Ecological board identification method and system

Abstract

The invention discloses an ecological board identification method and system, and relates to the technical field of intelligent perception, and the method comprises the steps: multiplying a stable mask by a height map in multi-modal data to obtain an effective height map, constructing a micro-fluctuation consistency vector, constructing a structure coupling vector for a response signal in the multi-modal data, and carrying out the recognition of an ecological board. Collecting a reference parameter set of a historical ecological board, and calculating a consistency score in combination with a process fingerprint vector, a micro-fluctuation consistency vector and a structure coupling vector to obtain a recognition result of the ecological board; gating fusion is performed on deviation degrees and reliability quantities of a texture domain, a micro-fluctuation domain and a knocking structure response domain, a consistency scoring mechanism is constructed, the reliability and engineering applicability of an ecological board identification result are improved, a texture reference map is constructed based on multi-frame gray median values, and the reliability of the ecological board identification result is improved. And stable image blocks are screened in combination with discrete variance of a sliding window, a highlight proportion and connected component constraints, so that stable extraction of the real process texture of the ecological board is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing technology, and in particular to a method and system for identifying ecological boards. Background Technology

[0002] Eco-friendly boards, as composite boards characterized by low formaldehyde release, good stability of decorative layers, and high structural consistency, are widely used in furniture manufacturing, interior decoration, and prefabricated buildings. As industrial vision technology gradually expands from single-frame imaging to multi-frame fusion, local statistical analysis, and texture direction modeling, it is used to characterize surface patterns, pressing processes, and defect morphology. Three-dimensional surface measurement has also evolved from simple roughness assessment to statistical distribution and spatial consistency characterization, used to depict micro-undulations and processing stability. Acoustic and vibration detection has evolved from traditional empirical methods of identifying boards by listening to sound to quantitative analysis based on parameters such as short-time energy attenuation, spectral peaks, and bandwidth, used to reflect the internal structure of the board, adhesive layer coupling, and resonance characteristics, in order to form a multi-source evidence chain.

[0003] However, existing technologies still have shortcomings. Although existing ecological board identification methods collect information from multiple sources such as texture, height, or impact at the same time and use independent thresholds or simple rules for judgment, they lack dynamic constraints and consistency fusion on the reliability of data in each domain. They are prone to misjudgment due to anomalies in a single feature. The analysis of board texture in existing technologies is based on grayscale statistics or feature extraction of a single frame or a small number of images. It is easily affected by changes in lighting, high light reflection, and local noise, and it is difficult to stably reflect the true process texture characteristics of the decorative layer of ecological board. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for identifying eco-friendly boards, which solves the problem that although existing eco-friendly board identification methods simultaneously collect multi-source information such as texture, height, or impact, and use independent thresholds or simple rules for judgment, they lack dynamic constraints and consistency fusion on the reliability of data in each domain, and are prone to misjudgment due to anomalies in a single feature. In the prior art, the analysis of board texture is based only on grayscale statistics or feature extraction of a single frame or a small number of images, which is easily affected by changes in lighting, high light reflection, and local noise, and is difficult to stably reflect the true process texture characteristics of the decorative layer of eco-friendly boards.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for identifying ecological boards, comprising, Extract the geometric center region of the ecological board to be tested and set the detection area. Collect multimodal data, divide the texture reference map in the multimodal data into equal-length windows to generate image blocks, calculate the index of each image block, screen stable image blocks, determine the adjacency relationship between stable image blocks, construct connected components, take the largest connected component to generate an initial binary mask, obtain a stable mask through morphological closing operation, and multiply it with the texture reference map to obtain a stable texture map. The stable texture map is subjected to grayscale erosion to obtain the marker map. The reconstruction result is obtained by geodesic dilatation reconstruction, and the residual map is calculated. The quadrilateral mesh point set is constructed and rotated to align. The pressing driving term and pressing anomaly response term are constructed. The original estimate is obtained by weighted least squares estimation. The final estimate is obtained by TV regular solver. The deviation intensity and deviation continuity values ​​are calculated to generate the process fingerprint vector. The effective height map is obtained by multiplying the stable mask with the height map in the multimodal data. A micro-undulation consistency vector is constructed, and a structural coupling vector is constructed from the response signal in the multimodal data. A set of benchmark parameters of historical ecological boards is collected. By combining the process fingerprint vector, the micro-undulation consistency vector, and the structural coupling vector, the consistency score is calculated to obtain the identification result of the ecological board.

[0007] As a preferred embodiment of the ecological board identification method of the present invention, the step of extracting the geometric center region of the ecological board to be tested and setting a detection area, and collecting multimodal data includes: The geometric center region of the ecological board under test is extracted and set as the detection area. Multimodal data of the detection area are collected, including texture reference map, height map and response signal.

