimensional projection imaging) and system architecture (single-view observation), resulting in their inability to obtain true, complete, and three-dimensional spatial temperature distribution information on the diamond crystal growth surface. Specifically, this manifests as: (1) the inability to accurately map temperature data to the specific location of the three-dimensional morphology; and (2) the inability to obtain the temperature of hidden surfaces (such as the side). This lack of core capability severely restricts the in-depth understanding and precise control of thermodynamic behavior during MPCVD growth.

[0009] Therefore, there is an urgent need to develop a new measurement technique that can realize in-situ reconstruction of the three-dimensional temperature field on the surface of diamond crystal growth. Summary of the Invention

[0010] The purpose of this application is to provide a multi-view diamond growth temperature field 3D reconstruction method and system, which can obtain a high-precision three-dimensional temperature field on the diamond crystal growth surface.

[0011] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a multi-view 3D reconstruction method for the temperature field of diamond growth, including: S1. Simultaneously acquire multi-view two-dimensional infrared radiation images of the deposition stage and three-dimensional geometric information of the diamond crystal surface, wherein the deposition stage is loaded with diamond crystals. S2. Based on the multi-view two-dimensional infrared radiation image and the three-dimensional geometric information, perform coordinate system calibration; S3. Process the two-dimensional infrared radiation images from each viewpoint, and invert the two-dimensional temperature distribution map corresponding to each viewpoint based on the two-dimensional infrared radiation images from each viewpoint. S4. Based on the three-dimensional geometric information, reconstruct the three-dimensional surface model of the diamond crystal; S5. Match and map the coordinates of the three-dimensional surface model in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system to obtain fused multi-view temperature data, and generate a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model.

[0012] Optionally, in step S3, processing the two-dimensional infrared radiation images from each viewpoint and inverting the two-dimensional temperature distribution map corresponding to each viewpoint based on the two-dimensional infrared radiation images from each viewpoint, for the target viewpoint, the following steps are used to invert and obtain the two-dimensional temperature distribution map corresponding to the target viewpoint: Subtract the background radiation signal from the two-dimensional infrared radiation image from the target's perspective; An emissivity directionality compensation model is used to correct the radiation signal after background subtraction; The corrected radiation signal is inverted into temperature values ​​using the Planck inverse function to obtain a two-dimensional temperature distribution map from the target's perspective; The target viewpoint can be any one of the multiple viewpoints.

[0013] Optionally, the expression for the emissivity directionality compensation model is: ; in, It is the directional coefficient. Let λ be the directional emissivity of the material at a wavelength λ and an observation angle θ. In the direction of the normal The reference emissivity.

[0014] Optionally, S4, based on the three-dimensional geometric information, reconstructing the three-dimensional surface model of the diamond crystal, specifically includes: The point cloud data corresponding to the three-dimensional geometric information is reconstructed into a triangular mesh model using the Poisson surface reconstruction algorithm or the moving cube algorithm.

[0015] Optionally, S5, matching and mapping the coordinates of the three-dimensional surface model in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system to obtain fused multi-view temperature data, and generating a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model, specifically including: S51. For the target vertex in the triangular mesh model, match and map the coordinates of the target vertex in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system, and calculate the projected coordinates of the target vertex in each viewpoint; wherein, the triangular mesh model includes a set of vertices, and the target vertex is any vertex in the set of vertices; S52. Determine the visibility of the target vertex under each viewpoint, and for the visible viewpoints, interpolate the temperature value of the target vertex from its corresponding two-dimensional temperature distribution map. S53. Perform a weighted average of the temperature values ​​of the target vertex under the visible view to obtain the fused temperature value of the target vertex. S54. Based on the fusion temperature value of each vertex in the vertex set, generate a three-dimensional temperature field on the diamond crystal growth surface.

[0016] Optionally, the following formula can be used to determine whether the m-th vertex is visible from the ith infrared thermal imager camera's field of view: ; in, It is the surface unit normal vector of the m-th vertex. Let m be the direction vector from the m-th vertex to the i-th infrared thermal imager camera. If the dot product of the surface unit normal vector of the m-th vertex and the direction vector from the m-th vertex to the i-th infrared thermal imager camera is greater than 1, then... If the m-th vertex is visible from the i-th infrared thermal imager camera's field of view, then it is visible; otherwise, it is invisible.

[0017] Optionally, after S5, it also includes: S6. Optimize and complete the generated three-dimensional temperature field, specifically including: for vertices where temperature data is missing due to being invisible from all perspectives, complete the temperature data based on the heat conduction physical model, and smooth the completed three-dimensional temperature field.

[0018] Optionally, the temperature data completion based on the heat conduction physical model specifically includes: On the triangular mesh model, using vertices with known temperature values ​​as boundary conditions, the equations are solved. Temperature estimation is performed on vertices with missing temperatures. in, It is a partial differential equation defined on a continuous surface. Representing any continuous point on the three-dimensional surface model M, it is an abstract position variable. This represents the temperature value of the temperature field T at position v on the surface.

[0019] Optionally, the three-dimensional temperature field can be smoothed using the following expression: ; in, Let m be the set of optimal temperature values ​​for vertex m on the triangular mesh model. A collection of multiple perspectives; For perspective indexing; Let m be the set of vertices in the triangular mesh model, where m is the index of the vertex. This is the visibility factor for the m-th vertex in the i-th view. If the vertex is visible in the view, it is 1; if the vertex is not visible in the view, it is 0. Let m be the original temperature value of vertex m in the two-dimensional infrared radiation image acquired from the i-th viewpoint. E represents the pair of adjacent vertices in the triangular mesh model, connected by mesh edges; E is the set of mesh edges. Let λ be the temperature value of vertex n, which is adjacent to vertex m in the triangular mesh model, and λ be the weight of the smoothing term.

