Method and system for predicting surface deformation of dark-red enameled pottery blank based on three-dimensional scanning and neural network

Abstract

The invention relates to a dark-red enameled pottery blank surface deformation prediction method and system based on three-dimensional scanning and a neural network. The method comprises the following steps: performing three-dimensional scanning on a dark-red enameled pottery blank to obtain blank surface point cloud data; determining geometric features of the surface of the blank; collecting a temperature field and a humidity field of the dark-red enameled pottery green body, and obtaining a pug matching vector; carrying out deformation field prediction through a condition generation type deformation network, outputting a deformation increment, and calculating a deformation danger score; and calculating a regional danger index, and generating a real-time thermodynamic diagram. By collecting parameters such as point cloud data, geometric features, a temperature field, a humidity field and a pug matching vector of a dark-red enameled pottery blank, the influence of each factor on deformation is considered, and the accuracy of deformation prediction can be improved; deformation field prediction is carried out through the conditional generation type deformation network, then the thermodynamic diagram is calculated and generated, deformation data can be visually seen, manual experiment adjustment is not needed, resources are saved, the yield is improved, and the rework rate is reduced.

Description

Technical Field

[0001] This invention relates to the field of casting technology, and in particular to a method and system for predicting surface deformation of raw purple clay teapot blanks based on three-dimensional scanning and neural networks. Background Technology

[0002] Before and after firing, depending on the raw materials and temperature / humidity, the defects in the raw clay body of a Zisha teapot will undergo non-linear stretching, scaling, and other deformations. Specifically, during the drying and firing stages, the clay composition, ambient temperature and humidity, etc., will affect the defective areas, causing non-linear stretching, compression, and warping deformations. In traditional Zisha teapot firing processes, problems often arise such as poor adhesion of the clay slurry for decorative patterns, leading to pattern detachment after firing; pattern distortion caused by surface stretching; and low yield. Traditional visual inspection or voxel filling methods are only suitable for detecting rigid defects and cannot predict elastic-plastic coupled deformation, resulting in offsetting or cracking of the decorative patterns and a low yield.

[0003] To overcome the problems in the traditional Yixing teapot firing process, the workers usually adjust the appearance, size, and curvature of the raw teapot by selecting raw materials, controlling temperature and humidity, and controlling firing time. Through continuous experimentation, they select satisfactory teapots from the finished products.

[0004] However, the traditional method of firing Yixing teapots requires strict control of parameters at each stage, and the experimental process is complicated and requires real-time manual monitoring, which often results in resource waste and low yield. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, a method and system for predicting surface deformation of raw Zisha teapots based on three-dimensional scanning and neural networks is provided, which can improve the yield, reduce the rework rate and save resources.

[0006] A method for predicting surface deformation of raw purple clay teapots based on 3D scanning and neural networks, the method comprising:

[0007] A three-dimensional scan of the raw clay teapot is performed to obtain point cloud data of the raw clay surface; the geometric features of the raw clay surface are determined based on the point cloud data of the raw clay surface.

[0008] The temperature and humidity fields of the raw purple clay teapot are collected, and the clay ratio vector of the raw purple clay teapot is obtained.

[0009] The point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay mix ratio vector are used as inputs. The deformation field is predicted through a conditional generative deformation network, and the deformation increment is output.

[0010] Based on the deformation increment, the deformation risk score is calculated point by point on the surface of the embryo according to the point cloud data of the embryo surface;

[0011] The region to be predicted is determined, and the deformation hazard scores corresponding to the region are obtained. Based on the deformation hazard scores, the regional hazard index is calculated, and a real-time heat map is generated.

[0012] In one embodiment, a three-dimensional scan of the unglazed Zisha teapot is performed to obtain point cloud data of the unglazed surface, including:

[0013] A three-dimensional scan of the raw purple clay teapot is performed to obtain the point cloud three-dimensional coordinates calculated in real time after scanning the raw purple clay teapot from various perspectives.

[0014] The iterative nearest point algorithm is used to transfer the three-dimensional coordinates of the point cloud to the global coordinate system, forming an initial point cloud dataset.

[0015] The initial point cloud dataset is optimized to obtain point cloud data of the embryo surface.

