A method for generating 3D cabinet based on image recognition

Abstract

The present invention discloses a method for generating a 3D cabinet based on image recognition, which relates to the technical field of image recognition processing, including obtaining a cabinet design hand-drawn sketch for grayscale processing, and using sparse matrix decomposition to perform denoising and enhancement processing on the hand-drawn sketch; extracting the sketch design lines and dimensions, and using a three-dimensional convolutional neural network to establish an initial three-dimensional cabinet model; optimizing the initial three-dimensional cabinet model based on spatial geometric reasoning and outputting the designed three-dimensional cabinet model; displaying the designed three-dimensional cabinet model and storing it in a database. The present invention obtains multi-view views of the design hand-drawn sketch for grayscale processing, and uses sparse decomposition and genetic algorithms to perform denoising and enhancement processing before extracting the design lines and dimension information, thereby greatly improving the accuracy and correctness of design scheme recognition, and generates an initial three-dimensional model using a three-dimensional convolutional neural network and optimizes it using spatial geometric reasoning, thereby effectively improving the accuracy and generation speed of the three-dimensional model.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition processing, and in particular to a method for generating a 3D cabinet based on image recognition. Background Art

[0002] With the rapid development of computer vision and artificial intelligence technology, image recognition has been widely used in various fields, especially in the field of 3D modeling and design. Traditional 3D modeling technology usually relies on professional software and the experience of designers. These methods are not only time-consuming, but also require a high technical threshold. In recent years, 3D model generation methods based on image recognition have gradually emerged, which automatically generate 3D models by analyzing the information in 2D images. Especially in the fields of furniture design and architectural design, the digital transformation of hand-drawn sketches has greatly improved the design efficiency and provided designers with a more intuitive means of creation. However, the existing technology still has shortcomings. Traditional denoising and image enhancement methods often rely on simple image processing algorithms, which cannot effectively deal with the common noise and blur problems in hand-drawn sketches, and lack optimization methods for the geometric structure of the model, resulting in the generated model being insufficient in design accuracy and operability. Summary of the invention

[0003] In view of the problems existing in the above-mentioned existing methods for generating 3D cabinets based on image recognition, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is that traditional denoising and image enhancement methods often rely on simple image processing algorithms, which cannot effectively deal with the common noise and blur problems in hand-drawn sketches, and lack optimization methods for the model's geometric structure, resulting in the generated models being insufficient in design accuracy and operability.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a method for generating a 3D cabinet based on image recognition, which includes obtaining a hand-drawn sketch of the cabinet design for grayscale processing, and using sparse matrix decomposition to perform denoising and enhancement processing on the hand-drawn sketch; extracting sketch design lines and dimensions, and using a three-dimensional convolutional neural network to establish an initial three-dimensional cabinet model; optimizing the initial three-dimensional cabinet model based on spatial geometric reasoning and outputting a designed three-dimensional cabinet model; displaying the designed three-dimensional cabinet model and storing it in a database.

[0006] As a preferred solution of the method for generating a 3D cabinet based on image recognition described in the present invention, the step of obtaining the cabinet design hand-drawn sketch and performing grayscale processing refers to using a scanning device to scan the cabinet design paper file to obtain the front view, side view, top view and cross-sectional view of the cabinet design hand-drawn sketch, performing grayscale processing on each pixel in the cabinet design hand-drawn sketch to form a single-channel grayscale image, and simultaneously obtaining the designer's identity as a watermark and adding it to the design hand-drawn sketch.