[0020] Secondly, this application provides a multi-view diamond growth temperature field 3D reconstruction system, including: The multi-view temperature measurement module has its sensing units distributed in different spatial orientations of the MPCVD deposition chamber, which is used to acquire multi-view two-dimensional infrared radiation images of the deposition stage carrying diamond crystals. A three-dimensional topography reconstruction module is used to acquire the three-dimensional geometric information of the diamond crystal surface; A synchronous triggering module is used to synchronously control the multi-view temperature measurement module and the three-dimensional shape reconstruction module to acquire data. The data processing module is connected to the multi-view temperature measurement module, the three-dimensional morphology reconstruction module, and the synchronous triggering module, respectively, and is used to process and fuse the synchronously acquired multi-view two-dimensional infrared radiation image and the three-dimensional geometric information to generate the three-dimensional temperature field of the diamond crystal growth surface.

[0021] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method and system for 3D reconstruction of the temperature field of diamond growth from multiple perspectives. By matching and fusing multi-view two-dimensional infrared radiation images with the three-dimensional morphological features of the diamond growth surface, the temperature value corresponding to the three-dimensional geometric information of the diamond crystal surface is obtained. This constructs a true and unambiguous three-dimensional temperature field of the diamond crystal growth surface, achieving a precise transition from two-dimensional temperature projection to a true three-dimensional temperature field, alleviating the technical problem in existing technologies that cannot accurately map temperature data to the specific location of the three-dimensional morphology. Simultaneously, by using multi-view two-dimensional infrared radiation images to observe the diamond crystal on the deposition stage from all angles, the temperature of the hidden surfaces (such as the sides) of the diamond crystal is obtained, thus constructing a complete three-dimensional temperature field of the diamond crystal growth surface. This alleviates the technical problem in existing technologies that cannot obtain the temperature of hidden surfaces (such as the sides), facilitating a deeper understanding and precise control of the thermodynamic behavior during MPCVD growth. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A schematic diagram of the functional modules of a multi-view diamond growth temperature field 3D reconstruction system provided in this application; Figure 2 A flowchart illustrating a multi-view 3D reconstruction method for diamond growth temperature field provided in this application; Figure 3 for Figure 2 A schematic diagram of the process for reconstructing a three-dimensional temperature field from multi-view two-dimensional data; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. To make the objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] This application provides a multi-view diamond growth temperature field 3D reconstruction system, including a multi-view temperature measurement module, a three-dimensional morphology reconstruction module, a synchronous triggering module, and a data processing module. Among them: The multi-view temperature measurement module, with its sensing units distributed in different spatial orientations within the MPCVD deposition chamber, is used to acquire multi-view two-dimensional infrared radiation images of the deposition stage carrying diamond crystals.

[0026] The multi-view temperature measurement module includes multiple infrared thermal imagers, which are equally spaced around the outside of the MPCVD deposition chamber at different depression and azimuth angles. Their optical axes pass through corresponding optical windows pointing towards the deposition stage. The optical windows are multiple high-temperature resistant germanium windows (Ge, coated with an anti-reflection film) to ensure infrared transmittance >90% at each viewing angle.

[0027] Specifically, 3-6 medium-wave infrared thermal imagers (MWIR) with a diameter of 3-5 μm are arranged around the MPCVD deposition chamber to observe the deposition stage at different angles. This application allows for observation of the deposition stage at equal intervals with a depression angle of 30-60° and an azimuth angle, aiming to simultaneously capture the top and side surfaces of the diamond crystal to solve the problem of three-dimensional temperature measurement of the crystal "growth mound" in MPCVD. An exemplary layout parameter for multiple infrared thermal imagers is: 4 thermal imagers, arranged at depression angles of 45°, and azimuth angles of 0°, 90°, 180°, and 270° respectively, with a field-of-view overlap rate of >50% between adjacent thermal imagers, an infrared camera resolution of 640×512 pixels, a NETD ≤20mK, and a three-dimensional topographic accuracy of ±0.05mm.

[0028] A synchronization trigger module is used to synchronously control the multi-view temperature measurement module and the three-dimensional morphology reconstruction module to acquire data. In this application, the synchronization trigger module is a synchronization trigger, because diamond growth is a dynamic process; the hardware-level synchronization trigger ensures that the morphology and temperature data reflect the growth state at the same instant.

[0029] A three-dimensional topography reconstruction module is used to acquire the three-dimensional geometric information of the diamond crystal surface. The three-dimensional topography reconstruction module in this application can be a three-dimensional topography reconstruction system, which acquires high-precision dense three-dimensional point cloud P3D by projecting coded structured light patterns and combining multi-view image acquisition and calculation. Specifically, the module includes the following units and processing steps: System calibration unit: Before the measurement begins, multiple industrial cameras and a digital projector arranged around the deposition chamber are calibrated using a high-temperature checkerboard calibration plate. At least two industrial cameras are used, forming a binocular or multi-view vision system. The calibration includes: The intrinsic parameters of each camera and the extrinsic parameters between cameras are calibrated to establish a spatial coordinate system. The intrinsic parameters include the camera matrix K and distortion coefficients D, and the extrinsic parameters include the rotation matrix R and the translation vector t.

[0030] Simultaneously, the reference plane equation A0x+B0y+C0z+D0=0 is fitted to the calibration plate image to provide a height reference for subsequent measurements. Where A... 0、 B 0、 C 0、 D0 is the coefficient of the plane equation.

[0031] Anti-interference image acquisition unit: Under the precise control of the synchronization trigger module, the projector and camera are driven to execute a specific acquisition sequence to overcome the interference of background radiation in high-temperature environments. Background image acquisition: First, the projector light source was turned off, and each camera simultaneously acquired two-dimensional images of the current deposition stage and diamond crystal under ambient light and high-temperature radiation, denoted as background image I. bg .

