A computer-aided design-based system for generating wood grain-like printing roller patterns and textures.
By using a fully automated system, we can solve the problems of low design efficiency, poor naturalness, and process distortion in the generation of wood grain patterns and textures. By adopting multi-scale feature fusion and seamless splicing technology, we can achieve high-quality and fast wood grain texture generation, which can meet the needs of the high-end market for integrated visual and tactile simulation.
Patent Information
- Application Number
- CN202611388417.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-08
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for generating wood grain printing roller patterns suffer from problems such as low design efficiency, poor texture naturalness, process distortion, and discontinuous texture splicing, making it difficult to meet the high-end market's demand for integrated simulation of visual effects and tactile experience.
A fully automated system was built, including modules for texture acquisition, feature extraction, morphological reconstruction, process compensation, and carving data generation. Multi-scale feature fusion, Fourier frequency domain adjustment, and style transfer algorithms were used, combined with seamless stitching technology, to generate high-quality carving data.
Significantly shortens the development cycle of new textures, improves the naturalness and realism of textures, solves the problem of process distortion, and realizes efficient, seamless splicing and three-dimensional depth information output of printing roller engraving, meeting the needs of the high-end market.
Smart Images

Figure CN123116378A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer-aided design technology, and in particular to a system for generating wood grain-like printing roller patterns and textures based on computer-aided design. Background Technology
[0002] With the widespread application of computer-aided design technology, the automated generation of wood grain printing roller patterns and textures has become an important research direction in the printing and packaging industry. As the core execution component of the printing system, the visual quality of the surface graphics of the printing roller directly affects the final printing effect. Wood grain textures, due to their combination of natural beauty and decorative diversity, have long enjoyed continuous growth in market demand in building materials, furniture, packaging and other fields.
[0003] Among them, the wood grain pattern and texture generation technology based on computer-aided design aims to replace traditional manual drawing with digital means to achieve efficient and accurate wood grain texture design and output. Its basic principle is to collect the texture feature data of natural wood, extract features and reconstruct patterns through image processing algorithms, and then map the processed texture units onto the surface of the printing roller to complete the generation of carving data.
[0004] However, existing technologies for generating wood grain roller patterns still have significant shortcomings: First, traditional methods rely excessively on designers' manual drawing and experience-based adjustments, with new texture development cycles lasting 6-8 months, resulting in low design efficiency and an inability to quickly respond to market demands for personalized and diverse wood grain patterns. Second, the generated textures are mostly simple, repetitive two-dimensional patterns, failing to reproduce the unique multi-scale characteristics and natural randomness of natural wood, easily leading to visual fatigue from texture homogenization. Third, the design stage does not fully consider the constraints of roller engraving and printing processes, making the generated textures prone to problems such as dot distortion, blurred edges, and uneven depth after laser engraving. Furthermore, existing splicing algorithms struggle to handle the continuity of complex wood grains, easily resulting in noticeable seams at the circumferential and axial splicing points of the rollers. Finally, this type of technology cannot generate three-dimensional wood grain textures with precise depth information, failing to meet the high-end market's demand for integrated simulation effects of visual and tactile experience. These problems collectively restrict the further development and application of wood grain roller pattern generation technology. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a computer-aided design-based wood grain printing roller pattern and texture generation system. By constructing a fully automated system that integrates high-precision mixed acquisition, multi-scale feature extraction, intelligent generation, full-process chain compensation, and seamless engraving output, it solves the pain points of low design efficiency, poor naturalness, and disconnect between design and production in traditional wood grain printing rollers. This system achieves high-quality and rapid output, meeting the high-end market's demand for integrated visual and tactile simulation effects.
[0006] In a first aspect, the present invention provides a wood grain printing roller pattern texture generation system based on computer-aided design, including a texture acquisition module, a feature extraction module, a morphology reconstruction module, a process compensation module, and an engraving data generation module; The texture acquisition module forms a specific observation angle with the surface of the printing roller substrate through an optical imaging unit, achieves point-by-point scanning with the help of a high-precision displacement guide rail, captures the reflection spectral characteristics of the wood surface using a spectrally optimized imaging sensor, and outputs the original texture image data. The feature extraction module includes an edge detection operator, a multi-scale decomposer, and a texture statistics unit, which are used to extract the contour features, hierarchical distribution, and periodic patterns of wood texture. The morphological reconstruction module employs a deep neural network generator to learn the texture distribution probability of natural wood through adversarial training, and uses latent spatial representation to synthesize diverse wood grain samples to construct two-dimensional planar texture patterns. The process compensation module establishes a dot gain model and a depth decay function based on the material properties of the printing roller and the laser engraving parameters. It then uses an inverse transformation algorithm to compensate for process distortion in the two-dimensional planar texture pattern, generating engravable intermediate data. The engraving data generation module maps the processed texture onto the cylindrical surface of the printing roller through a plate segmentation algorithm, eliminates circumferential and axial seams using seamless splicing logic, and outputs CNC instructions for the engraving machine with depth information.