[0008] As a preferred embodiment of the ecological board identification method of the present invention, the step of obtaining a stable mask through morphological closing operation and multiplying it with a texture reference map to obtain a stable texture map includes: The texture reference image is normalized to obtain a normalized grayscale image, which is then divided into image blocks. The index of the image blocks is calculated, and stable image blocks are obtained by filtering the index. Connected components are extracted from the stable image blocks, sorted in descending order, and the largest connected component is selected. If any pixel in the texture reference map is located in the area covered by the stable image block corresponding to the largest connected component, it is marked as 1; otherwise, it is marked as 0, thus obtaining an initial binary mask. Morphological closing operation is applied to the initial binary mask to smooth it, resulting in a stable mask. The stable mask and the texture reference map are multiplied pixel by pixel to obtain a stable texture map.

[0009] As a preferred embodiment of the identification method for the ecological board described in this invention, the step of obtaining the final estimated value using a TV regular solver, calculating the deviation intensity and deviation continuity value, and generating a process fingerprint vector includes: Based on the stable texture map, a grayscale erosion operator is used to perform erosion to obtain a marker map, and the stable texture map is set as a constraint mask map; The marker map is set as the initial reconstruction result, geodesic dilatation is used for reconstruction, and the residual map is calculated based on the reconstruction result and the constraint mask map. The texture direction is calculated based on the constraint mask image, a quadrilateral mesh point set is constructed, and the texture direction is used to rotate and align the quadrilateral mesh point set to obtain the rotated quadrilateral mesh point set. Based on the rotated quadrilateral grid point set, the main structure and residual of the grid domain are calculated, and the pressing driving term and the pressing abnormal response term are constructed respectively. The original estimate is obtained by weighted least squares estimation. The original estimate is calculated by TV regular solver. Based on the final estimate, the deviation intensity and deviation continuity values ​​are calculated, and the process fingerprint vector is constructed.

[0010] As a preferred embodiment of the ecological board identification method of the present invention, the step of multiplying the stable mask with the height map in the multimodal data to obtain an effective height map and constructing a micro-undulation consistency vector includes: The effective height map is obtained by multiplying the stable mask and the height map pixel by pixel. All pixels in the effective height map are traversed, and the pixels with a value of 1 are extracted. The height values ​​of the pixels are summed and averaged to obtain the average height. The root mean square roughness of the height of the pixels is then calculated using RMS. The skewness coefficient and kurtosis coefficient are calculated based on the height value and average height of the pixels. The average height, root mean square roughness, skewness coefficient and kurtosis coefficient are then concatenated to obtain the micro-undulation consistency vector.

[0011] As a preferred embodiment of the ecological board identification method of the present invention, the step of constructing a structural coupling vector by filtering the response signal in the multimodal data includes: The response signal is filtered using a bandpass filter to obtain a filtered signal. The filtered signal is then subjected to overlapping framing to obtain H frames. The short-time energy of each frame is calculated using the short-time energy formula, and the attenuation coefficient is calculated for each short-time energy. Perform an FFT operation on the filtered signal to obtain the amplitude spectrum. Perform median filtering on the amplitude spectrum to obtain the smoothed noise floor curve. Calculate the difference between the amplitude spectrum and the smoothed noise floor curve to obtain the residual spectrum. Extract the median of the residual spectrum. Calculate the absolute value of the difference between the residual spectrum and the median. Extract the median of the absolute value and multiply it by a correction constant to obtain the noise scale. Local maxima are obtained by peak detection on the amplitude spectrum. The significance is obtained by combining the noise scale. The significance is then filtered to obtain the main resonant frequency. The half-power bandwidth is then detected to obtain the half-power bandwidth. The structural coupling vector is obtained by performing a feature splicing operation on the attenuation coefficient, half-power bandwidth, and main resonant frequency.

[0012] As a preferred embodiment of the ecological board identification method of the present invention, the step of collecting a set of benchmark parameters of historical ecological boards, combining process fingerprint vectors, micro-fluctuation consistency vectors, and structural coupling vectors, and calculating a consistency score to obtain the ecological board identification result includes: The baseline parameter set of historical ecological boards is obtained through the API interface, and the deviation and normalized gating quantity are calculated by combining the process fingerprint vector, micro-ripple consistency vector and structural coupling vector. Multiplication is performed on the deviation degree and the corresponding normalized gating quantity. The multiplication results are summed and negative values ​​are taken. The consistency score is obtained by applying the natural exponential function to the negative value. If the consistency score is greater than or equal to the consistency score threshold, it is determined to be an ecological board; otherwise, it is determined to be a non-ecological board.

