Musical instrument production quality control method and system

Abstract

The invention discloses a musical instrument production quality control method and system, and relates to the technical field of musical instrument quality control. The advanced image processing and analysis technology is adopted, quality control over key links such as wood material selection, the polishing technology and coating detection in the musical instrument production process is achieved, the technical means such as high-resolution imaging, image denoising, the convolutional neural network, laser scanning and optical interference are utilized, wood texture features are accurately quantified, and the quality of the musical instrument is improved. And polishing parameters are optimized, and the coating quality is accurately evaluated, so that the quality stability and the production efficiency of musical instrument production are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of musical instrument quality control, and specifically to a production quality control method and system for musical instruments. Background Art

[0002] During the production process of musical instruments, traditional methods have many problems in the aspects of wood material selection, grinding process, and coating detection. Wood material selection relies on manual experience, and it is difficult to accurately quantify the characteristics of color, texture, and density distribution, resulting in insufficient consistency in material selection. Moreover, it is easily affected by surface defects and light conditions, affecting the reliability of texture classification. In the grinding process, when a laser scanner obtains the three-dimensional data of the body surface, the accuracy is easily affected by the surface reflectivity and ambient light. Especially when the coating is not uniform, the scanning data deviation is large, resulting in inaccurate adjustment of grinding parameters. In the coating detection link, although an optical interferometer can detect the coating thickness and defect distribution, on complex curved surfaces or high-gloss surfaces, the reliability of the detection results is reduced, and it is easily interfered by surface curvature and light scattering, making it difficult to accurately evaluate the coating quality. These problems seriously affect the quality stability and production efficiency of musical instrument production. Summary of the Invention

[0003] The purpose of the present invention is to provide a production quality control method and system for musical instruments, which utilize a variety of high-precision detection and analysis technologies to optimize the quality management process of musical instrument production and ensure the high-quality standards of the final products.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] The present application provides a production quality control method for musical instruments, including the following steps:

[0006] S1. Use a high-resolution camera to obtain the wood surface image, and remove surface defects and light condition interference through an image denoising algorithm; extract the texture features of the denoised image based on a convolutional neural network model to generate a high-precision texture feature vector;

[0007] S2. Input the texture feature vector into a preset texture classification model to obtain the classification result of the wood texture type; use a laser scanner to obtain the three-dimensional data of the body surface, and remove the influence of reflectivity and ambient light through a filtering algorithm;

[0008] S3. Analyze the filtered three-dimensional data, extract the depth and distribution information of the grinding marks, and calculate the adjustment value of the grinding parameters;

[0009] S4. Use an optical interferometer to obtain the interference image of the coating surface, reduce the interference of surface curvature and light scattering through an image enhancement algorithm, and based on the enhanced interference image, use an edge detection algorithm to identify the coating defect area and calculate the defect distribution;

[0010] S5. Combine the defect distribution with the preset coating thickness threshold to judge the qualification of the coating quality, and generate the optimized parameters for the body processing according to the grinding parameter adjustment value and the coating quality judgment result.

[0011] Furthermore, a high-resolution camera is used to obtain the wood surface image, and the surface defects and light condition interference are removed through an image denoising algorithm, specifically including:

[0012] Obtain the original wood surface image through a high-resolution camera, input the original image into the preset image denoising algorithm. When light interference is detected in the image, a filter is used to eliminate the light noise. When surface defects are detected, the defect recognition module marks the defect area, and the marked area is optimized by combining the denoising algorithm to obtain a clear wood surface image. Compare the optimized image with the preset quality standard to judge whether the image meets the requirements, and store the qualified image in the database;

[0013] Among them, removing the surface defects and light condition interference through the image denoising algorithm is based on multi-scale analysis and adaptive filtering technology. When light interference is detected in the image, the image is decomposed into sub-bands of different scales and directions through multi-scale analysis to separate the noise components and detail features in the image; when surface defects are detected in the image, the adaptive filtering technology is used to dynamically adjust the filtering parameters according to the local statistical information of the image.