[0007] Preferably, the texture acquisition module includes an imaging mechanism and a motion mechanism; The imaging mechanism includes a linear array camera or an area array camera, an LED multi-angle illumination array, and a spectral filter group; wherein, the LED multi-angle illumination array is used to eliminate shadows and highlight texture depth; the spectral filter group includes visible light band filters and near-infrared band filters, used to separate texture layers of different depths; The motion mechanism includes an X-axis linear module, a Y-axis linear module, and a rotary table; wherein, the X-axis linear module and the Y-axis linear module constitute a planar scanning coordinate system, which is used to realize full-coverage imaging of the surface to be processed on the printing roller; the rotary table is used to install the printing roller body and drive it to rotate around its own axis to realize circumferential texture splicing.
[0008] Preferably, the texture acquisition module is configured as a hybrid acquisition mode that combines contact and non-contact methods; in contact mode, a high-precision stylus sensor is used to measure the microscopic unevenness data of the roller surface; in non-contact mode, an optical microscope group is used to acquire surface texture images, and the data from the two modes are fused in the feature extraction module.
[0009] Preferably, the feature extraction module includes: The edge detection operator, based on the Canny algorithm or the Sobel operator, extracts the edges of the annual rings of wood texture and obtains the curvature distribution data of the grain lines. The multi-scale decomposer uses Gaussian pyramid or wavelet transform to decompose the original texture image into several sub-band images, which are used to characterize the macroscopic pattern, mesoscopic vessels and microscopic cellular structure of wood texture. The texture statistics generator calculates the contrast, second moment of the angle, entropy, and homogeneity feature parameters of the texture based on the gray-level co-occurrence matrix, and generates a texture feature statistical vector.
[0010] Preferably, the decision-making logic running inside the morphology reconstruction module is based on a multi-scale feature fusion model, and the calculation formula of the multi-scale feature fusion model is: ; In the formula, The generated wood grain pixel values are located at the reconstructed position. The probability distribution at the location; For the first Feature weight coefficients at each scale; For the first The feature tensor after network mapping at each scale is generated by the depth generator's first... The layer output is obtained; The kernel size of the multi-scale feature fusion function; Simultaneously, a condition-variable-based control algorithm is employed to control texture diversity and naturalness generation; the calculation formula for the control algorithm is as follows: ; In the formula, This is a vector of conditional variables, including wood species identifier, annual ring density range, and color offset. is the conditional affine transformation matrix of the generator, which is composed of the parameters of the fully connected layer; As a potential spatial dimension; To generate a noise vector, it must follow a standard normal distribution; This is the scaling parameter for the batch normalization layer, used to stabilize the network training process.
[0011] Preferably, the morphological reconstruction module employs a period consistency preservation algorithm based on Fourier frequency domain adjustment to concentrate the frequency domain energy of the generated texture within the natural periodic range of the annual ring lines; the expression for the period consistency evaluation index is: ; In the formula, The periodic consistency score of the generated texture in the frequency domain is set to [0,1], with the value being closer to 1 indicating a more concentrated periodic energy. Let be the logarithmic power spectrum of the texture image, and ,in, The result is a two-dimensional fast Fourier transform, and a one-dimensional spectrum curve is obtained by taking the radial average. The fundamental frequency of the tree ring texture is obtained by detecting the global peak position of the logarithmic power spectrum; This is the frequency tolerance bandwidth, used to cover the fundamental frequency energy distribution; The Nyquist frequency; The morphological reconstruction module also integrates a texture quality evaluation network, which is used to score the credibility of the intermediate generated textures. The state output equation of the evaluation network is: ; In the formula, Rate the quality of the current texture; For the feature extraction network The layer outputs activation values, and the activation function used is LeakyReLU; This is the downsampling step size of the pooling layer; This is the weight matrix of the fully connected classifier; This is the bias vector.
[0012] Preferably, the morphological reconstruction module is configured with a style transfer unit, which extracts multilayer perceptual features using a pre-trained VGG-19 network, calculates the cocorrelation of style representations using the Gram matrix, and the style loss function is calculated using the following formula: ; In the formula, This is the style loss value, used to measure the similarity between the generated texture and the target wood style; For the first The Gram matrix of the layer feature map is composed of the spatial average of the feature activations; The element indices of the Gram matrix satisfy the relation ; This is the style feature map of the target wood to be matched.
[0013] Preferably, the process compensation module includes a dot gain compensation unit and a depth mapping unit; The network expansion compensation unit establishes a compensation curve based on the improved Murray-Davis model, and the calculation formula is as follows: ; In the formula, The diameter of the compensated dots, The diameter of the dot pattern in the original design. This is the network expansion coefficient; The Gaussian sigma parameter for the laser point energy distribution is obtained through beam quality testing and calibration. The depth mapping unit establishes a non-linear mapping relationship between the engraving depth and the grayscale value. The mapping function uses a Sigmoid function or a gamma correction function, and the expression is: ; In the formula, To normalize the carving depth; Normalized grayscale value; This is the parameter representing the maximum steepness of the curve; This represents the gain center offset of the curve; In addition, the process compensation module is also equipped with a ripple effect suppression algorithm; the compensation formula of the suppression algorithm is: ; In the formula, As the direction compensation factor, The coefficient of difference in hardness between the radial and tangential directions of the wood, as determined in advance; The angle between the engraving scan path and the texture direction is obtained through texture direction analysis.