[0016] In one embodiment, the temperature and humidity fields of the raw purple clay teapot are collected, including:

[0017] Temperature and humidity data corresponding to the raw clay teapot are collected periodically by various sensors.

[0018] Based on the temperature data, a temperature sequence of temperature field changes over time is obtained; based on the humidity data, the humidity field distribution inside and on the surface of the raw purple clay teapot is obtained;

[0019] Using an interpolation algorithm, a three-dimensional temperature matrix is ​​generated based on the temperature sequence as a temperature field; a three-dimensional humidity matrix is ​​generated based on the humidity field distribution as a humidity field.

[0020] In one embodiment, the conditional generative deformation network includes a 3D sparse convolutional encoder, a FiLM conditional fusion layer, and a trilinear interpolation decoder; it takes the green embryo surface point cloud data, green embryo surface geometric features, temperature field, humidity field, and clay mix ratio vector as input, and performs deformation field prediction through the conditional generative deformation network to output the deformation increment, including:

[0021] The point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay proportion vector are converted into sparse voxel feature tensors as input.

[0022] The sparse voxel feature tensor is input into the 3D sparse convolutional encoder, and the 3D sparse convolutional encoder performs stepwise downsampling processing to obtain multi-scale feature maps with different resolutions.

[0023] The feature map with the lowest resolution is found from each of the multi-scale feature maps and pooled to generate a global feature vector. The global feature vector is then concatenated with the mud proportion vector through the FiLM conditional fusion layer. Finally, the trilinear interpolation decoder performs upsampling based on the multi-scale feature map to predict the displacement vector of each point and generate the deformation increment.

[0024] In one embodiment, based on the deformation increment, a deformation risk score is calculated point-by-point on the green embryo surface according to the green embryo surface point cloud data, including:

[0025] Based on the deformation increment, a deformation field is defined, and under the deformation field, the displacement gradient tensor of each point in the point cloud data of the green embryo surface is calculated;

[0026] The strain tensor is calculated based on the displacement gradient tensor, and the deformation hazard score is calculated based on the strain tensor.

[0027] In one embodiment, determining the region to be predicted and obtaining various deformation hazard scores corresponding to the region to be predicted includes:

[0028] Obtain the area selection instruction, and use the area temporarily selected by the user as the area to be predicted according to the area selection instruction;

[0029] Identify each point cloud data in the region to be predicted, and find each deformation hazard score corresponding to each point cloud data.

[0030] In one embodiment, the method further includes:

[0031] Display the heatmap and corresponding modification suggestions;

[0032] The modified purple clay teapot blank is obtained by modifying it according to the modification suggestions, and the surface point cloud data of the modified purple clay teapot blank is obtained by three-dimensional scanning again.

[0033] An updated heatmap is generated corresponding to the modified point cloud data of the raw Zisha teapot surface until no further modifications are suggested for the updated heatmap.

[0034] A system for predicting surface deformation of raw purple clay teapots based on 3D scanning and neural networks, the system comprising:

[0035] The scanning module is used to perform three-dimensional scanning on the raw clay teapot to obtain point cloud data of the raw clay surface; and to determine the geometric features of the raw clay surface based on the point cloud data of the raw clay surface.

[0036] The data acquisition module is used to acquire the temperature field and humidity field of the raw purple clay teapot and obtain the clay ratio vector of the raw purple clay teapot.

[0037] The deformation increment generation module is used to take the point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay mix ratio vector as input, and predict the deformation field through a conditional generative deformation network to output the deformation increment.

[0038] The score calculation module is used to calculate the deformation risk score of the green embryo surface point by point based on the deformation increment and the green embryo surface point cloud data.

[0039] The deformation prediction module is used to determine the area to be predicted, obtain the deformation hazard scores corresponding to the area to be predicted, calculate the area hazard index based on the deformation hazard scores, and generate a real-time heat map.

[0040] In one embodiment, the scanning module is further configured to: perform a three-dimensional scan of the raw purple clay teapot to obtain the three-dimensional coordinates of the point cloud calculated in real time after scanning the raw purple clay teapot from various perspectives; use an iterative nearest point algorithm to transfer the three-dimensional coordinates of the point cloud to the global coordinate system to form an initial point cloud dataset; and optimize the initial point cloud dataset to obtain the point cloud data of the raw clay teapot surface.