[0007] As a preferred solution of the method for generating a 3D cabinet based on image recognition according to the present invention, the denoising and enhancement processing of the hand-drawn sketch using sparse matrix decomposition includes:

[0008] Generate a random Gaussian matrix A from an independent Gaussian distribution with a dimension of m*n, and project the grayscale image into a low-dimensional space through the random Gaussian matrix A to obtain a compressed signal y;

[0009] The Gabor function for generating a redundant dictionary is defined as follows:

[0010] ;

[0011] in, is the atom of the redundant dictionary, is the scale parameter, For time shift, is the frequency parameter, is the phase shift, is the time-frequency parameter;

[0012] The time-frequency parameters of the Gabor function Discretize and generate redundant dictionaries D composed of different atoms:

[0013] ;

[0014] Where k is the number of atoms in the redundant dictionary;

[0015] Decompose the compressed signal y according to the redundant dictionary:

[0016] ,

[0017] in is a sparse coefficient vector;

[0018] Set initial residuals , a collection of atoms It is an empty set, and the iteration begins. In each iteration, the optimization objective function is defined based on the NSGA-Ⅲ multi-objective optimization algorithm as:

[0019] ;

[0020] in is the residual of the ith iteration, is the set of atoms in the i-th iteration, is the jth atom in the redundant dictionary, is the inner product of the atom and the residual, is the inner product of an atom and the atoms in its set of atoms;

[0021] In each iteration, a genetic algorithm is used to perform crossover and mutation operations to generate the next generation of candidate atoms, the optimal atom is selected based on the non-dominated sorting mechanism, the optimal atom is added to the atom set, and a redundant dictionary subset is formed based on the atom set. , and update the sparse coefficient vector and residual;

[0022] After reaching the preset maximum sparsity, the iteration stops and the redundant dictionary subset and sparse coefficient vector after iteration are output. ;

[0023] Use the iterated sparse coefficient vector Reconstructing compressed signals , the reconstructed compressed signal is converted back into a denoised and enhanced grayscale image through the inverse matrix of the random Gaussian matrix A.

[0024] As a preferred solution of the method for generating a 3D cabinet based on image recognition described in the present invention, wherein: the extraction of sketch design lines and dimensions refers to using the Canny algorithm to perform edge detection on the denoised and enhanced grayscale image to extract the edge image, using the Hough transform to detect and identify the design lines in the edge image, and using the OCR technology to perform character recognition on the denoised and enhanced grayscale image to extract dimension data, and marking the extracted dimension information corresponding to the design lines.

[0025] As a preferred solution of the method for generating a 3D cabinet based on image recognition described in the present invention, wherein: the use of a three-dimensional convolutional neural network to establish an initial three-dimensional cabinet model refers to grayscale processing and denoising of the front view, side view, top view and cross-sectional view in the hand-drawn sketch, and extracting design lines and size information;

[0026] Construct a 3D convolutional neural network, set the input as the design lines and size information of the front view, side view, top view and cross-sectional view, output as 3D point cloud data, define the mean square error as the loss function, and use the Adam optimizer to optimize the parameters of the 3D convolutional neural network;

[0027] The extracted design lines and dimension information are input into the trained 3D convolutional neural network to obtain the 3D point cloud data of the design scheme;

[0028] The Poisson surface reconstruction algorithm is used to generate an initial three-dimensional cabinet model according to the three-dimensional point cloud data of the design scheme, and the simplification algorithm is applied to simplify the initial three-dimensional cabinet model.

[0029] As a preferred solution of the method for generating a 3D cabinet based on image recognition according to the present invention, the method of optimizing the initial three-dimensional cabinet model based on spatial geometric reasoning and outputting the designed three-dimensional cabinet model refers to importing the initial three-dimensional cabinet model into the processing environment and calculating the normal vector for each face of the model:

[0030] ;

[0031] in is the normal vector, as well as are the three vertices on the face;

[0032] Define geometric constraints, use normal vectors and model vertices as initial particles, and use particle swarm optimization algorithm to adjust model vertices and normal vectors until the geometric constraints are met;

[0033] The updated model vertices and normal vectors are used to reconstruct the designed three-dimensional cabinet model for output.

[0034] As a preferred solution of the method for generating a 3D cabinet based on image recognition described in the present invention, the display of the designed three-dimensional cabinet model refers to identifying the designer according to the watermark in the design hand-drawn sketch, outputting the initial three-dimensional cabinet model and the designed three-dimensional cabinet model, and collecting feedback from the designer to optimize the details of the designed three-dimensional cabinet model until it fully meets the designer's requirements.