[0032] Acquiring coded structured light patterns: Subsequently, the projector sequentially projects preset coded structured light patterns, and each camera synchronously acquires the deformed patterns modulated by the crystal surface. The coded structured light patterns include a four-step phase-shifting sinusoidal fringe pattern I for high-precision positioning. φ0 3 and multiple Gray code diagrams for eliminating phase periodic ambiguity I gray .

[0033] Phase decoding unit: Processes the acquired image, calculating its absolute phase value pixel by pixel. This process includes: Background subtraction: Subtracting the background image pixel by pixel from the captured raw encoded image to eliminate the effects of high-temperature radiation and ambient light, resulting in a pure projector response light intensity I. clean (x,y)=I raw (x,y) I bg (x,y). Where I raw (x,y) represents the grayscale value at coordinates (x,y) in the original image. clean (x,y) represents the grayscale value at coordinates (x,y) in the clean image. bg (x,y) represents the grayscale value of the coordinates (x,y) in the background image.

[0034] Wrap-up phase calculation: Using the four-step phase shift map after background subtraction, the wrap-up phase is calculated through the arctangent function. .

[0035] in, To wrap the phase, , , , This is a detailed representation of a four-step phase shift diagram.

[0036] Phase unwrapping: Decode the captured Gray code image to obtain the stripe order k(x,y) corresponding to each pixel.

[0037] Absolute phase acquisition: By combining the wrapping phase and the fringe order, a unique absolute phase map is calculated. : .

[0038] 3D Reconstruction Unit: Based on the principles of multi-view stereo vision and absolute phase maps, it generates dense 3D point clouds. Phase-consistent stereo matching: Utilizing the uniqueness of absolute phase values, stereo matching is performed between camera images from different viewpoints. For any pixel in the left camera image, based on epipolar constraints, a search is conducted in the right camera image for its corresponding absolute phase value. The corresponding pixels that are equal or have the smallest error. This process establishes the projection correspondence of the same point on the crystal surface onto the imaging planes of different cameras.

[0039] Triangulation calculation: based on the established set of matching point pairs Based on the camera intrinsic and extrinsic parameters obtained by the system calibration unit, triangulation equations are constructed. By solving the following overdetermined system of equations, the three-dimensional coordinates (X, Y, Z) of each matching point in the world coordinate system are obtained: .

[0040] in, , Let u be the pixel coordinates of the point on the image, where u is the horizontal direction and v is the vertical direction. As a scale factor, Let be the intrinsic parameter matrix of the i-th camera. Let be the rotation matrix for the i-th camera. Let be the translation vector of the i-th camera.

[0041] Dense point cloud generation: Traverse all valid pixels in the image, repeat the above matching and calculation process, and finally generate dense three-dimensional point cloud data P3D covering the entire surface of the diamond crystal.

[0042] The data processing module is connected to the multi-view temperature measurement module, the three-dimensional morphology reconstruction module, and the synchronous triggering module, respectively, and is used to process and fuse the synchronously acquired multi-view two-dimensional infrared radiation image and the three-dimensional geometric information to generate the three-dimensional temperature field of the diamond crystal growth surface.

[0043] Through the synergistic operation of the aforementioned multi-view temperature measurement module, 3D morphology reconstruction module, synchronous triggering module, and data processing module, multi-view 2D temperature data is accurately mapped onto the 3D morphology of the diamond crystal, enabling the acquisition of the true temperature value of any point on the crystal surface in 3D space. This ensures that the output of this application is no longer a distorted 2D temperature image, but rather the true temperature value corresponding to every 3D spatial point on the diamond crystal growth surface. This represents a qualitative leap from approximate 2D temperature measurement to precise 3D perception, and completely solves the information confusion and distortion problems caused by traditional single-view infrared thermal imaging compressing 3D temperature information into a 2D plane. By deploying sensing units at different spatial orientations, the diamond crystal is observed without blind spots, collecting temperature data from all visible areas of the crystal's fully enclosed surface, including hidden areas that are long-term "black box" states (such as the sides). This achieves a breakthrough from local observation to obtaining complete temperature data for the entire surface. Through multi-view spatial collaborative observation, the inherent blind spots of single-view observation are completely eliminated, allowing simultaneous acquisition of temperature data from the top surface, sides, and multiple directions of the crystal. After data processing and fusion, a complete three-dimensional temperature field model covering the entire enclosed surface of the crystal can be generated, solving the problem that existing technologies cannot assess the thermal state of hidden areas. For example, for a diamond with a diameter of 10 mm, the system can reconstruct the temperature distribution over 95% of its surface area.

[0044] This application also provides a method for 3D reconstruction of the diamond growth temperature field from multiple perspectives, the method comprising: S1. Simultaneously acquire multi-view two-dimensional infrared radiation images of the deposition stage and three-dimensional geometric information of the diamond crystal surface, wherein the deposition stage is loaded with diamond crystals.

[0045] S2. Based on the multi-view two-dimensional infrared radiation image and the three-dimensional geometric information, perform coordinate system calibration.

[0046] S3. Process the two-dimensional infrared radiation images from each viewpoint, and invert the two-dimensional temperature distribution map corresponding to each viewpoint based on the two-dimensional infrared radiation images from each viewpoint.

[0047] S4. Based on the three-dimensional geometric information, reconstruct the three-dimensional surface model of the diamond crystal.

[0048] S5. Match and map the coordinates of the three-dimensional surface model in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system to obtain fused multi-view temperature data, and generate a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model.

[0049] This application proposes a multi-view diamond growth temperature field 3D reconstruction method, designed specifically for the characteristics of the MPCVD process environment. By simultaneously acquiring two-dimensional temperature images using multiple infrared thermal imagers placed at specific locations around the MPCVD deposition chamber, the two-dimensional temperature images are accurately mapped and fused onto the three-dimensional surface model of the diamond crystal, thereby obtaining the true temperature value corresponding to each three-dimensional surface point and constructing a true and unambiguous three-dimensional temperature field.