[0041] The aforementioned method and system for predicting surface deformation of raw Zisha teapots based on 3D scanning and neural networks improves the accuracy of deformation prediction by collecting point cloud data, geometric features, temperature field, humidity field, and clay ratio vector of the raw Zisha teapot. It takes into account the influence of various factors on deformation. By predicting the deformation field through a conditional generative deformation network, the system calculates the deformation hazard score and hazard index, and generates a heat map, which allows for a direct view of the deformation data. This eliminates the need for manual experimental adjustments, saves resources, increases the yield, and reduces the rework rate. Attached Figure Description

[0042] Figure 1 This is an application environment diagram of a method for predicting surface deformation of raw Zisha teapots based on 3D scanning and neural networks in one embodiment.

[0043] Figure 2 This is a flowchart illustrating a method for predicting surface deformation of a raw Zisha teapot based on 3D scanning and neural networks in one embodiment.

[0044] Figure 3 This is a structural block diagram of a Zisha teapot raw body surface deformation prediction system based on three-dimensional scanning and neural networks in one embodiment.

[0045] Figure 4 This is a flowchart illustrating a system for predicting surface deformation of a raw Zisha teapot based on 3D scanning and neural networks in one embodiment.

[0046] Figure 5This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] The surface deformation prediction method for raw Zisha teapots based on 3D scanning and neural networks provided in this application can be applied to, for example... Figure 1 The application environment shown. For example... Figure 1 As shown, the application environment includes computer device 110. Computer device 110 can perform 3D scanning of the raw clay teapot to obtain point cloud data of its surface; determine the geometric features of the raw clay surface based on the point cloud data; collect the temperature and humidity fields of the raw clay teapot and obtain the clay proportion vector; use the point cloud data, geometric features, temperature field, humidity field, and clay proportion vector as input, and predict the deformation field through a conditional generative deformation network, outputting the deformation increment; calculate the deformation hazard score point by point on the raw clay surface based on the deformation increment and the point cloud data; determine the area to be predicted, obtain the corresponding deformation hazard scores for each area, calculate the area hazard index based on each deformation hazard score, and generate a real-time heat map. The computer device 110 can be, but is not limited to, various personal computers, laptops, smartphones, robots, drones, tablets, portable devices, etc.

[0049] In one embodiment, such as Figure 2 As shown, a method for predicting surface deformation of a raw Zisha teapot based on 3D scanning and neural networks is provided, including the following steps:

[0050] Step 202: Perform a three-dimensional scan on the raw clay teapot to obtain point cloud data of the raw clay surface; determine the geometric features of the raw clay surface based on the point cloud data of the raw clay surface.

[0051] A 3D scanner can use structured light or laser triangulation to perform a 3D scan of the raw clay teapot and obtain point cloud data P0(x, y, z) on the surface of the raw clay teapot.

[0052] In one embodiment, the provided method for predicting the surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks may further include a three-dimensional scanning process. The specific process includes: performing a three-dimensional scan on the raw Zisha teapot to obtain the three-dimensional coordinates of the point cloud calculated in real time from various perspectives after scanning the raw Zisha teapot; using an iterative nearest point algorithm to transfer the three-dimensional coordinates of the point cloud to the global coordinate system to form an initial point cloud dataset; and optimizing the initial point cloud dataset to obtain the point cloud data of the raw teapot surface.

[0053] By setting up a multimodal acquisition module, including 3D scanning, RGB / hyperspectral reconstruction, and an environmental sensing unit, multimodal data related to the raw clay body of a Zisha teapot can be acquired. Specifically, the 3D scanning can use structured light or laser triangulation to obtain point cloud data P0(x, y, z) on the surface of the raw clay body.

[0054] Step 204: Collect the temperature and humidity fields of the raw Zisha teapot and obtain the clay ratio vector of the raw Zisha teapot.

[0055] The computer equipment can simultaneously acquire the temperature field T(x, y, z) and humidity field H(x, y, z) using a multimodal acquisition module, and read the clay proportion vector M. The clay proportion vector M can be the moisture content, percentage of purple clay, etc., and the specific parameters of M can be set according to actual conditions.