[0035] As a preferred solution of the method for generating a 3D cabinet based on image recognition described in the present invention, the storage in the database refers to obtaining the designer's identity, design hand-drawn sketches, and design three-dimensional cabinet models to form a data set and storing them in the database, and classifying and marking them according to the designer's identity and time, regularly performing integrity checks on the data stored in the database, and synchronously uploading the stored data to the cloud for backup after the detection is completed.

[0036] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the method for generating a 3D cabinet based on image recognition when executing the computer program.

[0037] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for generating a 3D cabinet based on image recognition.

[0038] The beneficial effects of the present invention are as follows: the present invention obtains multi-perspective views of the design hand-drawn sketch for grayscale processing, and uses sparse decomposition and genetic algorithm to perform denoising and enhancement processing before extracting the design lines and size information, thereby greatly improving the accuracy and correctness of design scheme recognition. The present invention generates an initial three-dimensional model by using a three-dimensional convolutional neural network and optimizes it using spatial geometric reasoning, thereby effectively improving the accuracy and generation speed of the three-dimensional model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0040] Figure 1 A flowchart of a method for generating a 3D cabinet based on image recognition is shown in FIG. DETAILED DESCRIPTION

[0041] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0042] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0044] Example 1

[0045] Reference Figure 1 , which is the first embodiment of the present invention, and which provides a method for generating a 3D cabinet based on image recognition. The method for generating a 3D cabinet based on image recognition includes:

[0046] S1. Obtain the cabinet design hand-drawn sketch for grayscale processing, and use sparse matrix decomposition to perform denoising and enhancement processing on the hand-drawn sketch;

[0047] Specifically, obtaining the cabinet design hand-drawn sketch for grayscale processing means using a scanning device to scan the cabinet design paper file to obtain the front view, side view, top view and cross-sectional view of the cabinet design hand-drawn sketch, and grayscale processing each pixel in the cabinet design hand-drawn sketch.

[0048] ;

[0049] in is the grayscale value, R, G and B are the pixel values ​​of the red, green and blue channels in the hand-drawn sketch respectively;

[0050] A single-channel grayscale image is formed, and the designer's identity is simultaneously obtained and added to the design hand-drawn sketch as a watermark.

[0051] By using a scanning device to scan paper design files, we can obtain front views, side views, top views and cross-sectional views to ensure that all angles of the design are fully recorded. The effectiveness of this step lies in that it converts traditional paper designs into digital formats, which is convenient for subsequent image processing and analysis. In addition, obtaining multiple views helps to fully understand the structure of the cabinet and lays the foundation for three-dimensional modeling. Grayscale processing of the scanned color image can remove unnecessary color information and highlight lines and shapes. In grayscale images, lines and edges are more obvious, which provides a better foundation for subsequent edge detection and feature extraction. By comparing different image processing algorithms, we can select the most suitable technology for the current task, thereby improving the accuracy of model generation. Adding a designer's identity watermark to the design hand-drawn sketch can not only effectively protect the designer's copyright, but also ensure the source of the design in data sharing and collaboration. The use of watermarks helps to build trust and prevent design works from being stolen or tampered with. In addition, identity watermarks also provide convenience in data management and version control, making design tracking and review more efficient. The single-channel grayscale image formed provides an operational data basis for the subsequent three-dimensional cabinet generation. Due to its simplified structure and obvious features, it can be more easily combined with other processing algorithms (such as edge detection and feature extraction), thereby supporting the automation and intelligence of subsequent steps. The conversion of this data format is a key link in the entire process and lays a good foundation for the realization of the final three-dimensional model.