[0050] Specifically, step S2 of this application, which involves calibrating the coordinate system based on the multi-view two-dimensional infrared radiation image and the three-dimensional geometric information, includes: S21. Prepare the calibration device: Use a high-temperature infrared calibration plate with a checkerboard pattern of known emissivity on its surface. It can be heated to multiple different temperatures so that feature points can be presented in both infrared radiation images and three-dimensional geometric information.

[0051] S22. Synchronous data acquisition: Under the same conditions, the two-dimensional infrared images of the calibration plate by each infrared thermal imager are acquired synchronously, and the three-dimensional geometric information of the calibration plate surface is obtained by the three-dimensional shape reconstruction module.

[0052] S23. Solving for the extrinsic parameters of the infrared thermal imager camera: For each infrared thermal imager, based on the known 3D coordinates of the checkerboard corner points in the 2D infrared image of the calibration board in the world coordinate system, the pose relationship of the infrared thermal imager camera relative to the unified world coordinate system, i.e., the extrinsic parameter matrix, is calculated using the PnP (Perspective-n-Point) algorithm. ; in, Let be a 3×3 rotation matrix, describing the rotation relationship from the world coordinate system to the i-th infrared thermal imager camera coordinate system. Let be a 3×1 translation vector, describing the translation from the origin of the world coordinate system to the origin of the i-th infrared thermal imager camera coordinate system. This represents the transformation matrix that transforms the coordinates of a corner point from the world coordinate system to the coordinate system of the i-th infrared thermal imager camera. This application includes a world coordinate system W with the center of the deposition stage as the origin, and coordinate systems of each infrared thermal imager camera. The three-dimensional topography coordinate system S in the three-dimensional topography reconstruction module.

[0053] S24. Determine the coordinate transformation relationship between the 3D topography reconstruction module and the infrared thermal imager camera: ; in, Let j be the coordinates of the j-th calibration point in the three-dimensional topographic coordinate system. Let j be the coordinates of the j-th calibration point in the world coordinate system. This represents the transformation matrix that transforms the coordinates of a corner point from the three-dimensional topographic coordinate system to the world coordinate system.

[0054] This application introduces a high-temperature infrared calibration plate as a common benchmark and utilizes the PnP algorithm and feature point matching to accurately solve the transformation relationship between each infrared thermal imager camera and the 3D topography reconstruction module relative to a unified world coordinate system. This fundamentally solves the spatial matching problem between infrared 2D radiation data and 3D geometric information, ensuring that temperature data can be mapped to the corresponding surface position of the 3D surface model with sub-pixel accuracy, laying a solid foundation for subsequent accurate temperature field reconstruction and analysis.

[0055] In step S3 of this application, processing two-dimensional infrared radiation images from various viewpoints and inverting the two-dimensional temperature distribution map corresponding to each viewpoint based on the two-dimensional infrared radiation images from various viewpoints, the following steps are used to invert the two-dimensional temperature distribution map corresponding to the target viewpoint: The background radiation signal in the two-dimensional infrared radiation image from the target's perspective is subtracted.

[0056] An emissivity directionality compensation model is used to correct the radiation signal after background subtraction.

[0057] The corrected radiation signal is inverted into temperature values ​​using the Planck inverse function, resulting in a two-dimensional temperature distribution map from the target's perspective.

[0058] The target viewpoint can be any one of the multiple viewpoints.

[0059] Specifically, a synchronous triggering module controls multiple infrared thermal imagers to simultaneously acquire two-dimensional infrared radiation images at time t. For the i-th infrared thermal imager, the following formula is used to evaluate the two-dimensional infrared radiation image acquired by the i-th infrared thermal imager at time t. Temperature inversion, radiation calibration, and emissivity correction are performed to obtain a two-dimensional temperature distribution map. x and y are the coordinates in the two-dimensional infrared radiation image.

[0060] ; in, It is Planck's inverse function. For the diamond crystal surface at a viewing angle Emissivity in direction, This is an algorithm term specifically used to subtract the plasma background radiation within the MPCVD deposition chamber. Let i be the two-dimensional temperature distribution map of the i-th infrared thermal imager, where i is the index of the infrared thermal imager and also the index of the viewing angle.

[0061] Step S3 above achieves interference removal and high-precision inversion of multi-view infrared temperature measurement signals in MPCVD plasma environment: it solves the problem of interference to multi-view infrared temperature measurement signals under strong plasma background radiation. Through a special background radiation subtraction model and emissivity directionality correction, it provides an accurate and consistent data source for three-dimensional fusion. This not only improves the accuracy and reliability of the final temperature data, but also significantly enhances the ability of this application to cope with complex targets and environments, providing an indispensable high-quality data guarantee for generating a reliable three-dimensional temperature field.

[0062] The expression for the emissivity directionality compensation model in step S3 is as follows: ; in, The directional coefficient is determined through calibration experiments; Let be the directional emissivity of the material at a wavelength λ and an observation angle θ (usually the angle between the observation direction and the surface normal). In the normal direction (usually 0°) The reference emissivity. This represents a quadratic relationship between directional emissivity and the deviation of the observation angle from the normal direction. This is a commonly used empirical model for the variation of emissivity of non-Lambertian surfaces with angle.