[0013] Secondly, the present invention provides an identification system for ecological boards, comprising, The detection area positioning and multimodal data acquisition module is used to locate the geometric center of the ecological board under test as the detection area, and uses an industrial camera and microphone to collect its texture reference map, height map and impact response signal respectively; The stable texture map generation module is used to process the texture baseline map, filter and analyze stable image patches to create a stable mask, and finally generate a stable texture map. The process fingerprint vector construction module is used to construct a process fingerprint vector based on a stable texture map through a series of calculations, including erosion, dilation reconstruction, residual map analysis, and other steps, combined with texture direction information. The micro-undulation consistency vector construction module is used to determine the effective height map using a stable mask and height map, and extract the average height, root mean square roughness, skewness coefficient and kurtosis coefficient from it to construct the micro-undulation consistency vector. The structural coupling vector construction and ecological board discrimination module is used to process the impact response signal to obtain the main resonance frequency and half-power bandwidth, and calculate the deviation degree by combining other features and determine whether it is an ecological board through the consistency score.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the method for identifying the ecological board as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for identifying the eco-board as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: This invention obtains an effective height map by multiplying a stable mask with the height map in multimodal data, constructs a micro-undulation consistency vector, constructs a structural coupling vector from the response signals in multimodal data, collects a set of benchmark parameters of historical ecological boards, and calculates a consistency score by combining the process fingerprint vector, the micro-undulation consistency vector, and the structural coupling vector to obtain the identification result of the ecological board; it improves the reliability and engineering applicability of the ecological board identification result and achieves stable extraction of the real process texture of the ecological board. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the method for identifying the ecological board in Example 1.

[0019] Figure 2 This is a schematic diagram of the ecological board identification system in Example 1. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] 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 phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a method for identifying ecological boards, including the following steps: S1. Extract the geometric center region of the ecological board to be tested and set the detection area. Collect multimodal data. Divide the texture reference map in the multimodal data into equal-length windows to generate image blocks. Calculate the index of each image block. Filter stable image blocks and determine the adjacency relationship between stable image blocks. Construct connected components. Take the largest connected component to generate an initial binary mask. Obtain the stable mask through morphological closing operation and multiply it with the texture reference map to obtain a stable texture map. Specifically, the geometric center region of the ecological board to be tested is extracted and a detection area is set. Multimodal data is collected, including: The geometric center region of the ecological board under test is extracted (using, for example, the centroid method) and set as the detection area, with a size of, for example, 120mm × 120mm. The two-dimensional coordinate system (X, Y) of the ecological board under test is defined with the center point of the detection area as the origin. Multimodal data is collected, including collecting K consecutive frames of images of the detection area using an industrial camera, and calculating the median value of the pixel grayscale of the K frames for any pixel position to obtain a texture reference map. The height map of the detection area is collected using an industrial camera, and a flexible tapping device is used to tap the origin. The response signal is collected through a microphone.

[0024] By using the geometric center of the ecological board under test as the sole detection area and establishing a unified two-dimensional coordinate reference system within this area, all subsequent texture information, micro-undulation information, and dynamic response information are collected and analyzed under the same spatial reference. This avoids non-process-related interference caused by differences in edge processing, cutting errors, or uneven splicing of decorative layers. Multiple consecutive frames of images are used to generate a texture reference map using pixel-level median statistics, effectively suppressing the influence of instantaneous illumination fluctuations, surface micro-reflections, and random noise on texture representation. This makes the obtained texture characterization closer to the actual decorative layer process state of the ecological board. The height map and the impact response signal are acquired synchronously within the same detection area, achieving spatially consistent sampling of static geometric features and dynamic mechanical responses. This provides a comparable and alignable data foundation for subsequent cross-modal correlation analysis, thereby significantly improving the stability and repeatability of the overall test results.