[0052] Further, the sparse matrix decomposition is used to perform hand-drawn sketch denoising and enhancement processing, including:

[0053] Generate a random Gaussian matrix A from an independent Gaussian distribution with a dimension of m*n, and project the grayscale image into a low-dimensional space through the random Gaussian matrix A to obtain a compressed signal y;

[0054] The Gabor function for generating a redundant dictionary is defined as follows:

[0055] ;

[0056] in, is the atom of the redundant dictionary, is the scale parameter, For time shift, is the frequency parameter, is the phase shift, is the time-frequency parameter;

[0057] Gabor function is a time-frequency analysis tool with good spatial locality and frequency selectivity. Gabor function is widely used in image processing and feature extraction, and can effectively capture edge and texture information in images.

[0058] The time-frequency parameters of the Gabor function Discretize and generate redundant dictionaries D composed of different atoms:

[0059] ;

[0060] Where k is the number of atoms in the redundant dictionary;

[0061] Decompose the compressed signal y according to the redundant dictionary:

[0062] ,

[0063] in is a sparse coefficient vector;

[0064] Set initial residuals , a collection of atoms It is an empty set and starts iteration, including two goals:

[0065] Maximize the matching degree between the atom and the residual, that is, the inner product of the atom and the current residual:

[0066] ;

[0067] Minimize the mutual interference between atoms, that is, the inner product of atoms with atoms in the collection of atoms:

[0068] ;

[0069] Combining the two objectives, the optimization objective function is defined based on the NSGA-Ⅲ multi-objective optimization algorithm in each iteration as:

[0070] ;

[0071] in is the residual of the ith iteration, is the set of atoms in the i-th iteration, is the jth atom in the redundant dictionary, is the inner product of the atom and the residual, is the inner product of an atom and the atoms in its set of atoms;

[0072] NSGA-III (Non-dominated Sorting Genetic Algorithm III) is an evolutionary algorithm for solving multi-objective optimization problems. The algorithm optimizes multiple objective functions simultaneously and selects non-dominated solutions to improve the performance and convergence speed of the algorithm. Its core is to maintain diversity and select the optimal solution, which is suitable for dealing with complex optimization problems.

[0073] In each iteration, a genetic algorithm is used to perform crossover and mutation operations to generate the next generation of candidate atoms, the optimal atom is selected based on the non-dominated sorting mechanism, the optimal atom is added to the atom set, and a redundant dictionary subset is formed based on the atom set. , and update the sparse coefficient vector and residual;

[0074] ;

[0075] ;

[0076] in is the updated sparse coefficient vector, is the updated residual, is the redundant dictionary subset composed of the atomic set of the i-th iteration, is the sparse coefficient vector of the i-th iteration;

[0077] After reaching the preset maximum sparsity, the iteration stops and the redundant dictionary subset and sparse coefficient vector after iteration are output. ;

[0078] Use the iterated sparse coefficient vector Reconstructing compressed signals , the reconstructed compressed signal is converted back into a denoised and enhanced grayscale image through the inverse matrix of the random Gaussian matrix A.

[0079] A random Gaussian matrix is ​​generated from an independent Gaussian distribution to project the grayscale image of the hand-drawn sketch. By multiplying the grayscale image with the random Gaussian matrix, the resulting compressed signal helps reduce the data dimension and computational complexity. This compression not only speeds up the subsequent processing speed, but also effectively retains the important information in the image, making data processing more efficient. The redundant dictionary is generated by using the Gabor function to generate multiple atoms by discretizing the time-frequency parameters. The diversity of the redundant dictionary enables it to more comprehensively describe the different features in the hand-drawn sketch, improving the flexibility and accuracy of image processing. By selecting appropriate scale and frequency parameters, the Gabor function can extract the most representative texture and edge features in the image. By sparsely decomposing the compressed signal and decomposing it using the redundant dictionary and sparse coefficient vector, the important features in the image can be effectively identified and retained. This process ensures the accuracy of the decomposition result by maximizing the matching degree between atoms and residuals and minimizing the mutual interference between atoms. At the same time, the NSGA-III multi-objective optimization algorithm provides an efficient solution, and selects the optimal atoms through a non-dominated sorting mechanism, further optimizing the quality of the sparse decomposition. After reaching the preset maximum sparsity, the iterated sparse coefficient vector is reconstructed into a compressed signal through the inverse matrix, and finally a denoised and enhanced grayscale image is output. This process not only improves the visual quality of the hand-drawn sketch and removes noise, but also ensures the retention of details, providing reliable data support for the subsequent generation of 3D models.