[0063] Diamond crystal surfaces are not ideal diffuse reflectors (Lambertian sources); their emissivity exhibits significant directionality. This application introduces an emissivity directionality compensation model to ensure the consistency and comparability of temperature values ​​retrieved when observing the same surface region from different angles. This is also an absolute prerequisite for reliable fusion of multi-view data. Compared to general infrared thermometry, which often assumes emissivity to be constant or only wavelength-dependent, this application models the characteristics of diamond growth surfaces, fundamentally improving the accuracy of raw temperature data from each viewpoint and laying a solid foundation for subsequent high-quality 3D reconstruction. The emissivity directionality compensation model is a deep adaptation to the specific application scenario and materials of MPCVD diamond growth, and is one of the significant technical features distinguishing this application from general 3D thermometry methods. The emissivity of the diamond surface varies with the observation angle. Significant changes necessitate model correction, which is a prerequisite for obtaining consistent and comparable temperature data from multiple perspectives—a point often overlooked in general object temperature measurement.

[0064] This application achieves a breakthrough in acquiring complete temperature data of the entire surface from local observation: by simultaneously observing with multiple thermal imagers arranged in different spatial orientations, blind spots are eliminated, temperature data of all visible areas of the fully enclosed surface of the crystal are collected, and through algorithm optimization and completion based on physical laws, a complete and enclosed three-dimensional temperature distribution model covering the entire crystal growth surface is finally generated.

[0065] S4 of this application involves reconstructing a three-dimensional surface model of the diamond crystal based on the aforementioned three-dimensional geometric information, specifically including: The point cloud data corresponding to the three-dimensional geometric information is reconstructed into a triangular mesh model using the Poisson surface reconstruction algorithm or the moving cube algorithm.

[0066] Specifically, this application acquires two-dimensional infrared radiation images using an infrared thermal imager, while simultaneously using a three-dimensional topography reconstruction module to obtain a three-dimensional point cloud of the diamond crystal. k is the point cloud index, and a triangular mesh model is generated from the point cloud using Poisson surface reconstruction or Marching Cubes algorithms. ;in, For the set of vertices, The index of the vertex. Given a set of triangular facets, calculate the surface normal vector for each vertex. .

[0067] This application utilizes either the Poisson surface reconstruction algorithm or the moving cube algorithm to efficiently and accurately convert discrete 3D point cloud data into a continuous triangular mesh model. This process completely recovers the continuous geometric morphology of the measured diamond crystal surface, including its complex curved surfaces and edges, providing an accurate spatial representation of the temperature field. Specifically, the Poisson surface reconstruction algorithm effectively handles noise and generates a watertight, smooth surface, suitable for complex, smooth objects; the moving cube algorithm is classic and efficient, clearly presenting isosurface details. Both algorithms can complete the reconstruction task from point cloud to surface with high quality, ensuring the geometric accuracy of the 3D model.

[0068] In S5 of this application, the coordinates of the three-dimensional surface model in the coordinate system are matched and mapped with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system to obtain fused multi-view temperature data, and a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model is generated, specifically including: S51. For the target vertex in the triangular mesh model, match and map the coordinates of the target vertex in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system, and calculate the projected coordinates of the target vertex in each viewpoint; wherein, the triangular mesh model includes a set of vertices, and the target vertex is any vertex in the set of vertices.

[0069] S52. Determine the visibility of the target vertex under each viewpoint, and for the visible viewpoints, interpolate the temperature value of the target vertex from the corresponding two-dimensional temperature distribution map.

[0070] S53. Perform a weighted average of the temperature values ​​of the target vertex under the visible view to obtain the fused temperature value of the target vertex.

[0071] S54. Based on the fusion temperature value of each vertex in the vertex set, generate a three-dimensional temperature field on the diamond crystal growth surface.

[0072] This application's S4, by calculating the projected coordinates of each vertex under all viewing angles and determining its visibility, can identify which infrared cameras can observe each surface point, effectively solving the occlusion problem existing under a single viewing angle. This ensures that even on the surface of a complex diamond crystal growth cavity, every region can be covered by at least one viewing angle, thus generating a complete, blind-spot-free three-dimensional temperature field. Furthermore, it uses a weighted average fusion of temperature values ​​from multiple visible viewing angles of the same vertex, rather than simply using a single data point. This strategy significantly improves the reliability and accuracy of the temperature data: on the one hand, it smooths out random noise that may exist in single-view measurements; on the other hand, by fusing multi-angle observation information, it reduces excessive reliance on the absolute accuracy of individual camera calibration or measurement, enhancing the overall robustness of the system.

[0073] Specifically, S5 first performs a projection transformation: for the m-th vertex Calculate its projected coordinates in the two-dimensional infrared radiation image acquired from the i-th infrared thermal imager camera's viewpoint: ; in, Let be the intrinsic parameter matrix of the i-th infrared thermal imager camera.

[0074] Secondly, visibility is determined: the Z-buffer algorithm is used to determine the visibility of a vertex from different viewpoints. The m-th vertex... Conditions under which the i-th infrared thermal imager camera is visible from its field of view: ; in, It is the m-th vertex The surface unit normal vector, Let m be the direction vector from the m-th vertex to the i-th infrared thermal imager camera. If the dot product of the surface unit normal vector of the m-th vertex and the direction vector from the m-th vertex to the i-th infrared thermal imager camera is greater than 1, then... If the m-th vertex is visible from the i-th infrared thermal imager camera's field of view, then it is visible; otherwise, it is invisible.

[0075] Furthermore, temperature interpolation is performed: for visible vertices, from their projected coordinates... From bilinear interpolation Obtain the temperature value of vertex m. .

[0076] Finally, multi-view temperature fusion is performed: for each vertex, the temperature values ​​from all visible views are fused. ; in, To obtain the fusion temperature value of the m-th vertex, As a collection of multiple perspectives, For perspective indexing, Let be the weight of the m-th vertex. And the weight of the m-th vertex is considered as follows: 1. Viewpoint quality weight: The smaller the angle between the viewpoint and the normal vector, the greater the weight. ; 2. Distance weight: The closer to the camera, the greater the weight. ; 3. Signal-to-noise ratio weight: Adjusted based on the local signal-to-noise ratio of the image. Final weights: ; in, As for perspective weight, Distance weights Let i be the coordinate system of the i-th infrared thermal imager camera. For signal-to-noise ratio weights.