[0014] Furthermore, based on the convolutional neural network model, extract the texture features of the denoised image to generate a high-precision texture feature vector, specifically including:

[0015] Use the convolutional neural network model to extract the feature of the denoised wood surface image to obtain the texture feature data. According to the extracted texture feature data, generate the texture feature vector, and determine the dimension and numerical distribution of the feature vector;

[0016] Match the generated texture feature vector with the preset texture feature standard library to judge whether the feature vector meets the standard;

[0017] Among them, when the feature vector meets the standard, the feature vector data is stored in the texture feature database. When the feature vector does not meet the standard, optimize the feature extraction process and re-extract the texture feature data;

[0018] Through the optimized feature extraction process, generate a new texture feature vector, perform the standard matching again, and determine the high-precision texture feature vector according to the matching result, and store it in the texture feature database.

[0019] Furthermore, input the texture feature vector into the preset texture classification model to obtain the classification result of the wood texture type, specifically including:

[0020] Using the texture feature vector as the input data, processing it through a preset texture classification model to obtain the classification result of the wood texture type. According to the classification result, it is judged whether the wood texture type matches the preset standard. When it matches, the classification result is stored in the texture type database;

[0021] When the classification result does not match the preset standard, adjust the parameters of the texture classification model, re - classify the texture type. Through the adjusted texture classification model, obtain the classification result of the wood texture type again, and judge whether it matches the preset standard;

[0022] According to the matching result, determine the final classification data of the wood texture type, store it in the texture type database, generate the statistical distribution information of the wood texture type, and determine the characteristic law of the wood texture type according to the statistical distribution information.

[0023] Furthermore, use a laser scanner to obtain the three - dimensional data of the body surface of the instrument, and remove the influence of reflectivity and ambient light through a filtering algorithm. Specifically, it includes:

[0024] Use a laser scanner to perform high - precision scanning on the body of the instrument to obtain the three - dimensional point cloud data of the surface. The scanner emits a laser beam and captures the reflected light, and calculates the three - dimensional coordinates of each point by accurately measuring the flight time or phase difference of the light;

[0025] Then, apply a filtering algorithm to reduce the influence of ambient light and the reflectivity of the body surface on the data. Eliminate the influence of reflectivity through region segmentation and reflectivity correction, and at the same time use light intensity compensation and time filtering techniques to reduce the interference of ambient light;

[0026] Use the processed point cloud data for surface curvature analysis. By calculating the curvature of the surface points, analyze the geometric characteristics of the surface, which is used to identify the unevenness and texture features of the surface, and extract the key features of the body surface of the instrument;

[0027] Based on the features extracted from different perspectives, through data registration technology, align and fuse the data from different perspectives in a unified coordinate system to eliminate the influence brought by perspective differences; according to the fused point cloud data, generate a triangular mesh model of the body surface of the instrument;

[0028] Perform surface smoothing processing on the triangular mesh model of the body surface of the instrument. By analyzing the reconstructed triangular mesh model, detect the defects on the body surface of the instrument, and evaluate the consistency of the surface, and check whether there are uneven or abnormal areas.

[0029] Furthermore, analyze the filtered three - dimensional data, extract the depth and distribution information of the sanding marks, and calculate the adjustment value of the sanding parameters. Specifically, it includes:

[0030] According to the filtered three-dimensional data, use a depth analysis method to extract the depth information of the grinding marks. For the extracted depth information, analyze the distribution characteristics of the grinding marks to obtain distribution information. Conduct a statistical analysis of the extracted distribution characteristics to quantify the distribution characteristics of the marks;

[0031] Based on the depth information and distribution information, calculate the adjustment value of the grinding parameters. According to the adjustment value, determine the specific value of the grinding parameters. When the adjustment value exceeds the preset threshold, recalculate the parameter adjustment value, and finally generate an adjustment plan for the grinding parameters.

[0032] Furthermore, use an optical interferometer to obtain the interference image of the coating surface, and reduce the interference of surface curvature and light scattering through an image enhancement algorithm, specifically including:

[0033] Obtain the optical interference image of the coating surface, extract the optical information data in the image, and use an image enhancement algorithm to process the optical information data to reduce the influence of the curvature value and scattering degree on the image value;

[0034] According to the processed image value, analyze the distribution characteristics of the interference amount on the coating surface. When the interference amount exceeds the preset threshold, reprocess the image value using the enhancement method;

[0035] Based on the processing result, determine a suitable algorithm type and data source matching scheme. Through the matching of the algorithm type and data source, generate the final image data of the coating surface.