[0080] S2, extracting the sketch design lines and dimensions, and using a three-dimensional convolutional neural network to build an initial three-dimensional cabinet model;

[0081] Specifically, extracting the sketch design lines and dimensions refers to using the Canny algorithm to perform edge detection on the denoised and enhanced grayscale image to extract the edge image, using the Hough transform to detect and identify the design lines in the edge image, and using the OCR technology to perform character recognition on the denoised and enhanced grayscale image to extract dimension data, and marking the extracted dimension information with the design lines.

[0082] The Canny algorithm is used to detect the edge of the denoised and enhanced grayscale image, which can effectively extract the important edge information in the image. Through image smoothing and gradient calculation, the algorithm improves the edge detection accuracy. The generation of edge images provides a clear basis for subsequent line recognition and ensures the integrity of design information. The Hough transform is used to detect the straight lines of the edge image, which can identify the key lines in the design sketch. These lines usually represent the outline and structure of the design. The high efficiency of the Hough transform enables the accurate extraction of straight line information even in complex backgrounds. By comparing the identified lines with the original design sketch, the accuracy and integrity of the extracted information can be ensured. The OCR technology is used to perform character recognition on the denoised and enhanced grayscale image, and the dimension information is extracted from the image to ensure that all design annotations are digitized. This step improves the efficiency of the design process by converting handwritten or printed dimension text into a machine-readable format. The extracted dimension information is marked correspondingly with the design lines, so that the actual size of each line can be clarified. This not only improves the accuracy of the design, but also provides the necessary dimension parameters for the subsequent three-dimensional model generation, ensuring the feasibility of the model in practical applications.

[0083] Furthermore, using a three-dimensional convolutional neural network to establish an initial three-dimensional cabinet model means grayscale processing and denoising the front view, side view, top view, and cross-sectional view in the hand-drawn sketch, and extracting design lines and size information;

[0084] Construct a 3D convolutional neural network, set the input as the design lines and size information of the front view, side view, top view and cross-sectional view, output as 3D point cloud data, define the mean square error as the loss function, and use the Adam optimizer to optimize the parameters of the 3D convolutional neural network;

[0085] The extracted design lines and dimension information are input into the trained 3D convolutional neural network to obtain the 3D point cloud data of the design scheme;

[0086] The Poisson surface reconstruction algorithm is used to generate an initial three-dimensional cabinet model according to the three-dimensional point cloud data of the design scheme, and the simplification algorithm is applied to simplify the initial three-dimensional cabinet model.

[0087] Grayscale processing and denoising of the front view, side view, top view and cross-sectional view in the hand-drawn sketch can significantly improve the quality of the image. Grayscale processing makes the edges and lines more prominent by reducing color information; denoising eliminates interference information in the sketch to ensure the clarity of the design information. The beneficial effect of this step is that it provides clean basic data for subsequent feature extraction and model building, improves the accuracy of model generation, and extracts design lines and size information from the denoised grayscale image, so that the 3D convolutional neural network can receive clear input data. This process ensures the integrity and accuracy of the design information used in model building. , which improves the expressiveness and details of subsequent 3D models. The effectiveness of this stage not only improves the quality of data input, but also lays a solid foundation for 3D modeling. 3DCNN is set to receive design lines and size information and generate corresponding 3D point cloud data. By using mean square error as the loss function, the Adam optimizer is used to optimize the network parameters to ensure that the network can learn the best feature mapping. The beneficial effect of this step is that it can efficiently extract key features from multi-view information and accurately generate 3D models, solving the problem of information loss in traditional methods. The extracted design lines and size information are input into the trained 3D CNN is used to obtain the three-dimensional point cloud data of the design scheme. The significance of this process is that it can effectively convert the two-dimensional information of the hand-drawn sketch into three-dimensional data, which provides the necessary basis for the further processing of the model. The generation of point cloud data not only reflects the design intention, but also provides rich information for subsequent surface reconstruction. The Poisson surface reconstruction algorithm is used to convert the three-dimensional point cloud data into the initial three-dimensional cabinet model. The number of polygons is reduced by simplifying the algorithm, and the rendering efficiency and storage efficiency are improved. The beneficial effect of this process is that it not only generates a three-dimensional model that meets the design requirements, but also improves the processing performance of the model through optimization, ensuring good interactivity and visual effects in practical applications.