[0077] The final weight in this application is the combined product of three dimensions: normal angle, distance, and signal-to-noise ratio. This means that for a given vertex, this application does not simply take the average of multiple viewpoints, but rather simultaneously pursues the most accurate observation angle, the highest image resolution, and the clearest signal. This allows the fusion process to fully utilize the complementary information from multiple viewpoints, achieving a measurement quality that cannot be achieved by a single viewpoint or simple averaging.

[0078] Following S5, this application also includes: S6, optimizing and completing the generated three-dimensional temperature field, specifically including: completing the temperature data of vertices that are missing due to being invisible from all perspectives, based on the heat conduction physical model, and smoothing the completed three-dimensional temperature field.

[0079] For components with extremely complex structures, such as the interior of diamond crystal growth cavities, deep holes, or the backs of densely packed blades, there may still be certain surface areas that are absolute blind spots in the field of view of all cameras. Step S5 uses a thermal conductivity physics model to complete the data and intelligently infers the temperature of these "black hole" regions, thereby generating a physically complete three-dimensional temperature field. Furthermore, the use of a thermal conductivity physics model for completion ensures that the completed temperature values ​​not only consider the spatial location of surrounding known points but also adhere to the fundamental physical laws of energy conservation and heat diffusion. For example, it reasonably infers the heat flow influence from high-temperature regions to adjacent low-temperature regions, making the completed temperature field smoother and more continuous in its overall distribution, and consistent with thermodynamic principles. Its reliability is far higher than that of purely mathematical interpolation results.

[0080] Specifically, for invisible vertices (invisible from all views), diffusion-based completion based on the heat conduction equation is used. This method, based on the physical laws of heat conduction within the crystal, is more consistent with the physical reality and accuracy of the diamond growth thermal field than simple geometric interpolation (which may be used in comparison files). The temperature data completion based on the heat conduction physical model is as follows: On the triangular mesh model, using vertices with known temperature values ​​as boundary conditions, the equations are solved. Temperature estimation is performed on vertices with missing temperatures. in, It is a partial differential equation defined on a continuous surface. Representing any continuous point on the three-dimensional surface model M, it is an abstract position variable. This represents the function value of the temperature field T at the surface location v, i.e., the temperature value.

[0081] Among them, smoothing optimization: noise is smoothed using Laplacian smoothing. ; in, As vertices The set of neighboring vertices.

[0082] The smoothing optimization process described above effectively suppresses minor errors in multi-view data registration, noise in individual images, or discontinuities that may occur at fusion boundaries. This improves the visual quality and numerical stability of the temperature field, making isotherms clearer and temperature gradients more reasonable, facilitating direct observation and qualitative analysis.

[0083] This application uses the following expression to smooth and optimize the three-dimensional temperature field: ; in, The set of optimal temperature values ​​for vertex m on a triangular mesh model. A collection of multiple perspectives; For perspective indexing; Let m be the set of vertices in the triangular mesh model, where m is the index of the vertex. This is the visibility factor for the m-th vertex in the i-th view. If the vertex is visible in the view, it is 1; if the vertex is not visible in the view, it is 0. Let m be the temperature value of vertex m in a two-dimensional infrared radiation image acquired from the i-th viewpoint. E represents the pair of adjacent vertices in the triangular mesh model, connected by mesh edges; E is the set of mesh edges. Let λ be the temperature value of vertex n, which is adjacent to vertex m in the triangular mesh model, and λ be the weight of the smoothing term.

[0084] This expression defines a global optimization objective that solves for the optimal temperature value of all vertices by minimizing the objective function. The temperature difference between adjacent vertices is penalized, and λ is the weight of the smoothing term.

[0085] First item: Data fidelity item The significance of this item is to obtain the final three-dimensional temperature. It aims to get as close as possible to the values ​​observed by all the cameras that can "see" it. This forces the optimization results to stay within the original measurement data.

[0086] The second term: smoothing prior terms. The significance of this item is: based on the common sense of physics (temperature fields are usually continuous and smooth in space), constraints are introduced to avoid unreasonable drastic fluctuations or noise in the results, while also providing reasonable interpolation guidance in sparse data regions.

[0087] Overall goal: To find a set of temperature values ​​{ It is faithful to all multi-view observation data and satisfies the physical laws of spatial smoothness.

[0088] Following S6, this application also includes: S7, three-dimensional temperature field analysis and visualization: Temperature gradient calculation: Calculate the surface temperature gradient field. ; in, For containing vertices A collection of dough pieces, For the area of ​​the patch, This is a constant gradient calculated on a single triangular facet f. This formula is a standard and robust method for calculating continuous physical fields (gradients) from discrete triangular meshes, enabling high-precision quantification of key physical quantities. For the first time, it achieves accurate calculation of the three-dimensional temperature gradient vector at every point on the MPCVD diamond growth surface, accurately identifying critical areas of maximum gradient or assessing the uniformity of the gradient across the entire surface.

[0089] The temperature gradient at vertex m is calculated by considering the gradients of all surrounding triangular faces. The gradient was obtained by area-weighted averaging. Larger areas contribute more to the vertex gradient. This achieves a transformation from temperature to a key thermodynamic quantity: the temperature gradient is the most direct and important physical quantity for analyzing thermal stress, heat flow, and the formation and evolution of crystal defects (such as dislocations). This application can accurately calculate the three-dimensional gradient vector at every point on the MPCVD diamond growth surface and provides quantitative process evaluation indicators: the uniformity, magnitude, and direction of the gradient can be directly correlated with the growth quality.