[0025] Furthermore, a stable mask is obtained through morphological closing operations, and then multiplied with the texture reference map to obtain a stable texture map, including: The texture reference image is subjected to grayscale normalization to obtain a normalized grayscale image; Set equal-length windows and sliding steps (the same in both horizontal and vertical directions). Starting from the top left corner of the normalized grayscale image, divide the image into n image blocks (each window corresponds to one image block) based on the sliding steps and the window size. Calculate metrics for each image patch, including mean gray level, variance, and highlight ratio; Extract the normalized grayscale values ​​of all pixels in the Cth image block, sum them, and then divide by the number of pixels in the Cth image block to obtain the mean grayscale value. Calculate the difference between the normalized gray level and the mean gray level of the Cth image block. After squaring the difference, iterate through all pixels and sum the squares. Divide the sum by the number of pixels in the Cth image block to obtain the discrete variance. Sort the normalized grayscale values ​​of the Cth image block in descending order, filter the maximum normalized grayscale value, calculate the product of the maximum normalized grayscale value and the scaling factor (set based on empirical rules) to obtain the solidification threshold. If the normalized grayscale value is greater than the solidification threshold, it is marked as a highlight hit; otherwise, it is marked as 0. Iterate through all the normalized grayscale values ​​of the Cth image block to obtain the highlight hit count. Divide the highlight hit count by the number of pixels in the Cth image block to obtain the highlight ratio. Based on the fixed threshold method, a highlight ratio threshold is set. If the highlight ratio is less than the highlight ratio threshold and the discrete variance is greater than or equal to the minimum value of the discrete variance and less than or equal to the maximum value of the discrete variance (obtained by sorting the discrete variance in descending order), then the Cth image block is marked as a stable image block. Otherwise, it is deleted. All image blocks are traversed to obtain all stable image blocks. If any stable image block Y satisfies any of the following conditions with other stable image blocks, then the other stable image blocks are determined to be adjacent image blocks. The conditions include: if the absolute value of the difference between the x-coordinates of any two stable image blocks is equal to the sliding step size and the absolute value of the difference between the y-coordinates is equal to 0, they are determined to be adjacent in the horizontal direction; if the absolute value of the difference between the x-coordinates of any two stable image blocks is equal to 0 and the absolute value of the difference between the y-coordinates is equal to the sliding step size, they are determined to be adjacent in the vertical direction; if the absolute value of the difference between the x-coordinates of any two stable image blocks is equal to the sliding step size and the absolute value of the difference between the y-coordinates is equal to the sliding step size, they are determined to be adjacent in the diagonal direction. Traverse all other stable image blocks to obtain all adjacent image blocks of stable image block Y, which are the connected components. Repeat the above operation for all stable image blocks to obtain the connected components of all stable image blocks. The connected components of all stable image patches are sorted in descending order according to the number of stable image patches in the connected components to obtain the largest connected component. If any pixel in the texture reference map is located in the area covered by the stable image block corresponding to the largest connected component, it is marked as 1; otherwise, it is marked as 0. This yields the initial binary mask. The initial binary mask is then smoothed using morphological closing operations to obtain the stable mask. A stable texture map is obtained by multiplying the stable mask and the texture reference map pixel by pixel.

[0026] By introducing a stable image patch selection mechanism based on statistical characteristics onto the texture baseline map, texture analysis no longer relies on single-point or global mean, but instead shifts to a systematic evaluation of local texture stability. This effectively distinguishes between genuine process textures in the decorative layer and pseudo-texture regions caused by specular highlights, reflections, or local contamination. Combining connectivity analysis and morphological closing operations, only spatially continuous and predominantly stable regions are retained, avoiding interference from scattered noise blocks on texture determination results. This ensures the integrity and consistency of the spatial structure of the texture regions ultimately involved in the analysis. The pixel-by-pixel fusion of the stable mask and the texture baseline map ensures that all subsequent texture-related operations are limited to reliable regions. From an engineering perspective, this achieves adaptive purification of texture data, laying a solid foundation for the extraction of fine process features.

[0027] S2. Perform grayscale erosion on the stable texture map to obtain the marker map, obtain the reconstruction result through geodesic dilatation reconstruction, calculate the residual map, construct and rotate the aligned quadrilateral mesh point set, construct the pressing driving term and the pressing anomaly response term, obtain the original estimate through weighted least squares estimation, obtain the final estimate using the TV regular solver, calculate the deviation intensity and deviation continuity value, and generate the process fingerprint vector. Specifically, the TV regularized solver is used to obtain the final estimate, calculate the deviation intensity and deviation continuity values, and generate a process fingerprint vector, including: Based on the stable texture map, a grayscale erosion operator is used to perform erosion to obtain the marker map. The stable texture map is then set as the constraint mask map, and the formula is as follows:

[0028] in, For the labeled image, For pixel coordinates, For grayscale erosion operators, To stabilize the texture map; The marker map is set as the initial reconstruction result, and geodesic dilatation is used for reconstruction to obtain the reconstruction result, using the following formula:

[0029] in, For iteration The reconstruction results For iteration The reconstruction results To constrain the mask image, This is the grayscale dilation operator; The residual map is calculated based on the reconstruction results and the constraint mask map, using the following formula:

[0030] in, For residual graphs, For the reconstruction result; Texture orientation is calculated based on the constraint mask image, and a quadrilateral mesh point set is constructed. The texture orientation is then used to rotate and align the quadrilateral mesh point set to obtain the rotated quadrilateral mesh point set. The formula is as follows:

[0031] in, For the structure tensor, For a quadrilateral grid point set, and Horizontal direction and vertical direction The gradient is obtained by applying the Sobel operator to the constraint mask map. The stable pixel set represents all pixels marked as 1 in the stable texture map. For texture direction, The largest eigenvalue, The coordinates of the grid vertices are obtained by dividing the circumscribed matrix into equal parts. and For row indexes and column indexes, and For the number of grid rows and columns, and These are the rotated x and y coordinates. and The x and y coordinates of the grid vertices; The outer matrix refers to the outer matrix obtained by filtering the maximum and minimum x and y coordinates of the stable texture map. Within this outer matrix, the area is divided into equal intervals to create a grid. This involves calculating the difference between the maximum and minimum x and y coordinates to obtain the horizontal length, and calculating the difference between the maximum and minimum y coordinates to obtain the vertical length. The horizontal grid step size is defined as the ratio of the horizontal length to the vertical length minus 1, and the vertical grid step size is defined as the ratio of the vertical length to the horizontal length minus 1. Therefore, the formula for the grid vertex coordinates is:

[0032] in, and Let x and y be the coordinates of the grid vertices. and The minimum x-coordinate and minimum y-coordinate, and The horizontal grid step size and the vertical grid step size; Based on the rotated quadrilateral grid point set, the principal structure and residual of the grid domain are calculated (using bilinear interpolation), as follows:

[0033] in, For the main structure of the grid domain, Residual of the grid domain and These are the rotated x and y coordinates; Based on the main structure of the mesh domain and the mesh domain remnants, the bonding driving term and the bonding anomaly response term are constructed respectively, with the following formula:

[0034] in, For the bonding drive term, the formula for calculating the bonding anomaly response term is the same as that for calculating the bonding drive term. Change to ; Based on the neighborhood of the center point in the rotated quadrilateral grid (with radius r, expanding outwards by r centered on the center point), traverse each point in the neighborhood, extract the bonding driver term and bonding anomaly response term for each point, and obtain the original estimated value through weighted least squares estimation, as shown in the formula:

[0035] in, Let be the objective function. The original estimated value to be calculated, Let x and y be the x and y coordinates of the points within the domain. The neighborhood of the center point in the rotated quadrilateral grid set. As weight, To address the issue of abnormal response items, The original estimate of the quadrilateral grid point set after rotation is obtained by minimizing the objective function. It is a very small positive number; The original estimates are used to calculate the final estimate using the TV regular solver (ADMM iterative solver), and the formula is as follows:

[0036] in, This is the final estimate of the quadrilateral grid point set after rotation. For regularization weights; The deviation intensity and deviation continuity values ​​are calculated based on the final estimated values, using the following formulas:

[0037] in, and These represent the deviation intensity and deviation continuity values; The formula for calculating continuity score is:

[0038] in, For continuous scoring, The number of pixels in a stable texture map; After normalizing the continuity score, the mean of the reconstructed main structure, the standard deviation of the reconstructed main structure, the deviation intensity, the deviation continuity value, and the directional consistency vector, a vector splicing operation is performed to obtain the process fingerprint vector.

[0039] By separating the main structure and residuals based on morphological reconstruction of stable texture maps, the surface texture of eco-boards is clearly divided into overall structural components dominated by the lamination process and detailed deviation components caused by local anomalies. This breaks through the limitations of traditional texture analysis, which is limited to the statistical level. By introducing a mesh modeling method with texture direction alignment, different boards can still be compared and analyzed in a unified structural coordinate system even when there are differences in texture direction. This significantly improves the comparability of cross-sample process consistency assessment. Combined with local weighted estimation and total variational regularization constraints, the true change trend driven by the process is preserved while suppressing random noise. This makes the final estimation result have both smoothness and structural preservation ability. The process fingerprint vector constructed in this way can not only characterize the overall process level of the eco-board decorative layer, but also reflect the intensity and continuity of local process deviations, and has high engineering interpretability and discrimination stability.

[0040] S3. Multiply the stable mask with the height map in the multimodal data to obtain the effective height map, construct the micro-ripple consistency vector, construct the structural coupling vector for the response signal in the multimodal data, collect the benchmark parameter set of historical ecological boards, and combine the process fingerprint vector, micro-ripple consistency vector and structural coupling vector to calculate the consistency score and obtain the identification result of the ecological board. Specifically, the stable mask is multiplied with the height map in the multimodal data to obtain the effective height map, and a micro-undulation consistency vector is constructed, including: The effective height map is obtained by multiplying the stable mask and the height map pixel by pixel. Traverse all pixels in the effective height map, extract the pixels with a value of 1, sum and average the height values ​​of the pixels to obtain the average height, and use RMS to calculate the root mean square roughness of the pixel height. The skewness and kurtosis coefficients are calculated based on the pixel height and average height, using the following formula:

[0041] in, and These are the skewness coefficient and the kurtosis coefficient. The number of pixels represented as 1. For pixels The height value, Average height The root mean square roughness; The average height, root mean square roughness, skewness coefficient, and kurtosis coefficient are concatenated to obtain a micro-undulation consistency vector.

[0042] By introducing a stable mask into height map analysis, only height information consistent with the stable texture area is statistically analyzed. This allows micro-undulation features to directly reflect the true geometric shape of the decorative layer within the reliable area, avoiding interference from non-decorative factors such as edge warping and substrate defects on the height analysis results. By employing multi-dimensional statistical features to jointly characterize the surface micro-undulation state, it can not only describe the overall height level and roughness, but also depict the asymmetry and sharpness of the height distribution. This comprehensively reflects the combined effects of process factors such as pressure distribution, substrate springback, and adhesive layer spreading during the lamination process. The resulting micro-undulation consistency vector maintains physical interpretability while exhibiting high resolution for subtle process differences between different batches of eco-boards.