[0056] In one embodiment, the method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks may further include the process of collecting temperature and humidity data. Specifically, the process includes: periodically collecting temperature and humidity data corresponding to the raw Zisha teapot using various sensors; obtaining a temperature sequence of temperature field changes over time based on the temperature data; obtaining the humidity field distribution inside and on the surface of the raw Zisha teapot based on the humidity data; generating a three-dimensional temperature matrix as the temperature field based on the temperature sequence using an interpolation algorithm; and generating a three-dimensional humidity matrix as the humidity field based on the humidity field distribution.

[0057] Among them, RGB / hyperspectral reconstruction can be used for 3D reconstruction texture display without the need for defect analysis; in the environmental sensing unit, the infrared thermal imager measures the temperature field T(x, y, z); and the relative humidity sensing array measures the humidity field H(x, y, z).

[0058] Step 206: The point cloud data of the green body surface, the geometric features of the green body surface, the temperature field, the humidity field, and the clay mix ratio vector are used as inputs. The deformation field is predicted through a conditional generative deformation network, and the deformation increment is output.

[0059] The conditional generative deformation network comprises a 3D sparse convolutional encoder, a FiLM conditional fusion layer, and a trilinear interpolation decoder. The computer can input data such as the green surface point cloud P0, temperature field T, humidity field H, and clay mix ratio vector M into the conditional generative deformation network, outputting the deformation increment ΔG = {Δx, Δy, Δz, σ} after time Δt, where σ is the local deformation confidence level. The conditional generative deformation network uses L = λ1||ΔG - GT||2 + λ2Lap as the loss function.

[0060] In one embodiment, the method for predicting the surface deformation of a Zisha teapot blank based on 3D scanning and neural networks may further include a deformation field prediction process. This process includes: converting the point cloud data of the blank surface, the geometric features of the blank surface, the temperature field, the humidity field, and the clay proportion vector to obtain a sparse voxel feature tensor as input; inputting the sparse voxel feature tensor into a 3D sparse convolutional encoder, and performing progressive downsampling processing through the 3D sparse convolutional encoder to obtain multi-scale feature maps of different resolutions; finding the feature map with the lowest resolution among the various multi-scale feature maps and performing pooling processing to generate a global feature vector; concatenating the global feature vector with the clay proportion vector through a FiLM conditional fusion layer, and then performing upsampling processing based on the multi-scale feature maps through a trilinear interpolation decoder to predict the displacement vector of each point and generate the deformation increment.

[0061] In this embodiment, a 3D sparse convolutional encoder is used to extract geometric features; T, H, and M are injected through a FiLM layer; a trilinear interpolation decoder is used for trilinear interpolation upsampling and outputs the deformation increment ΔG.

[0062] Step 208: Based on the deformation increment, calculate the deformation risk score point by point on the green embryo surface according to the green embryo surface point cloud data.

[0063] Computer equipment can generate a predicted deformation point cloud P based on the deformation increment. t And calculate the deformation risk score.

[0064] Specifically, in one embodiment, a method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks may further include a process of calculating a deformation hazard score. The specific process includes: defining a deformation field based on the deformation increment; calculating the displacement gradient tensor of each point in the point cloud data of the raw teapot surface under the deformation field; calculating the strain tensor based on the displacement gradient tensor; and calculating the deformation hazard score based on the strain tensor.

[0065] Specifically, for any point i on the surface of a raw Zisha teapot, the computer equipment can calculate the deformation risk score ρ point by point on the surface of the raw teapot. i =||Δd i ||·σ i , where ||Δdi || represents the magnitude of the displacement vector predicted by the network, i.e., the deformation amplitude; σ i This represents the network's confidence level in the displacement, with a value ranging from 0 to 1. Higher values ​​indicate greater confidence.

[0066] Step 210: Determine the area to be predicted, obtain the deformation hazard scores corresponding to the area to be predicted, calculate the area hazard index based on the deformation hazard scores, and generate a real-time heat map.

[0067] The user-selected area Ω is taken as the area to be predicted. The calculated area risk index can be expressed as ID_Ω=Σ_{i∈Ω}ρ i The symbol / |Ω| is used to generate a heatmap. Here, |Ω| represents the area of ​​the selected region, calculated as: number of points × area per point. The heatmap is generated based on the ID, prompting the user whether repairs are needed or if additional features can be added.