[0090] Isothermal surface extraction: Isothermal surfaces at specific temperatures are extracted for analyzing temperature distribution characteristics. By analyzing the shape, spacing, and topology of the isothermal surfaces, heating uniformity, thermal field symmetry, and local anomaly detection can be immediately determined.

[0091] Heat flux density estimation: Estimate the surface heat flux distribution by combining the thermal conductivity coefficient.

[0092] This application provides a foundation for calculating key physical quantities to support in-depth process analysis: based on a high-resolution three-dimensional temperature field, it can accurately calculate the three-dimensional temperature gradient vector field and local heat flux density distribution on the crystal surface for the first time, and supports the extraction of arbitrary cross-sections and isothermal surfaces for analysis. This provides an unprecedented quantitative tool for in-depth analysis of heat transport, thermal stress distribution, and defect formation mechanisms during the growth process. Simulation and verification show that, under typical growth conditions, the surface normal temperature gradient calculated in this application has an accuracy better than 10°C / mm, providing direct and reliable data input for quantitative analysis of thermal stress distribution and prediction of dislocation multiplication and crack initiation, elevating process analysis from qualitative observation to quantitative physical modeling.

[0093] This application can also process continuous time series. Reconstructing the three-dimensional temperature field sequence This study analyzes the spatiotemporal evolution of the temperature field. By providing real-time or near-real-time output of the three-dimensional temperature field spatiotemporal evolution sequence, it offers high-dimensional, accurate temperature big data with strong correlation to spatial morphology for the closed-loop control of the MPCVD growth process. This lays a crucial sensing foundation for realizing process reverse design based on three-dimensional thermal field simulation and defect early warning and adaptive control based on thermal field anomalies.

[0094] This application constructs a high-dimensional data engine for intelligent process control: the three-dimensional temperature field spatiotemporal sequence output by this application, such as a 1Hz update rate, constitutes a high-dimensional data source characterizing the evolution of the thermodynamic state during the growth process. By extracting feature parameters (such as the movement trajectory of the high-temperature zone and the evolution of gradient field uniformity), it can be directly linked with process parameters (microwave power, gas pressure), providing the possibility for achieving adaptive control based on three-dimensional thermal field feedback. This lays an indispensable perceptual foundation for the intelligent upgrade of MPCVD process from "parameter trial and error" to "model-driven" intelligent control.

[0095] One implementation example of this application is: integrating with a three-dimensional topography reconstruction system, directly utilizing the three-dimensional topography data output by the relevant technology "three-dimensional topography reconstruction system", and adding two infrared thermal imagers (cost optimization scheme), ensuring the correspondence between topography and temperature data through time synchronization, with reconstruction accuracy: temperature ±8°C, space ±0.1mm, generating a three-dimensional temperature field on the diamond crystal growth surface.

[0096] This application advances diamond growth temperature monitoring from the era of "two-dimensional images" to the era of "three-dimensional physical fields" through multi-view fusion and three-dimensional reconstruction. The accurate, complete, and high-dimensional temperature data it generates has transformative significance for deepening the understanding of growth mechanisms and achieving precise process control.

[0097] like Figure 1 As shown, the synchronous triggering module and data processing module of the multi-view diamond growth temperature field 3D reconstruction system are located at the top, responsible for synchronous control and calculation. The middle section contains the deposition chamber, the diamond crystal, and multiple infrared thermal imagers and 3D morphology reconstruction modules deployed around it. The bottom section is a visualization module, displaying the final generated 3D temperature field and derived analysis results. Solid arrows indicate control and data flow, while dashed arrows indicate the observation optical path. This layout effectively integrates information, avoids excessive horizontal length, and ensures the readability of the text in the figure.

[0098] like Figure 2 The diagram shows a flowchart of a multi-view diamond growth temperature field 3D reconstruction method.

[0099] like Figure 3 As shown, it explains Figure 2This diagram illustrates the process of reconstructing a 3D temperature field from multi-view 2D data. The left side shows 2D infrared radiation images from different viewpoints. The triangular mesh model at the top center provides the spatial reference for reconstruction. The core processing flow consists of four steps: 1. Coordinate mapping and projection: Projecting the 3D mesh vertices onto the respective 2D image planes; 2. Visibility determination and temperature interpolation: Determining whether a vertex is visible from each viewpoint and interpolating the temperature from the image; 3. Multi-view weighted fusion: For each vertex, fusing the temperature values ​​from all visible viewpoints, considering factors such as viewpoint angle and distance; 4. Optimization and completion: Performing diffusion completion on missing data and smoothing optimization. The final output is a 3D temperature field as shown on the right, which can be visually displayed using color rendering and can be further used to calculate higher-order physical quantities such as temperature gradient fields and isothermal surfaces.

[0100] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores multi-view two-dimensional infrared radiation images and three-dimensional geometric information. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a multi-view 3D reconstruction method for diamond growth temperature fields.

[0101] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for 3D reconstruction of the temperature field of diamond growth from multiple perspectives, characterized in that, The method includes: S1. Simultaneously acquire multi-view two-dimensional infrared radiation images of the deposition stage and three-dimensional geometric information of the diamond crystal surface, wherein the deposition stage is loaded with diamond crystals. S2. Based on the multi-view two-dimensional infrared radiation image and the three-dimensional geometric information, perform coordinate system calibration; S3. Process the two-dimensional infrared radiation images from each viewpoint, and invert the two-dimensional temperature distribution map corresponding to each viewpoint based on the two-dimensional infrared radiation images from each viewpoint. S4. Based on the three-dimensional geometric information, reconstruct the three-dimensional surface model of the diamond crystal; S5. Match and map the coordinates of the three-dimensional surface model in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system to obtain fused multi-view temperature data, and generate a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model.