[0043] Furthermore, a structural coupling vector is constructed by filtering the response signals in the multimodal data, including: The response signal is filtered using a bandpass filter (such as an IIR or FIR filter) to obtain the filtered signal. The filtered signal is subjected to overlapping framing to obtain H frames. The short-time energy of each frame is calculated using the short-time energy formula. The attenuation coefficient is calculated for each short-time energy using the following formula:

[0044] in, The attenuation coefficient is... For the first The time of each frame For the first The average time of each frame, For the first Short-time energy of each frame. For short-term energy Perform natural logarithm operations Then calculate the mean. Number of frames; Perform an FFT operation on the filtered signal to obtain the amplitude spectrum. Perform median filtering on the amplitude spectrum to obtain the smoothed noise floor curve. Calculate the difference between the amplitude spectrum and the smoothed noise floor curve to obtain the residual spectrum. Extract the median of the residual spectrum. Calculate the absolute value of the difference between the residual spectrum and the median. Extract the median of the absolute value and multiply it by a correction constant (set based on historical experimental experience) to obtain the noise scale. Local maxima are obtained by peak detection on the amplitude spectrum. The significance is then calculated by combining this with a noise scale, using the following formula:

[0045] in, For significance, For the first A local maximum point, For residual spectrum The noise scale; Filter the significance values ​​that are greater than or equal to the significance threshold (based on the fixed threshold method), sort them in descending order, and filter the frequency corresponding to the maximum significance value, which is then set as the principal resonance frequency. The half-power bandwidth is obtained by detecting the main resonant frequency; The structural coupling vector is obtained by performing a feature splicing operation on the attenuation coefficient, half-power bandwidth, and main resonant frequency.

[0046] By performing targeted filtering and time-frequency joint analysis on the impact response signal, the dynamic attenuation characteristics and resonance behavior of the eco-board under stimulated conditions can be effectively characterized. This can reflect the bonding state between the decorative layer and the substrate and the overall structural integrity. The introduction of attenuation characteristics allows the structural analysis to go beyond frequency position and further reflect the material coupling characteristics during energy dissipation. The combined use of resonance frequency and bandwidth can distinguish between abnormal responses caused by changes in structural stiffness and local defects. This structural coupling vector supplements visual and geometric information from the perspective of mechanical response, forming an indirect perception of the internal structural state of the eco-board and effectively improving the sensitivity of the identification process to hidden process defects.

[0047] Furthermore, a set of baseline parameters for historical eco-friendly boards is collected, and combined with process fingerprint vectors, micro-ripple consistency vectors, and structural coupling vectors, a consistency score is calculated to obtain the identification results of the eco-friendly boards, including: The baseline parameter set of historical ecological boards is obtained through the API interface, including the mean and diagonal variance matrix of the process fingerprint vector, the mean and diagonal variance matrix of the micro-fluctuation consistency vector, and the mean and diagonal variance matrix of the structural coupling vector. The deviation is calculated by using Mahalanobis distance formula to calculate the distance between the mean and diagonal variance matrix of the process fingerprint vector in the benchmark parameter set and the process fingerprint vector in the previous step, the mean and diagonal variance matrix of the micro-fluctuation consistency vector in the benchmark parameter set and the micro-fluctuation consistency vector in the previous step, and the mean and diagonal variance matrix of the structural coupling vector in the benchmark parameter set and the structural coupling vector in the previous step, respectively, to obtain the process fingerprint deviation, micro-fluctuation deviation and structural coupling deviation. The reliability metrics for the three domains are calculated, including the texture domain reliability, the undulation domain reliability (the number of effective height pixels, representing the number of pixels with a value of 1 extracted from all pixels in the effective height map), and the tap domain reliability (obtained using MSE based on the attenuation coefficient). The formula is as follows:

[0048] in, For texture domain reliability, To stabilize the number of image patches, For the first The highlight ratio of a stable image patch; Based on the three-domain reliability quantity, the gated quantity is calculated and normalized to obtain the normalized gated quantity, as shown in the formula:

[0049] in, , as well as This is a gated quantity for the reliability of the texture domain, the reliability of the fluctuation domain, and the reliability of the knock domain. This is the sensitivity coefficient. and For the reliability quantities in the fluctuation domain and the knock domain, This is a reliability measure for the reference fluctuation domain (set based on historical experimental experience). The attenuation coefficient; Multiplication is performed on the deviation degree and the corresponding normalized gating quantity, the multiplication results are summed and negative values ​​are taken, and the consistency score is obtained by applying the natural exponential function to the negative values. If the consistency score is greater than or equal to the consistency score threshold (set based on the fixed threshold method), it is determined to be an ecological board; otherwise, it is determined to be a non-ecological board.