[0014] Preferably, the engraving data generation module includes a seamless splicing unit; The seamless splicing unit employs a periodic texture extension algorithm along the circumferential direction of the printing roller to maintain the continuity of texture features at both ends. The extension algorithm uses least squares error optimization to find the optimal splicing position. For the axial direction of the printing roller, an overlapping area feature matching algorithm is used to eliminate splicing misalignment. The similarity metric for the matching algorithm uses a structural similarity index, calculated using the following formula: ; In the formula, For texture similarity of overlapping regions, and This represents the average brightness of the two texture images within the overlap region. and For variance, It is a weighted linear combination of covariances.
[0015] Secondly, the present invention also provides a method for generating wood grain-like printing roller patterns and textures based on computer-aided design, using a computer-aided design-based wood grain-like printing roller pattern and texture generation system as described above, the method comprising: S100: The original texture image of the wood specimen is acquired through multi-angle LED illumination and spectral filter group of the texture acquisition module, and the image clarity of each point is ensured by the autofocus mechanism. S200: Extract the outline of the annual ring pattern through the edge detection operator of the feature extraction module, obtain macro, meso and micro three-level texture features through the multi-scale decomposer, and calculate the gray-level co-occurrence matrix feature parameters through the texture statistics unit to form a multi-dimensional feature vector. S300, the morphological reconstruction module calls the pre-trained deep generator network based on the multi-dimensional feature vector, uses conditional variables to control the directionality, annual ring density and color distribution of the generated texture, and maintains periodic consistency through Fourier frequency domain adjustment; S400: The generated two-dimensional planar texture data is transmitted to the process compensation module. First, dot gain compensation is performed, then depth mapping is performed, and finally horizontal stripe effect suppression is performed to obtain the compensated engraving intermediate data. S500 uses the seamless splicing unit of the engraving data generation module to map the texture onto the cylindrical surface of the printing roller, generating CNC engraving instructions with depth grading information to generate wood grain-like printing roller patterns.
[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. By combining contact and non-contact acquisition modes, the texture features of wood at different levels are captured. A multi-scale feature fusion model is used in conjunction with a period consistency preservation algorithm based on Fourier frequency domain adjustment to generate textures. A style transfer unit is integrated for accurate style matching, which significantly shortens the development cycle of new wood grain textures and significantly improves the naturalness and realism of the generated textures. It perfectly replicates the hierarchical features of natural wood from macroscopic patterns to microscopic structures, effectively solving the core contradiction of low efficiency and insufficient texture naturalness of traditional methods.
[0017] 2. By building a built-in full-process chain compensation module, a compensation curve is established based on an improved dot gain model. A nonlinear function is used to construct the mapping relationship between engraving depth and grayscale value. A horizontal grain effect suppression algorithm is designed to correct the engraving depth attenuation caused by wood anisotropy, effectively solving the problem of process distortion in laser engraving. At the same time, it supports adaptive parameter adjustment for various printing roller materials, greatly improving the first-pass yield rate and effectively solving the industry pain point of design and production disconnect.
[0018] 3. By designing a dual seamless splicing algorithm that combines circumferential periodic extension and axial structural similarity matching, and automatically generating corresponding three-dimensional depth information based on multi-scale features, the system achieves global seamless splicing of the roller's circumference and axis. At the same time, it directly outputs CNC commands with depth information that can be recognized by laser engraving machines, breaking through the limitations of traditional two-dimensional design and meeting the demand of the high-end market for integrated simulation effects of "vision + touch". In addition, the system also supports distributed collaborative generation and user interactive adjustment, further enhancing its adaptability to diverse industrial application scenarios. Attached Figure Description
[0019] Figure 1 This is a structural block diagram of a computer-aided design-based wood grain printing roller pattern texture generation system.
[0020] Figure 2 This is a schematic diagram of the hardware structure of the texture acquisition module in a computer-aided design-based wood grain roller pattern texture generation system.
[0021] Figure 3 This is a flowchart illustrating a method for generating wood grain-like printing roller patterns and textures based on computer-aided design. Detailed Implementation
[0022] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. It should be noted that relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.
[0023] Example 1 Please see Figures 1-3 This invention provides a computer-aided design-based wood grain printing roller pattern texture generation system, applicable to high-end customized scenarios in the printing roller engraving and printing packaging industries. It aims to resolve the contradiction between the low efficiency and poor texture naturalness of traditional manual drawing through an automated texture generation system, achieving high-quality and rapid output of wood grain printing roller patterns. The system includes a texture acquisition module, a feature extraction module, a morphology reconstruction module, a process compensation module, and an engraving data generation module. Each module achieves high-bandwidth, low-latency data interaction through a high-speed data bus, and relies on a multi-core processor architecture to ensure real-time computing performance.