2. The multi-view diamond growth temperature field 3D reconstruction method according to claim 1, characterized in that, S3. In the step of processing the two-dimensional infrared radiation images from each viewpoint and inverting the two-dimensional temperature distribution map corresponding to each viewpoint based on the two-dimensional infrared radiation images from each viewpoint, for the target viewpoint, the following steps are used to invert and obtain the two-dimensional temperature distribution map corresponding to the target viewpoint: Subtract the background radiation signal from the two-dimensional infrared radiation image from the target's perspective; An emissivity directionality compensation model is used to correct the radiation signal after background subtraction; The corrected radiation signal is inverted into temperature values ​​using the Planck inverse function to obtain a two-dimensional temperature distribution map from the target's perspective; The target viewpoint can be any one of the multiple viewpoints.

3. The multi-view diamond growth temperature field 3D reconstruction method according to claim 2, characterized in that, The expression for the emissivity directionality compensation model is: ； in, It is the directional coefficient. Let λ be the directional emissivity of the material at a wavelength λ and an observation angle θ. In the direction of the normal The reference emissivity.

4. The multi-view diamond growth temperature field 3D reconstruction method according to claim 1, characterized in that, S4. Based on the aforementioned three-dimensional geometric information, reconstruct the three-dimensional surface model of the diamond crystal, specifically including: The point cloud data corresponding to the three-dimensional geometric information is reconstructed into a triangular mesh model using the Poisson surface reconstruction algorithm or the moving cube algorithm.

5. The multi-view diamond growth temperature field 3D reconstruction method according to claim 1, characterized in that, S5. Match and map the coordinates of the three-dimensional surface model in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system to obtain fused multi-view temperature data, and generate a three-dimensional temperature field of the diamond crystal growth surface associated with the three-dimensional surface model, specifically including: S51. For the target vertex in the triangular mesh model, match and map the coordinates of the target vertex in the coordinate system with the coordinates of the two-dimensional temperature distribution map corresponding to each viewpoint in the coordinate system, and calculate the projected coordinates of the target vertex in each viewpoint; wherein, the triangular mesh model includes a set of vertices, and the target vertex is any vertex in the set of vertices; S52. Determine the visibility of the target vertex under each viewpoint, and for the visible viewpoints, interpolate the temperature value of the target vertex from its corresponding two-dimensional temperature distribution map. S53. Perform a weighted average of the temperature values ​​of the target vertex under the visible view to obtain the fused temperature value of the target vertex. S54. Based on the fusion temperature value of each vertex in the vertex set, generate a three-dimensional temperature field on the diamond crystal growth surface.

6. The multi-view diamond growth temperature field 3D reconstruction method according to claim 5, characterized in that, The following formula is used to determine whether the m-th vertex is visible from the ith infrared thermal imager camera's field of view: ; in, It is the surface unit normal vector of the m-th vertex. Let m be the direction vector from the m-th vertex to the i-th infrared thermal imager camera. If the dot product of the surface unit normal vector of the m-th vertex and the direction vector from the m-th vertex to the i-th infrared thermal imager camera is greater than 1, then... If the m-th vertex is visible from the i-th infrared thermal imager camera's field of view, then it is visible; otherwise, it is invisible.

7. The multi-view diamond growth temperature field 3D reconstruction method according to claim 5, characterized in that, Following S5, it also includes: S6. Optimize and complete the generated three-dimensional temperature field, specifically including: for vertices where temperature data is missing due to being invisible from all perspectives, complete the temperature data based on the heat conduction physical model, and smooth the completed three-dimensional temperature field.

8. The multi-view diamond growth temperature field 3D reconstruction method according to claim 7, characterized in that, The temperature data completion based on the heat conduction physical model is as follows: On the triangular mesh model, using vertices with known temperature values ​​as boundary conditions, the equations are solved. Temperature estimation is performed on vertices with missing temperatures. in, It is a partial differential equation defined on a continuous surface. Representing any continuous point on the three-dimensional surface model M, it is an abstract position variable. This represents the temperature value of the temperature field T at position v on the surface.

9. The multi-view diamond growth temperature field 3D reconstruction method according to claim 7, characterized in that, The following expression is used to smooth and optimize the three-dimensional temperature field: ； in, Let m be the set of optimal temperature values ​​for vertex m on the triangular mesh model. A collection of multiple perspectives; For perspective indexing; Let m be the set of vertices in the triangular mesh model, where m is the index of the vertex. This is the visibility factor for the m-th vertex in the i-th view. If the vertex is visible in the view, it is 1; if the vertex is not visible in the view, it is 0. Let m be the original temperature value of vertex m in the two-dimensional infrared radiation image acquired from the i-th viewpoint. E represents the pair of adjacent vertices in the triangular mesh model, connected by mesh edges; E is the set of mesh edges. Let λ be the temperature value of vertex n, which is adjacent to vertex m in the triangular mesh model, and λ be the weight of the smoothing term.

10. A multi-view diamond growth temperature field 3D reconstruction system, using the multi-view diamond growth temperature field 3D reconstruction method according to any one of claims 1-9, comprising: The multi-view temperature measurement module has its sensing units distributed in different spatial orientations of the MPCVD deposition chamber, which is used to acquire multi-view two-dimensional infrared radiation images of the deposition stage carrying diamond crystals. A three-dimensional topography reconstruction module is used to acquire the three-dimensional geometric information of the diamond crystal surface; A synchronous triggering module is used to synchronously control the multi-view temperature measurement module and the three-dimensional shape reconstruction module to acquire data. The data processing module is connected to the multi-view temperature measurement module, the three-dimensional morphology reconstruction module, and the synchronous triggering module, respectively, and is used to process and fuse the synchronously acquired multi-view two-dimensional infrared radiation image and the three-dimensional geometric information to generate the three-dimensional temperature field of the diamond crystal growth surface.

Images