[0050] By introducing a set of historical ecological board benchmark parameters, the multi-domain characteristics of the current ecological board under test can be statistically compared with mature process samples, avoiding the randomness risk brought about by single-sample threshold judgment. Parallel calculation of multi-domain deviation allows the three types of information—texture process, micro-undulation morphology, and structural response—to play independent roles in the judgment process. At the same time, the information contribution of different domains is adaptively adjusted through a reliability gating mechanism, so that the overall judgment can still maintain robustness even in the case of uneven data quality or local anomalies. The final consistency score comprehensively reflects the synergy of the ecological board in terms of appearance process, geometric morphology, and mechanical response, giving the identification results clear engineering basis and statistical support, and significantly improving the accuracy, reliability, and generalizability of ecological board identification.

[0051] This embodiment also provides an identification system for ecological boards, including: The detection area positioning and multimodal data acquisition module is used to locate the geometric center of the ecological board under test as the detection area, and uses an industrial camera and microphone to collect its texture reference map, height map and impact response signal respectively; The stable texture map generation module is used to process the texture baseline map, filter and analyze stable image patches to create a stable mask, and finally generate a stable texture map. The process fingerprint vector construction module is used to construct a process fingerprint vector based on a stable texture map through a series of calculations, including erosion, dilation reconstruction, residual map analysis, and other steps, combined with texture direction information. The micro-undulation consistency vector construction module is used to determine the effective height map using a stable mask and height map, and extract the average height, root mean square roughness, skewness coefficient and kurtosis coefficient from it to construct the micro-undulation consistency vector. The structural coupling vector construction and ecological board discrimination module is used to process the impact response signal to obtain the main resonance frequency and half-power bandwidth, and calculate the deviation degree by combining other features and determine whether it is an ecological board through the consistency score.

[0052] This embodiment also provides a computer device applicable to the method for identifying eco-boards, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for identifying eco-boards as proposed in the above embodiment.

[0053] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0054] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the identification method for the ecological board as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0055] In summary, this invention obtains an effective height map by multiplying a stable mask with the height map in multimodal data, constructs a micro-undulation consistency vector, filters the response signal in the multimodal data to construct a structural coupling vector, collects a set of benchmark parameters for historical eco-boards, and calculates a consistency score by combining the process fingerprint vector, the micro-undulation consistency vector, and the structural coupling vector to obtain the eco-board identification result. This improves the reliability and engineering applicability of the eco-board identification result and achieves stable extraction of the true process texture of the eco-board.

[0056] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method of identifying an ecological board, characterized by: include, Extract the geometric center region of the ecological board to be tested and set the detection area. Collect multimodal data, divide the texture reference map in the multimodal data into equal-length windows to generate image blocks, calculate the index of each image block, screen stable image blocks, determine the adjacency relationship between stable image blocks, construct connected components, take the largest connected component to generate an initial binary mask, obtain a stable mask through morphological closing operation, and multiply it with the texture reference map to obtain a stable texture map. The stable texture map is subjected to grayscale erosion to obtain the marker map. The reconstruction result is obtained by geodesic dilatation reconstruction, and the residual map is calculated. The quadrilateral mesh point set is constructed and rotated to align. The pressing driving term and pressing anomaly response term are constructed. The original estimate is obtained by weighted least squares estimation. The final estimate is obtained by TV regular solver. The deviation intensity and deviation continuity values ​​are calculated to generate the process fingerprint vector. The effective height map is obtained by multiplying the stable mask with the height map in the multimodal data. A micro-undulation consistency vector is constructed, and a structural coupling vector is constructed from the response signal in the multimodal data. A set of benchmark parameters of historical ecological boards is collected. By combining the process fingerprint vector, the micro-undulation consistency vector, and the structural coupling vector, the consistency score is calculated to obtain the identification result of the ecological board.

2. The method for identifying ecological boards as described in claim 1, characterized in that: The process of extracting the geometric center region of the ecological board under test and setting the detection area, and collecting multimodal data includes: The geometric center region of the ecological board under test is extracted and set as the detection area. Multimodal data of the detection area are collected, including texture reference map, height map and response signal.

3. The method for identifying ecological boards as described in claim 2, characterized in that: The process of obtaining a stable mask through morphological closing operations and multiplying it with a texture reference map to obtain a stable texture map includes: The texture reference map is subjected to grayscale normalization to obtain a normalized grayscale map, which is then divided into image blocks. The index of the image blocks is calculated, and stable image blocks are obtained by filtering the index. Connected components are extracted from the stable image blocks, sorted in descending order, and the largest connected component is selected. If any pixel in the texture reference map is located in the area covered by the stable image block corresponding to the largest connected component, it is marked as 1; otherwise, it is marked as 0, thus obtaining an initial binary mask. Morphological closing operation is applied to the initial binary mask to smooth it, resulting in a stable mask. The stable mask and the texture reference map are multiplied pixel by pixel to obtain a stable texture map.