[0024] At the hardware architecture level, the texture acquisition module forms a specific observation angle with the surface of the printing roller substrate through the optical imaging unit, realizes point-by-point scanning with the help of high-precision displacement guide rail, captures the reflection spectrum characteristics of the wood surface using a spectrally optimized imaging sensor, ensures imaging clarity through an automatic focusing adjustment mechanism, automatically switches between high-resolution or wide-field acquisition modes in different texture areas, and outputs raw texture image data. Specifically, the texture acquisition module includes an imaging mechanism and a motion mechanism. The imaging mechanism includes a line scan camera or area scan camera, an LED multi-angle illumination array, and a spectral filter group. The LED multi-angle illumination array is configured with multiple light source units distributed in a 45-degree ring to eliminate shadows and highlight texture depth. The spectral filter group includes visible light filters and near-infrared filters to separate texture layers of different depths. The motion mechanism includes an X-axis linear module, a Y-axis linear module, and a rotary table. The X-axis linear module and the Y-axis linear module form a planar scanning coordinate system to achieve full-coverage imaging of the surface to be processed on the printing roller. The rotary table is used to mount the printing roller body and drive it to rotate around its own axis to achieve circumferential texture stitching.
[0025] Furthermore, the texture acquisition module of the present invention is configured as a hybrid acquisition mode that combines contact and non-contact methods. In contact mode, a high-precision stylus sensor is used to measure the microscopic unevenness data of the roller surface. In non-contact mode, an optical microscope group is used to acquire surface texture images. The data from the two modes are fused in the feature extraction module to achieve complete capture of wood texture features.
[0026] The feature extraction module, as the core of the system's front-end processing, integrates an edge detection operator, a multi-scale decomposer, and a texture statistical analyzer. These components extract the contour features, hierarchical distribution, and periodic patterns of the wood texture, respectively, and transmit the extracted feature vector data to the morphological reconstruction module. The edge detection operator, based on the Canny algorithm or the Sobel operator, extracts the edges of the annual rings of the wood texture, obtaining the curvature distribution data of the grain lines. The multi-scale decomposer uses Gaussian pyramids or wavelet transforms to decompose the original texture image data into sub-band images of at least three scales, used to characterize the macroscopic grain, mesoscopic vessels, and microscopic cellular structure of the wood texture. The texture statistical analyzer calculates the contrast, second moment of the angle, entropy, and homogeneity parameters of the texture based on the gray-level co-occurrence matrix, generating a texture feature statistical vector.
[0027] The morphology reconstruction module, as the core generation unit of the system, employs a deep neural network generator. Through adversarial training, it learns the probability distribution of natural wood textures and utilizes latent spatial representation to synthesize diverse wood grain samples, constructing two-dimensional planar texture patterns. The decision-making logic within the morphology reconstruction module is based on a multi-scale feature fusion model, the relevant calculation formulas of which are as follows: ; In the formula, The generated wood grain pixel values are located at the reconstructed position. The probability distribution at a given location serves as the core output for texture synthesis. For the first The feature weight coefficients at each scale represent the proportion of contribution of that scale to the final texture; For the first The feature tensor after network mapping at each scale is generated by the depth generator's first... The layer output is obtained; The kernel size of the multi-scale feature fusion function is 3×3 or 5×5.
[0028] To precisely control the diversity and naturalness of the generated textures, this invention employs a condition variable-based control algorithm, the calculation formula of which is as follows: ; In the formula, This is a vector of conditional variables, including wood species identifiers, annual ring density ranges, and color offsets. is the conditional affine transformation matrix of the generator, which is composed of the parameters of the fully connected layer; The potential spatial dimension is usually taken as 512 or 1024; To generate a noise vector, it must follow a standard normal distribution; This is the scaling parameter for the batch normalization layer, used to stabilize the network training process.
[0029] Furthermore, the morphological reconstruction module of this invention employs a periodicity preservation algorithm based on Fourier frequency domain adjustment to concentrate the frequency domain energy of the generated texture within the natural periodic range of the annual ring lines; the expression for the periodicity evaluation index is: ; In the formula, To generate a periodic consistency score for the texture in the frequency domain, the value range is [0,1]. The closer the value is to 1, the more concentrated the periodic energy is and the better the periodicity of the annual rings. Let be the logarithmic power spectrum of the texture image, and ,in, The result is a two-dimensional fast Fourier transform, and a one-dimensional spectrum curve is obtained by taking the radial average. The fundamental frequency of the tree ring texture is automatically detected by the global peak position of the logarithmic power spectrum; This is the frequency tolerance bandwidth, typically taken as ±15% of the fundamental frequency, used to cover the energy distribution near the fundamental frequency; It is the Nyquist frequency.
[0030] The morphological reconstruction module also integrates a texture quality evaluation network, which is used to score the credibility of the intermediate generated textures. The state output equation of the evaluation network is: ; In the formula, Outputs the quality score for the current texture, ranging from 0 to 1; For the feature extraction network The layer outputs activation values, and the activation function used is LeakyReLU; This is the downsampling step size for the pooling layer, typically set to 2; This is the weight matrix of the fully connected classifier; This is the bias vector.