[0088] S3, optimizing the initial three-dimensional cabinet model based on spatial geometric reasoning and outputting the designed three-dimensional cabinet model;

[0089] Specifically, optimizing the initial three-dimensional cabinet model based on spatial geometric reasoning and outputting the designed three-dimensional cabinet model means importing the initial three-dimensional cabinet model into the processing environment and calculating the normal vector for each face of the model:

[0090] ;

[0091] in is the normal vector, as well as are the three vertices on the face;

[0092] Define geometric constraints, such as perpendicularity, parallelism, and proportionality, use normal vectors and model vertices as initial particles, and use the particle swarm optimization algorithm to adjust model vertices and normal vectors until the geometric constraints are met;

[0093] The updated model vertices and normal vectors are used to reconstruct the designed three-dimensional cabinet model for output.

[0094] After the initial 3D cabinet model is imported into the processing environment, the normal vector is calculated for each face of the model. This process helps to determine the direction of each face and provides basic data for the subsequent definition of geometric constraints. Accurate normal vectors can ensure the consistency of lighting and rendering effects and improve visual expression. Geometric constraints such as verticality, parallelism and proportional relationships can ensure that the model meets actual use requirements in design. For example, ensure that the faces of the cabinet are vertical to enhance structural stability. The definition of these geometric constraints helps to avoid errors in the design process and ensure the feasibility of the model in practical applications. The normal vector and model vertices are used as initial particles, and the vertices and normal vectors of the model are adjusted through the particle swarm optimization algorithm. The PSO algorithm can effectively explore in high-dimensional space and quickly converge to the optimal solution. The beneficial effects of this process are The result is that it can automatically adjust the model structure according to the set geometric constraints, reduce manual intervention, and improve design efficiency and accuracy. After meeting all geometric constraints, it uses the updated model vertices and normal vectors to reconstruct the designed three-dimensional cabinet model. The reconstruction process can effectively integrate the optimized parameters to generate the final design model. This process ensures the visual and functional rationality of the model and meets the design requirements. The generated three-dimensional cabinet model can be directly used in subsequent production and manufacturing processes to improve the practicality of the design. At the same time, through the digital model, it can be easily modified, shared and displayed to meet the needs of modern design processes. The implementation of this method provides an efficient and reliable solution for fields such as furniture design and interior design, and promotes the intelligent and automated process of design work.

[0095] S4, display the designed three-dimensional cabinet model and store it in the database;

[0096] Specifically, displaying the designed 3D cabinet model means identifying the designer based on the watermark in the design hand-drawn sketch, outputting the initial 3D cabinet model and the designed 3D cabinet model, and collecting feedback from the designer to optimize the details of the designed 3D cabinet model until it fully meets the designer's requirements.