4. The method for identifying ecological boards as described in claim 3, characterized in that: The process of obtaining the final estimate using a TV regularized solver, calculating the deviation intensity and deviation continuity values, and generating a process fingerprint vector includes: Based on the stable texture map, a grayscale erosion operator is used to perform erosion to obtain a marker map, and the stable texture map is set as a constraint mask map; The marker map is set as the initial reconstruction result, geodesic dilatation is used for reconstruction, and the residual map is calculated based on the reconstruction result and the constraint mask map. The texture direction is calculated based on the constraint mask image, a quadrilateral mesh point set is constructed, and the texture direction is used to rotate and align the quadrilateral mesh point set to obtain the rotated quadrilateral mesh point set. Based on the rotated quadrilateral grid point set, the main structure and residual of the grid domain are calculated, and the pressing driving term and the pressing abnormal response term are constructed respectively. The original estimate is obtained by weighted least squares estimation. The original estimate is calculated by TV regular solver. Based on the final estimate, the deviation intensity and deviation continuity values ​​are calculated, and the process fingerprint vector is constructed.

5. The method for identifying ecological boards as described in claim 4, characterized in that: The step of multiplying the stable mask with the height map in the multimodal data to obtain an effective height map and constructing a micro-undulation consistency vector includes: The effective height map is obtained by multiplying the stable mask and the height map pixel by pixel. All pixels in the effective height map are traversed, and the pixels with a value of 1 are extracted. The height values ​​of the pixels are summed and averaged to obtain the average height. The root mean square roughness of the height of the pixels is then calculated using RMS. The skewness coefficient and kurtosis coefficient are calculated based on the height value and average height of the pixels. The average height, root mean square roughness, skewness coefficient and kurtosis coefficient are then concatenated to obtain the micro-undulation consistency vector.

6. The method for identifying ecological boards as described in claim 5, characterized in that: The process of filtering response signals from multimodal data to construct a structural coupling vector includes: The response signal is filtered using a bandpass filter to obtain a filtered signal. The filtered signal is then subjected to overlapping framing to obtain H frames. The short-time energy of each frame is calculated using the short-time energy formula, and the attenuation coefficient is calculated for each short-time energy. Perform an FFT operation on the filtered signal to obtain the amplitude spectrum. Perform median filtering on the amplitude spectrum to obtain the smoothed noise floor curve. Calculate the difference between the amplitude spectrum and the smoothed noise floor curve to obtain the residual spectrum. Extract the median of the residual spectrum. Calculate the absolute value of the difference between the residual spectrum and the median. Extract the median of the absolute value and multiply it by a correction constant to obtain the noise scale. Local maxima are obtained by peak detection on the amplitude spectrum. The significance is obtained by combining the noise scale. The significance is then filtered to obtain the main resonant frequency. The half-power bandwidth is then detected to obtain the half-power bandwidth. The structural coupling vector is obtained by performing a feature splicing operation on the attenuation coefficient, half-power bandwidth, and main resonant frequency.

7. The method for identifying ecological boards as described in claim 6, characterized in that: The collection of historical ecological board benchmark parameter sets, combined with process fingerprint vectors, micro-ripple consistency vectors, and structural coupling vectors, calculates a consistency score to obtain the ecological board identification result, including: The baseline parameter set of historical ecological boards is obtained through the API interface, and the deviation and normalized gating quantity are calculated by combining the process fingerprint vector, micro-ripple consistency vector and structural coupling vector. Multiplication is performed on the deviation degree and the corresponding normalized gating quantity. The multiplication results are summed and negative values ​​are taken. The consistency score is obtained by applying the natural exponential function to the negative value. If the consistency score is greater than or equal to the consistency score threshold, it is determined to be an ecological board; otherwise, it is determined to be a non-ecological board.

8. An ecological board identification system, based on the ecological board identification method according to any one of claims 1 to 7, characterized in that: include, The detection area positioning and multimodal data acquisition module is used to locate the geometric center of the ecological board under test as the detection area, and uses an industrial camera and microphone to collect its texture reference map, height map and impact response signal respectively; The stable texture map generation module is used to process the texture baseline map, filter and analyze stable image patches to create a stable mask, and finally generate a stable texture map. The process fingerprint vector construction module is used to construct a process fingerprint vector based on a stable texture map through a series of calculations, including erosion, dilation reconstruction, residual map analysis, and other steps, combined with texture direction information. The micro-undulation consistency vector construction module is used to determine the effective height map using a stable mask and height map, and extract the average height, root mean square roughness, skewness coefficient and kurtosis coefficient from it to construct the micro-undulation consistency vector. The structural coupling vector construction and ecological board discrimination module is used to process the impact response signal to obtain the main resonance frequency and half-power bandwidth, and calculate the deviation degree by combining other features and determine whether it is an ecological board through the consistency score.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for identifying the ecological board according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for identifying the ecological board according to any one of claims 1 to 7.

Images