[0031] The morphological reconstruction module is equipped with a style transfer unit. This unit extracts multilayer perceptual features using a pre-trained VGG-19 network, calculates the cocorrelation of style representations using the Gram matrix, and the style loss function is calculated using the following formula: ; In the formula, This is the style loss value, used to measure the similarity between the generated texture and the target wood style; For the first The Gram matrix of the layer feature map is composed of the spatial average of the feature activations; The element indices of the Gram matrix satisfy the relation ; This is the style feature map of the target wood to be matched.
[0032] The process compensation module, as a crucial link between morphological reconstruction and engraving generation, establishes a dot gain model and a depth attenuation function based on the material properties of the printing roller and laser engraving parameters. It then uses an inverse transformation algorithm to compensate for process distortion in the planar texture pattern, generating engraved intermediate data. The process compensation module includes a dot gain compensation unit and a depth mapping unit. The dot gain compensation unit establishes a compensation curve based on an improved Murray-Davis model, calculated using the following formula: ; In the formula, The diameter of the compensated dots, The diameter of the dot pattern in the original design. The dot gain factor is a dimensionless constant related to the surface roughness of the printing roller and the laser power density, with a value ranging from 1.05 to 1.35. The Gaussian sigma parameter for the laser point energy distribution is obtained through beam quality testing and calibration.
[0033] The depth mapping unit establishes a non-linear mapping relationship between the engraving depth and the grayscale value. The mapping function uses a Sigmoid function or a gamma correction function, and the expression is: ; In the formula, To normalize the carving depth, the range is 0 to 1; The input is a normalized gray value, ranging from 0 to 1; This is the maximum steepness parameter of the curve, with a value ranging from 8 to 15; This is the gain center offset of the curve, usually set to 0.5.
[0034] The process compensation module is also equipped with an algorithm to suppress the horizontal grain effect. Due to the anisotropy caused by the difference in the tracheid arrangement direction in the wood grain, the horizontal grain effect will cause the carving depth to decrease in the horizontal grain direction. The compensation formula of the suppression algorithm is as follows: ; In the formula, As the direction compensation factor, The coefficient of difference in hardness between the radial and tangential directions of the wood is determined in advance and is usually taken as 1.2 to 1.8; The angle between the engraving scan path and the texture direction is obtained through texture direction analysis.
[0035] The engraving data generation module, as the system's output terminal, maps the processed texture onto the cylindrical surface of the printing roller using a panel segmentation algorithm. It then eliminates circumferential and axial seams using seamless splicing logic, outputting CNC commands for the engraving machine with depth information. The engraving data generation module includes a seamless splicing unit. For the circumferential direction of the printing roller, the seamless splicing unit employs a periodic texture extension algorithm to maintain the continuity of texture features at both ends. The extension algorithm uses least squares error optimization to find the optimal splicing position. For the axial direction of the printing roller, an overlapping area feature matching algorithm is used to eliminate splicing misalignment. The similarity metric for the matching algorithm uses a structural similarity index, calculated using the following formula: ; In the formula, For texture similarity of overlapping regions, and This represents the average brightness of the two texture images within the overlap region. and For variance, It is a weighted linear combination of covariances, and the closer the value is to 1, the better the splicing effect.
[0036] The system also features the following auxiliary control functions: Adaptive sampling logic dynamically adjusts the scanning resolution based on the complexity of the texture region, using a sampling density of over 1200 dpi in high-frequency detail areas and 300 to 600 dpi in flat areas to balance scanning efficiency and detail preservation; online calibration logic automatically calls a standard grayscale target for color calibration after each specified number of roller scans, updating the response curve parameters of the imaging sensor using regression analysis; and roller surface projection logic maps the planar texture data to the cylindrical surface of the roller using an equidistant cylindrical projection algorithm. The formula for calculating the projection transformation is: ; In the formula, Let be the arc length coordinates of the printing roller surface. The coordinates of the texture plane are rectangular coordinates. This is the effective radius of the printing roller.
[0037] In practical implementation, this invention pre-sets multiple working modes to address complex texture generation requirements. In the high naturalness mode, priority is given to ensuring the integrity of multi-scale features in the generated texture. By increasing the iteration count of the depth generator and enabling the style transfer unit, a wood grain texture with rich micro-details is obtained. In the high-efficiency mode, a lightweight network structure is adopted and the sampling resolution is reduced to achieve fast texture generation, suitable for large-scale standardized production scenarios.
[0038] The technical effects of the present invention will be quantitatively demonstrated through specific embodiments below.
[0039] In this embodiment, the experimental object is a standard carbon steel printing roller with an effective radius of 100 mm and an effective length of 500 mm. The system uses a texture acquisition module to scan and acquire an oak specimen at a resolution of 1200 dpi.
[0040] The system was set up to scan the oak specimen under multi-angle LED illumination. The texture acquisition module detected an average spacing of 5.2 mm between annual rings on the wood surface, a vessel diameter distribution ranging from 150 to 350 micrometers, and a cell texture scale of approximately 20 to 80 micrometers. After edge detection, the feature extraction module extracted an average annual ring curvature distribution of 0.15 mm. -1 Multi-scale decomposition yields three sub-band images, with a contrast parameter of 4.85 and an angular second moment of 0.62 calculated from the gray-level co-occurrence matrix.