[0036] Furthermore, based on the enhanced interference image, use an edge detection algorithm to identify the coating defect area and calculate the defect distribution, specifically including:

[0037] Use an edge detection algorithm to extract the edge value from the enhanced interference image, determine the boundary range of the defect area, segment the defect area according to the edge value, obtain the image value of the defect area, and calculate the area and position of each defect area;

[0038] Use a spatial distribution analysis algorithm to process the image value of the defect area, generate the distribution value of the defect, and judge the aggregation characteristics of the defect. When the distribution value exceeds the preset threshold, use a clustering algorithm to reclassify the defect area to determine the type and quantity of the defect;

[0039] According to the type and quantity of the defect, use a feature extraction algorithm to extract the feature value of the defect from the interference image, generate a detailed description of the defect, and according to the detailed description of the defect, use a classification algorithm to classify the defect to determine the severity of the defect.

[0040] According to the severity of the defect, generate a spatial distribution map of the coating defect and judge the overall quality status of the coating.

[0041] Further, in combination with the defect distribution and a preset coating thickness threshold, the qualification of the coating quality is judged, and based on the adjustment value of the polishing parameters and the judgment result of the coating quality, the optimized parameters for the body processing are generated, specifically including:

[0042] Obtain the defect distribution data, combine it with the preset coating thickness threshold, judge the qualification of the coating quality. When the coating quality is unqualified, determine the adjustment value of the polishing parameters according to the defect distribution data and the thickness distribution data;

[0043] Based on the adjustment value and the judgment result of the coating quality, generate the optimized parameters for the body processing, and then use the clustering algorithm to classify the defect areas to determine the aggregation characteristics and distribution rules of the defects;

[0044] According to the aggregation characteristics and distribution rules of the defects, adjust the adjustment value of the polishing parameters, optimize the body processing parameters, generate a coating quality improvement plan based on the optimized parameters and the judgment result of the coating quality, and generate the final optimized parameters for the body processing according to the improvement plan and the processing parameters.

[0045] The present invention provides a production quality control system for musical instruments, used to implement a production quality control method for musical instruments, including:

[0046] An image acquisition and preprocessing module, which uses a high-resolution camera to obtain the original image of the wood surface, and applies an image denoising algorithm to eliminate the interference of surface defects and light conditions. Through multi-scale analysis and adaptive filtering techniques, the noise components in the image are accurately located and removed;

[0047] A texture feature extraction and classification module, which uses a convolutional neural network model to extract features from the denoised wood surface image to generate high-precision texture feature vectors. By constructing a CNN model containing multiple convolutional layers, pooling layers and fully connected layers, the local and high-level features of the image are extracted, and then the generated texture feature vectors are matched with a preset texture feature standard library to judge whether the feature vectors meet the standards and classify the wood texture types;

[0048] A three-dimensional scanning and surface analysis module, which uses a laser scanner to obtain the three-dimensional data of the body surface, and removes the influence of reflectivity and ambient light through a filtering algorithm. Through techniques such as region segmentation, reflectivity correction, light intensity compensation and temporal filtering to reduce interference, surface curvature analysis is performed using the processed point cloud data, the key features of the body surface are extracted, and a triangular mesh model is generated for surface smoothing processing and defect detection;

[0049] Coating analysis and defect identification module, which uses an optical interferometer to obtain the interference image of the coating surface, reduces the interference of surface curvature and light scattering through an image enhancement algorithm, uses an edge detection algorithm to extract edge values from the enhanced interference image, determines the boundary range of the defect area, calculates the area and position of each defect area, then uses a spatial distribution analysis algorithm to process the image values of the defect area, generates the distribution value of the defect, judges the aggregation characteristics of the defect, and uses a clustering algorithm to reclassify the defect area;

[0050] Quality assessment and parameter optimization module, which combines the defect distribution with the preset coating thickness threshold to judge the qualification of the coating quality, generates the optimized parameters for the body processing according to the grinding parameter adjustment value and the coating quality judgment result, classifies the defect areas through a clustering algorithm, determines the aggregation characteristics and distribution rules of the defects, and adjusts the grinding parameters according to this information, optimizes the body processing parameters, generates a coating quality improvement plan, and generates the final optimized parameters for the body processing according to the improvement plan and the processing parameters.