[0097] By identifying the designer's watermark, the copyright and ownership of the design can be ensured. This process not only protects the designer's intellectual property rights, but also provides clear responsibility for subsequent feedback and optimization. When giving feedback, designers can more clearly express their needs for the model, thereby improving the efficiency of design communication. The effectiveness of watermark recognition lies in that it provides a personalized customization basis for the model display process, so that each designer's unique style and requirements can be reflected in the final model, and the initial 3D cabinet model and the designed 3D cabinet model are output to ensure that designers can evaluate and provide feedback at two stages. The initial model provides a basic reference for the design, while the optimized model reflects the final effect of the design. This process helps to improve the transparency and auditability of the design, allowing designers to understand the progress of the design more intuitively. After the model is output, the opinions of designers are collected through an effective feedback mechanism. Designers can make specific suggestions for improvement on the appearance, function and practicality of the model. The significance of this feedback process is that it provides important data support for the optimization of model details, so that the final product can better meet user needs. Detail optimization is performed based on the feedback from designers to ensure that the model meets the design requirements in all aspects. This optimization process may involve adjustments to the size, shape, material and other aspects of the model to ensure that the final design is both beautiful and practical. Detail optimization not only improves the visual appeal of the model, but also enhances its functionality, ensuring that the design can be smoothly implemented in practical applications. The effectiveness of this process lies in that through continuous iterative optimization, the final design not only meets aesthetic standards, but also meets the needs of designers in terms of practicality.

[0098] Furthermore, storing in a database means obtaining the designer's identity, design hand-drawn sketches, and design three-dimensional cabinet models to form a data set and storing them in the database, and classifying and marking them according to the designer's identity and time, regularly performing integrity checks on the data stored in the database, and synchronously uploading the stored data to the cloud for backup after the check is completed.

[0099] The system automatically obtains the designer's identity and forms a complete data set in combination with the design hand-drawn sketch and the three-dimensional cabinet model. The effectiveness of this process lies in the ability to centrally store design-related information to ensure data consistency and traceability. As a key part of the data set, the designer's identity not only provides copyright and ownership protection, but also facilitates classification and query in data management. The data is classified and marked according to the designer's identity and time, making subsequent data retrieval and analysis more efficient. This classification system can quickly locate the design records of specific designers, facilitating team collaboration and project management. The implementation of classification marking ensures the orderliness of the data structure and helps with subsequent analysis and decision-making. The data stored in the database is regularly checked for integrity to ensure that the data has not been damaged or lost. Integrity testing not only verifies the accuracy of the data through methods such as checksums, but may also include data recovery strategies to ensure that data can be quickly recovered in unexpected situations. The advantage of this process is that it significantly improves the security and reliability of the data and reduces the risk of work interruption due to data corruption.

[0100] Example 2

[0101] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0102] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0103] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0104] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit with a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0105] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for generating a 3D cabinet based on image recognition, characterized in that: include, Obtain the cabinet design hand-drawn sketch for grayscale processing, and use sparse matrix decomposition to perform denoising and enhancement processing on the hand-drawn sketch; Extract the sketch design lines and dimensions, and use the 3D convolutional neural network to build the initial 3D cabinet model; Optimize the initial 3D cabinet model based on spatial geometric reasoning and output the designed 3D cabinet model; Display the designed 3D cabinet model and store it in the database; Optimizing the initial three-dimensional cabinet model based on spatial geometric reasoning and outputting the designed three-dimensional cabinet model refers to importing the initial three-dimensional cabinet model into the processing environment and calculating the normal vector for each face of the model: ； in is the normal vector, , as well as are the three vertices on the face; Define geometric constraints, use normal vectors and model vertices as initial particles, and use particle swarm optimization algorithm to adjust model vertices and normal vectors until the geometric constraints are met; The updated model vertices and normal vectors are used to reconstruct the designed three-dimensional cabinet model for output.

2. The method for generating a 3D cabinet based on image recognition according to claim 1, characterized in that: The obtaining of the cabinet design hand-drawn sketch for grayscale processing refers to using a scanning device to scan the cabinet design paper file to obtain the front view, side view, top view and cross-sectional view of the cabinet design hand-drawn sketch, performing grayscale processing on each pixel in the cabinet design hand-drawn sketch to form a single-channel grayscale image, and simultaneously obtaining the designer's identity as a watermark and adding it to the design hand-drawn sketch.