[0041] The morphological reconstruction module retrieves a pre-trained deep generator network based on the feature vectors. Conditional variables are set as follows: wood species is oak, annual ring density is medium-high (3 to 6 rings / cm), and color shift is warm. The central control module drives the generator to perform texture synthesis, at which point the measured multi-scale feature fusion weights are configured as follows: , , The periodic consistency evaluation score reached 0.87, and the output score of the texture quality evaluation network reached 0.91.
[0042] Comparative Case 1: Under the same wood specimen conditions, a traditional fractal geometry-based wood grain generation method was used. Due to the lack of a multi-scale feature fusion mechanism, the generated texture could only represent macroscopic patterns, with mesoscopic vessels and microscopic cellular structures completely missing. The texture quality score was only 0.58, and the periodic consistency score was only 0.42. The generated texture showed obvious digital artificial traces.
[0043] Comparison Case 2: The wood grain texture is generated using traditional hand-drawing methods, and the carving data is adapted based on experience.
[0044] As can be seen from the experimental data in the table, the automatic generation system for wood grain roller patterns provided by this invention, under the conditions of this embodiment, has a new texture development cycle of only 2.5 weeks, which is more than 90% shorter than the 28 weeks in comparative case 2, successfully solving the problem of low efficiency in traditional manual drawing. The cycle consistency score reaches 0.87, indicating that the generated wood grain texture has a cycle distribution characteristic similar to natural wood in the frequency domain; the texture quality score reaches 0.91, indicating that the multi-scale feature fusion and style transfer unit effectively improves the naturalness and realism of the generated texture. The dot distortion rate is controlled at 3.2%, and the depth unevenness is controlled at 2.8%, indicating that the process compensation module effectively solves the process distortion problem after laser engraving. The splicing seam visibility is zero, proving that the seamless splicing unit effectively eliminates the splicing seams in the circumferential and axial directions of the roller. The output of three-dimensional depth information allows the generated engraving data to be directly used for high-end three-dimensional engraving processing, meeting the market's deep demand for integrated simulation effects of visual effects and tactile experience.
[0045] In contrast, the traditional fractal generation method in Comparative Case 1, lacking the ability to model multi-scale features, achieved a texture quality score of only 0.58 and a periodic consistency score of only 0.42, resulting in poor texture naturalness and a tendency to produce texture homogenization. While the traditional hand-drawing method in Comparative Case 2 maintained a certain level of texture naturalness, its development cycle was as long as 28 weeks, failing to meet the requirements for rapid market response. Furthermore, its dot distortion rate and depth unevenness were as high as 12.4% and 15.6% respectively, indicating that process constraints were not fully considered during the design phase, leading to unsatisfactory engraving results.
[0046] Furthermore, to address the compatibility issues of printing rollers made of special materials, the process compensation module of this invention features adaptive compensation functionality. When the system identifies the printing roller material as copper or nickel, the central control module automatically adjusts the dot gain compensation coefficient and the steepness parameters of the depth mapping curve based on the material's hardness characteristics and laser absorption rate parameters, ensuring ideal engraving effects can be achieved on printing rollers made of different materials. Simultaneously, the online calibration logic of the texture acquisition module automatically performs standard grayscale target calibration after each new printing roller is installed, using regression analysis to update the response curve parameters of the imaging sensor, eliminating acquisition errors caused by light source aging or sensor drift.
[0047] In multi-texture collaborative generation applications, this system can be configured with multiple distributed texture generation units for long or wide printing rollers. The morphology reconstruction module employs a spatial partition-based collaborative generation algorithm, which differentiates the texture feature vectors of each partition. The partition located in the middle of the printing roller is primarily responsible for the complete presentation of the main patterns, while the partitions at the ends focus on the texture connection with adjacent printing rollers, thereby maximizing texture consistency at the batch printing roller level. This collaborative logic follows the following control equations: ; In the formula, It is a comprehensive texture quality evaluation index for printing rollers, used to measure the quantitative value of the synergistic effect of multiple zones; This represents the total number of spatial partitions. For the first The weight coefficient of each partition under the current working condition is dynamically allocated according to the degree of influence of the partition on the overall visual effect; This is the texture quality score for the corresponding partition under the current feature configuration.
[0048] In terms of system security, this invention uses a high-speed data bus to communicate with the engraving machine control system in real time. When an abnormal shutdown signal or insufficient air pressure warning is detected, the central control module will trigger the highest priority emergency command. Through the internally preset caching mechanism, all unfinished engraving data will be saved to the safe storage area in a very short time (usually less than 100ms) to ensure that the continuity of subsequent processing is not affected.
[0049] In summary, this invention successfully resolves the contradiction between low efficiency and poor texture naturalness in manual drawing through a precise texture acquisition architecture, multi-dimensional feature extraction algorithms, and a morphological reconstruction strategy based on a rigorous mathematical model. It not only achieves complete capture of wood texture features at the physical level, but also achieves deep integration of feature patterns and process constraints at the algorithm level, providing a brand-new technical approach for high-quality output of printing roller engraving.