[0068] In one embodiment, a method for predicting surface deformation of a raw Zisha teapot based on 3D scanning and neural networks may further include a process of obtaining deformation hazard scores. The specific process includes: obtaining a region selection instruction and using the user-selected region as the region to be predicted according to the region selection instruction; determining each point cloud data in the region to be predicted and finding each deformation hazard score corresponding to each point cloud data.

[0069] In one embodiment, the provided method for predicting the surface deformation of a Zisha teapot blank based on 3D scanning and neural networks may further include a process of modifying the blank according to the prediction results. The specific process includes: displaying a heat map and displaying modification suggestions corresponding to the heat map; obtaining the modified Zisha teapot blank according to the modification suggestions, and then performing a 3D scan again to obtain the modified Zisha teapot blank surface point cloud data; generating an updated heat map corresponding to the modified Zisha teapot blank surface point cloud data, until the updated heat map has no modification suggestions.

[0070] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0071] In one embodiment, such as Figure 3As shown, a surface deformation prediction system for raw Zisha teapots based on 3D scanning and neural networks is provided, including: a scanning module 310, a data acquisition module 320, a deformation increment generation module 330, a fraction calculation module 340, and a deformation prediction module 350, wherein:

[0072] The scanning module 310 is used to perform three-dimensional scanning on the raw clay teapot to obtain point cloud data of the raw clay surface; and to determine the geometric features of the raw clay surface based on the point cloud data of the raw clay surface.

[0073] The data acquisition module 320 is used to collect the temperature and humidity fields of the raw clay body of the Zisha teapot and obtain the clay ratio vector of the raw clay body of the Zisha teapot.

[0074] The deformation increment generation module 330 is used to take the green embryo surface point cloud data, green embryo surface geometric features, temperature field, humidity field, and clay proportion vector as input, predict the deformation field through a conditional generative deformation network, and output the deformation increment.

[0075] The score calculation module 340 is used to calculate the deformation risk score point by point on the surface of the green embryo based on the deformation increment and the point cloud data of the green embryo surface.

[0076] The deformation prediction module 350 is used to determine the area to be predicted, obtain the deformation hazard scores corresponding to the area to be predicted, calculate the area hazard index based on the deformation hazard scores, and generate a real-time heat map.

[0077] In one embodiment, the scanning module 310 is further used to perform a three-dimensional scan of the raw purple clay teapot, obtain the three-dimensional coordinates of the point cloud calculated in real time after scanning the raw purple clay teapot from various perspectives; use the iterative nearest point algorithm to transfer the three-dimensional coordinates of the point cloud to the global coordinate system to form an initial point cloud dataset; and optimize the initial point cloud dataset to obtain the point cloud data of the raw clay surface.

[0078] In one embodiment, the data acquisition module 320 is further configured to periodically acquire temperature and humidity data corresponding to the raw clay teapot body through various sensors; obtain a temperature sequence of temperature field changes over time based on the temperature data; obtain the humidity field distribution inside and on the surface of the raw clay teapot body based on the humidity data; generate a three-dimensional temperature matrix as the temperature field based on the temperature sequence using an interpolation algorithm; and generate a three-dimensional humidity matrix as the humidity field based on the humidity field distribution.

[0079] In one embodiment, the conditional generative deformation network includes a 3D sparse convolutional encoder, a FiLM conditional fusion layer, and a trilinear interpolation decoder. The deformation increment generation module 330 is further used to convert the point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay mix ratio vector to obtain a sparse voxel feature tensor as input. The sparse voxel feature tensor is input into the 3D sparse convolutional encoder, which performs progressive downsampling processing to obtain multi-scale feature maps of different resolutions. The feature map with the lowest resolution is found from each multi-scale feature map and pooled to generate a global feature vector. The global feature vector is concatenated with the clay mix ratio vector through the FiLM conditional fusion layer, and then upsampled based on the multi-scale feature map by the trilinear interpolation decoder to predict the displacement vector of each point and generate the deformation increment.

[0080] In one embodiment, the score calculation module 340 is further configured to define a deformation field based on the deformation increment, calculate the displacement gradient tensor of each point in the point cloud data of the green embryo surface under the deformation field, calculate the strain tensor based on the displacement gradient tensor, and calculate the deformation hazard score based on the strain tensor.