3. The method for generating a 3D cabinet based on image recognition according to claim 2, characterized in that: The denoising and enhancement processing of hand-drawn sketches using sparse matrix decomposition includes: Generate a random Gaussian matrix A from an independent Gaussian distribution with a dimension of m*n, and project the grayscale image into a low-dimensional space through the random Gaussian matrix A to obtain a compressed signal y; The Gabor function for generating a redundant dictionary is defined as follows: ； in, is the atom of the redundant dictionary, is the scale parameter, For time shift, is the frequency parameter, is the phase shift, is the time-frequency parameter; The time-frequency parameters of the Gabor function Discretize and generate redundant dictionaries D composed of different atoms: ； Where k is the number of atoms in the redundant dictionary; Decompose the compressed signal y according to the redundant dictionary: ； in is a sparse coefficient vector; Set initial residuals , a collection of atoms It is an empty set, and the iteration begins. In each iteration, the optimization objective function is defined based on the NSGA-Ⅲ multi-objective optimization algorithm as: ； in is the residual of the i-th iteration, is the set of atoms in the i-th iteration, is the jth atom in the redundant dictionary, is the inner product of the atom and the residual, is the inner product of an atom and the atoms in its set of atoms; In each iteration, a genetic algorithm is used to perform crossover and mutation operations to generate the next generation of candidate atoms, the optimal atom is selected based on the non-dominated sorting mechanism, the optimal atom is added to the atom set, and a redundant dictionary subset is formed based on the atom set. , and update the sparse coefficient vector and residual; After reaching the preset maximum sparsity, the iteration stops and the redundant dictionary subset and sparse coefficient vector after iteration are output. ; Use the iterated sparse coefficient vector Reconstructing compressed signals , the reconstructed compressed signal is converted back into a denoised and enhanced grayscale image through the inverse matrix of the random Gaussian matrix A.

4. The method for generating a 3D cabinet based on image recognition according to claim 3, characterized in that: The extracting of sketch design lines and dimensions refers to using the Canny algorithm to perform edge detection on the denoised and enhanced grayscale image to extract the edge image, using the Hough transform to detect and identify the design lines in the edge image, and using the OCR technology to perform character recognition on the denoised and enhanced grayscale image to extract dimension data, and marking the extracted dimension information with the design lines accordingly.

5. The method for generating a 3D cabinet based on image recognition according to claim 4, characterized in that: The use of a three-dimensional convolutional neural network to establish an initial three-dimensional cabinet model refers to grayscale processing and denoising of the front view, side view, top view and cross-sectional view in the hand-drawn sketch, and extracting design lines and size information; Construct a 3D convolutional neural network, set the input as the design lines and size information of the front view, side view, top view and cross-sectional view, output as 3D point cloud data, define the mean square error as the loss function, and use the Adam optimizer to optimize the parameters of the 3D convolutional neural network; The extracted design lines and dimension information are input into the trained 3D convolutional neural network to obtain the 3D point cloud data of the design scheme; The Poisson surface reconstruction algorithm is used to generate an initial three-dimensional cabinet model according to the three-dimensional point cloud data of the design scheme, and the simplification algorithm is applied to simplify the initial three-dimensional cabinet model.

6. The method for generating a 3D cabinet based on image recognition according to claim 5, characterized in that: The display of the designed three-dimensional cabinet model refers to identifying the designer based on the watermark in the design hand-drawn sketch, outputting the initial three-dimensional cabinet model and the designed three-dimensional cabinet model, and collecting feedback from the designer to optimize the details of the designed three-dimensional cabinet model until it fully meets the designer's requirements.

7. The method for generating a 3D cabinet based on image recognition according to claim 6, characterized in that: The storing in the database refers to obtaining the designer's identity, design hand-drawn sketches, and design three-dimensional cabinet models to form a data set and storing it in the database, and classifying and marking them according to the designer's identity and time, regularly performing integrity checks on the data stored in the database, and synchronously uploading the stored data to the cloud for backup after the check is completed.

8. A computer device comprising: Memory and processor; The memory stores a computer program, wherein the processor implements the steps of the method for generating a 3D cabinet based on image recognition according to any one of claims 1 to 7 when executing the computer program.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for generating a 3D cabinet based on image recognition according to any one of claims 1 to 7 are implemented.

Images