[0050] In another extended embodiment of the invention, the morphological reconstruction module is also configured to support user-interactive texture adjustment. When a user requests specific modifications to the generated texture, the system performs a semantic search in the latent space using a feature vector interpolation algorithm to quickly locate the texture region that meets the user's needs and perform local replacement. This combination of user-interactive adjustment and automatic generation constitutes the system's dual creation mode.
[0051] The system also possesses self-learning capabilities based on historical data. Through the texture database module, the system can accumulate and store successfully generated texture samples and their feature parameters. By clustering analysis of this historical data, the system can automatically extract feature template libraries for common wood species, and quickly match the most similar historical templates for fine-tuning when generating new textures, further improving generation efficiency.
[0052] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A computer-aided design-based wood grain-like plate roll pattern texture generation system, characterized by, It includes a texture acquisition module, a feature extraction module, a morphology reconstruction module, a process compensation module, and a carving data generation module; The texture acquisition module forms a specific observation angle with the surface of the printing roller substrate through an optical imaging unit, achieves point-by-point scanning with the help of a high-precision displacement guide rail, captures the reflection spectral characteristics of the wood surface using a spectrally optimized imaging sensor, and outputs the original texture image data. The feature extraction module includes an edge detection operator, a multi-scale decomposer, and a texture statistics unit, which are used to extract the contour features, hierarchical distribution, and periodic patterns of wood texture. The morphological reconstruction module employs a deep neural network generator to learn the texture distribution probability of natural wood through adversarial training, and uses latent spatial representation to synthesize diverse wood grain samples to construct two-dimensional planar texture patterns. The process compensation module establishes a dot gain model and a depth decay function based on the material properties of the printing roller and the laser engraving parameters. It then uses an inverse transformation algorithm to compensate for process distortion in the two-dimensional planar texture pattern, generating engravable intermediate data. The engraving data generation module maps the processed texture onto the cylindrical surface of the printing roller through a plate segmentation algorithm, eliminates circumferential and axial seams using seamless splicing logic, and outputs CNC instructions for the engraving machine with depth information.
2. The computer aided design based wood grain class roll pattern texture generation system of claim 1, wherein, The texture acquisition module includes an imaging mechanism and a motion mechanism; The imaging mechanism includes a linear array camera or an area array camera, an LED multi-angle illumination array, and a spectral filter group; wherein, the LED multi-angle illumination array is used to eliminate shadows and highlight texture depth; the spectral filter group includes visible light band filters and near-infrared band filters, used to separate texture layers of different depths; The motion mechanism includes an X-axis linear module, a Y-axis linear module, and a rotary table; wherein, the X-axis linear module and the Y-axis linear module constitute a planar scanning coordinate system, which is used to realize full-coverage imaging of the surface to be processed on the printing roller; the rotary table is used to install the printing roller body and drive it to rotate around its own axis to realize circumferential texture splicing.
3. The computer-aided design based wood grain-like rule texture generation system of claim 1, wherein, The texture acquisition module is configured as a hybrid acquisition mode that combines contact and non-contact methods. In contact mode, a high-precision stylus sensor is used to measure the microscopic unevenness data of the roller surface. In non-contact mode, an optical microscope group is used to acquire surface texture images. The data from the two modes are fused in the feature extraction module.
4. The wood grain-like printing roller pattern texture generation system based on computer-aided design according to claim 3, characterized in that, The feature extraction module includes: The edge detection operator, based on the Canny algorithm or the Sobel operator, extracts the edges of the annual rings of wood texture and obtains the curvature distribution data of the grain lines. The multi-scale decomposer uses Gaussian pyramid or wavelet transform to decompose the original texture image into several sub-band images, which are used to characterize the macroscopic pattern, mesoscopic vessels and microscopic cellular structure of wood texture. The texture statistics generator calculates the contrast, second moment of the angle, entropy, and homogeneity feature parameters of the texture based on the gray-level co-occurrence matrix, and generates a texture feature statistical vector.
5. The wood grain-like printing roller pattern texture generation system based on computer-aided design according to claim 4, characterized in that, The decision-making logic running inside the morphology reconstruction module is based on a multi-scale feature fusion model, and the calculation formula of the multi-scale feature fusion model is as follows: ; In the formula, The generated wood grain pixel values are located at the reconstructed position. The probability distribution at the location; For the first Feature weight coefficients at each scale; For the first The feature tensor after network mapping at each scale is generated by the depth generator's first... The layer output is obtained; The kernel size of the multi-scale feature fusion function; Simultaneously, a condition-variable-based control algorithm is employed to control texture diversity and naturalness generation; the calculation formula for the control algorithm is as follows: ; In the formula, This is a vector of conditional variables, including wood species identifier, annual ring density range, and color offset. is the conditional affine transformation matrix of the generator, which is composed of the parameters of the fully connected layer; As a potential spatial dimension; To generate a noise vector, it must follow a standard normal distribution; This is the scaling parameter for the batch normalization layer, used to stabilize the network training process.