[0081] In one embodiment, the score calculation module 340 is further configured to obtain a region selection instruction, and use the user-selected temporary region as the region to be predicted according to the region selection instruction; determine each point cloud data in the region to be predicted, and find each deformation hazard score corresponding to each point cloud data.

[0082] In one embodiment, the deformation prediction module 350 is further configured to display a heat map and display modification suggestions corresponding to the heat map; obtain the modified purple clay teapot blank according to the modification suggestions, and perform a three-dimensional scan again to obtain the surface point cloud data of the modified purple clay teapot blank; generate an updated heat map corresponding to the surface point cloud data of the modified purple clay teapot blank, until the updated heat map has no modification suggestions.

[0083] In one embodiment, the application process of a Zisha teapot green body surface deformation prediction system based on 3D scanning and neural networks may include:

[0084] After the green embryo is pressed by the mold, the scanner acquires the point cloud data of the green embryo surface;

[0085] The craftsman wears an AR terminal and selects the area where he wants to pile up flowers. The system generates a defect heat map based on the geometric and physical characteristics of the raw materials. The terminal prompts "Red areas are recommended to be repaired with a needle".

[0086] The craftsman followed the instructions;

[0087] Scan and evaluate again. If the system no longer indicates a red area, the current embryo is considered to meet the requirements of the next step of the fabrication process, and the process ends.

[0088] In this embodiment, the process of a Zisha teapot green surface deformation prediction system based on 3D scanning and neural networks is as follows: Figure 4 As shown, the process specifically includes: performing a 3D scan of the raw Zisha teapot; determining the geometric features of the raw teapot surface through a feature extraction module, and collecting temperature and humidity fields, as well as clay ratio vectors; predicting using a conditional generative deformation network, where the user temporarily selects an area using an AR / stylus pen, the system outputs the area's danger index, and generates a real-time heat map; if the conditions are met, the craftsman proceeds to the next step of applying decorative patterns; if the conditions are not met, a correction instruction is generated to perform tasks such as tinting, adjusting the clay slurry, and modifying the mold, and then performing a 3D scan on the adjusted raw Zisha teapot.

[0089] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a method for predicting the surface deformation of raw purple clay teapots based on 3D scanning and neural networks. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0090] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0091] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a method for predicting surface deformation of raw Zisha teapots based on three-dimensional scanning and neural networks.

[0092] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a method for predicting surface deformation of raw purple clay teapots based on three-dimensional scanning and neural networks.

[0093] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0094] 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.

[0095] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks, characterized in that, The method includes: A three-dimensional scan of the raw clay teapot is performed to obtain point cloud data of the raw clay surface; the geometric features of the raw clay surface are determined based on the point cloud data of the raw clay surface. The temperature and humidity fields of the raw purple clay teapot are collected, and the clay ratio vector of the raw purple clay teapot is obtained. The point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay mix ratio vector are used as inputs. The deformation field is predicted through a conditional generative deformation network, and the deformation increment is output. Based on the deformation increment, the deformation risk score is calculated point by point on the surface of the embryo according to the point cloud data of the embryo surface; The region to be predicted is determined, and the deformation hazard scores corresponding to the region are obtained. Based on the deformation hazard scores, the regional hazard index is calculated, and a real-time heat map is generated.

2. The method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks according to claim 1, characterized in that, A 3D scan of the raw purple clay teapot was performed to obtain point cloud data of its surface, including: A three-dimensional scan of the raw purple clay teapot is performed to obtain the point cloud three-dimensional coordinates calculated in real time after scanning the raw purple clay teapot from various perspectives. The iterative nearest point algorithm is used to transfer the three-dimensional coordinates of the point cloud to the global coordinate system, forming an initial point cloud dataset. The initial point cloud dataset is optimized to obtain point cloud data of the embryo surface.

3. The method for predicting surface deformation of raw Zisha teapots based on three-dimensional scanning and neural networks according to claim 1, characterized in that, The temperature and humidity fields of the raw purple clay teapot were collected, including: Temperature and humidity data corresponding to the raw clay teapot are collected periodically by various sensors. Based on the temperature data, a temperature sequence of temperature field changes over time is obtained; based on the humidity data, the humidity field distribution inside and on the surface of the raw purple clay teapot is obtained; Using an interpolation algorithm, a three-dimensional temperature matrix is ​​generated based on the temperature sequence as a temperature field; a three-dimensional humidity matrix is ​​generated based on the humidity field distribution as a humidity field.