6. The wood grain-like printing roller pattern texture generation system based on computer-aided design according to claim 5, characterized in that, The morphological reconstruction module employs a period consistency preservation algorithm based on Fourier frequency domain adjustment to concentrate the frequency domain energy of the generated texture within the natural periodic range of the annual ring lines; the expression for the period consistency evaluation index is: ; In the formula, The periodic consistency score of the generated texture in the frequency domain is set to [0,1], with the value being closer to 1 indicating a more concentrated periodic energy. Let be the logarithmic power spectrum of the texture image, and ,in, The result is a two-dimensional fast Fourier transform, and a one-dimensional spectrum curve is obtained by taking the radial average. The fundamental frequency of the tree ring texture is obtained by detecting the global peak position of the logarithmic power spectrum; This is the frequency tolerance bandwidth, used to cover the fundamental frequency energy distribution; The Nyquist frequency; The morphological reconstruction module also integrates a texture quality evaluation network, which is used to score the credibility of the intermediate generated textures. The state output equation of the evaluation network is: ; In the formula, Rate the quality of the current texture; For the feature extraction network The layer outputs activation values, and the activation function used is LeakyReLU; This is the downsampling step size of the pooling layer; This is the weight matrix of the fully connected classifier; This is the bias vector.
7. The wood grain-like printing roller pattern texture generation system based on computer-aided design according to claim 6, characterized in that, The morphological reconstruction module is equipped with a style transfer unit. This style transfer unit extracts multilayer perceptual features using a pre-trained VGG-19 network, calculates the cocorrelation of style representations using the Gram matrix, and the style loss function is calculated using the following formula: ; In the formula, This is the style loss value, used to measure the similarity between the generated texture and the target wood style; For the first The Gram matrix of the layer feature map is composed of the spatial average of the feature activations; The element indices of the Gram matrix satisfy the relation ; This is the style feature map of the target wood to be matched.
8. The wood grain-like printing roller pattern texture generation system based on computer-aided design according to claim 7, characterized in that, The process compensation module includes a dot gain compensation unit and a depth mapping unit; The network expansion compensation unit establishes a compensation curve based on the improved Murray-Davis model, and the calculation formula is as follows: ; In the formula, The diameter of the compensated dots, The diameter of the dot pattern in the original design. This is the network expansion coefficient; The Gaussian sigma parameter for the laser point energy distribution is obtained through beam quality testing and calibration. The depth mapping unit establishes a non-linear mapping relationship between the engraving depth and the grayscale value. The mapping function adopts a Sigmoid function or a gamma correction function, and the expression is: ; In the formula, To normalize the carving depth; Normalized grayscale value; This is the parameter representing the maximum steepness of the curve; This represents the gain center offset of the curve; In addition, the process compensation module is also equipped with a ripple effect suppression algorithm; the compensation formula of the suppression algorithm is: ; In the formula, As the direction compensation factor, The coefficient of difference in hardness between the radial and tangential directions of the wood, as determined in advance; The angle between the engraving scan path and the texture direction is obtained through texture direction analysis.
9. The wood grain-like printing roller pattern texture generation system based on computer-aided design according to claim 8, characterized in that, The engraving data generation module includes a seamless splicing unit; The seamless splicing unit employs a periodic texture extension algorithm along the circumferential direction of the printing roller to maintain the continuity of texture features at both ends. The extension algorithm uses least squares error optimization to find the optimal splicing position. For the axial direction of the printing roller, an overlapping area feature matching algorithm is used to eliminate splicing misalignment. The similarity metric for the matching algorithm uses a structural similarity index, calculated using the following formula: ; In the formula, For texture similarity of overlapping regions, and This represents the average brightness of the two texture images within the overlap region. and For variance, It is a weighted linear combination of covariances.
10. A method for generating wood grain-like printing roller patterns based on computer-aided design, characterized in that: The method, applied to a computer-aided design-based wood grain roller pattern texture generation system as described in any one of claims 1-9, comprises: S100: The original texture image of the wood specimen is acquired through multi-angle LED illumination and spectral filter group of the texture acquisition module, and the image clarity of each point is ensured by the autofocus mechanism. S200: Extract the outline of the annual ring pattern through the edge detection operator of the feature extraction module, obtain macro, meso and micro three-level texture features through the multi-scale decomposer, and calculate the gray-level co-occurrence matrix feature parameters through the texture statistics unit to form a multi-dimensional feature vector. S300, the morphological reconstruction module calls the pre-trained deep generator network based on the multi-dimensional feature vector, uses conditional variables to control the directionality, annual ring density and color distribution of the generated texture, and maintains periodic consistency through Fourier frequency domain adjustment; S400: The generated two-dimensional planar texture data is transmitted to the process compensation module. First, dot gain compensation is performed, then depth mapping is performed, and finally horizontal stripe effect suppression is performed to obtain the compensated engraving intermediate data. S500 uses the seamless splicing unit of the engraving data generation module to map the texture onto the cylindrical surface of the printing roller, generating CNC engraving instructions with depth grading information to generate wood grain-like printing roller patterns.