4. The method for predicting surface deformation of raw Zisha teapots based on three-dimensional scanning and neural networks according to claim 1, characterized in that, The conditional generative deformation network includes a 3D sparse convolutional encoder, a FiLM conditional fusion layer, and a trilinear interpolation decoder. It takes the green embryo surface point cloud data, green embryo surface geometric features, temperature field, humidity field, and clay mix vector as input, and predicts the deformation field through the conditional generative deformation network, outputting the deformation increment, including: The point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay proportion vector are converted into sparse voxel feature tensors as input. The sparse voxel feature tensor is input into the 3D sparse convolutional encoder, and the 3D sparse convolutional encoder performs stepwise downsampling processing to obtain multi-scale feature maps with different resolutions. The feature map with the lowest resolution is found from each of the multi-scale feature maps and pooled to generate a global feature vector. The global feature vector is then concatenated with the mud proportion vector through the FiLM conditional fusion layer. Finally, the trilinear interpolation decoder performs upsampling based on the multi-scale feature map to predict the displacement vector of each point and generate the deformation increment.

5. The method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks according to claim 1, characterized in that, Based on the deformation increment, a deformation risk score is calculated point-by-point on the embryo surface using the point cloud data of the embryo surface, including: Based on the deformation increment, a deformation field is defined, and under the deformation field, the displacement gradient tensor of each point in the point cloud data of the green embryo surface is calculated; The strain tensor is calculated based on the displacement gradient tensor, and the deformation hazard score is calculated based on the strain tensor.

6. The method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks according to claim 1, characterized in that, Determine the region to be predicted and obtain the deformation hazard scores corresponding to the region to be predicted, including: Obtain the area selection instruction, and use the area temporarily selected by the user as the area to be predicted according to the area selection instruction; Identify each point cloud data in the region to be predicted, and find each deformation hazard score corresponding to each point cloud data.

7. The method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks according to claim 1, characterized in that, The method further includes: Display the heatmap and corresponding modification suggestions; The modified purple clay teapot blank is obtained by modifying it according to the modification suggestions, and the surface point cloud data of the modified purple clay teapot blank is obtained by three-dimensional scanning again. An updated heatmap is generated corresponding to the modified point cloud data of the raw Zisha teapot surface until no further modifications are suggested for the updated heatmap.

8. A system for predicting surface deformation of raw purple clay teapots based on three-dimensional scanning and neural networks, characterized in that, The system includes: The scanning module is used to perform three-dimensional scanning on the raw clay teapot to obtain point cloud data of the raw clay surface; and to determine the geometric features of the raw clay surface based on the point cloud data of the raw clay surface. The data acquisition module is used to acquire the temperature field and humidity field of the raw purple clay teapot and obtain the clay ratio vector of the raw purple clay teapot. The deformation increment generation module is used to take the point cloud data of the green embryo surface, the geometric features of the green embryo surface, the temperature field, the humidity field, and the clay mix ratio vector as input, and predict the deformation field through a conditional generative deformation network to output the deformation increment. The score calculation module is used to calculate the deformation risk score of the green embryo surface point by point based on the deformation increment and the green embryo surface point cloud data. The deformation prediction module is used to determine the area to be predicted, obtain the deformation hazard scores corresponding to the area to be predicted, calculate the area hazard index based on the deformation hazard scores, and generate a real-time heat map.

9. The method for predicting surface deformation of a raw Zisha teapot based on three-dimensional scanning and neural networks according to claim 8, characterized in that, The scanning module is also used to: perform three-dimensional scanning on the raw purple clay teapot and obtain the point cloud three-dimensional coordinates calculated in real time after scanning the raw purple clay teapot from various perspectives; The iterative nearest point algorithm is used to transfer the three-dimensional coordinates of the point cloud to the global coordinate system, forming an initial point cloud dataset. The initial point cloud dataset is optimized to obtain point cloud data of the embryo surface.

Images