[0051] The beneficial effects of the present invention are as follows:

[0052] By using a high-resolution camera to obtain the wood surface image and applying an image denoising algorithm to remove surface defects and light condition interference, the problem that traditional wood material selection relies on manual experience and it is difficult to accurately quantify the characteristics of color, texture and density distribution is solved; by using multi-scale analysis and adaptive filtering techniques, such as wavelet transform and NL-Means algorithm, the noise components and detail features are separated from the image, the details of the wood texture are retained, the texture features of the denoised image are extracted based on the convolutional neural network model, a high-precision texture feature vector is generated, and by matching with the preset texture feature standard library, the automatic classification of wood texture types is realized, the reliability and consistency of texture classification are improved, and the quality control level of wood material selection is improved;

[0053] By using a laser scanner to obtain the three-dimensional data of the body surface and applying a filtering algorithm to remove the influence of reflectivity and ambient light, the problem of large scanning data deviation in the grinding process is solved; through region segmentation, reflectivity correction, light intensity compensation and time filtering techniques, the interference of ambient light and surface reflectivity on the data is reduced, the key features of the body surface are extracted by using the processed point cloud data for surface curvature analysis, and a triangular mesh model is generated for surface smoothing processing and defect detection. Based on the filtered three-dimensional data, the depth and distribution information of the grinding marks are extracted, the grinding parameter adjustment value is calculated, so as to optimize the grinding parameters and ensure the quality and consistency of the body surface;

[0054] An optical interferometer is used to obtain the interference image of the coating surface, and an image enhancement algorithm is used to reduce the interference of surface curvature and light scattering, solving the problem of reduced reliability of the detection results in the coating detection link. The edge detection algorithm is used to identify the coating defect area, calculate the defect distribution, and the clustering algorithm is used to reclassify the defect area to determine the type and quantity of the defects. Combining the defect distribution with the preset coating thickness threshold, the qualification of the coating quality is judged, and according to the adjustment value of the grinding parameters and the coating quality judgment result, the optimized parameters for the body processing are generated, not only improving the evaluation accuracy of the coating quality, but also providing a scientific basis for the improvement of the coating quality, thus enhancing the overall quality of the musical instrument. Description of the Drawings

[0055] For better understanding and implementation, the technical solution of the present application will be described in detail below with reference to the drawings.

[0056] Figure 1 It is a schematic flowchart of a method for controlling the production quality of a musical instrument provided in Embodiment 1 of the present application;

[0057] Figure 2 It is a schematic flowchart of a method for controlling the production quality of a musical instrument provided in Embodiment 1 of the present application to generate a high-precision texture feature vector;

[0058] Figure 3 It is a schematic flowchart of a method for controlling the production quality of a musical instrument provided in Embodiment 1 of the present application to obtain the classification result of the wood texture type;

[0059] Figure 4 It is a schematic structural diagram of a system for controlling the production quality of a musical instrument provided in Embodiment 2 of the present application. Detailed Embodiments

[0060] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the exemplary embodiments will be described in detail herein, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of methods and systems consistent with some aspects of the present application as detailed in the appended claims.

[0061] The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The singular forms "a", "the" and "said" used in the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0062] The following will, in conjunction with the accompanying drawings and preferred embodiments, provide a detailed description of the specific implementation manners, features and effects of the present invention.

[0063] Embodiment 1

[0064] Please refer to Figures 1 - 3 , this embodiment provides a method for controlling the production quality of musical instruments, including the following steps:

[0065] S1. Use a high-resolution camera to obtain the wood surface image, and remove surface defects and light condition interference through an image denoising algorithm; extract the texture features of the denoised image based on a convolutional neural network model to generate a high-precision texture feature vector;

[0066] Furthermore, using a high-resolution camera to obtain the wood surface image and removing surface defects and light condition interference through an image denoising algorithm specifically includes:

[0067] Obtain the original wood surface image through a high-resolution camera, input the original image into a preset image denoising algorithm. When light interference is detected in the image, use a filter to eliminate the light noise. When surface defects are detected, mark the defect area through a defect recognition module, and optimize the marked area in combination with the denoising algorithm to obtain a clear wood surface image. Compare the optimized image with the preset quality standard to determine whether the image meets the requirements, and store the qualified image in the database.

[0068] Among them, removing surface defects and light condition interference through the image denoising algorithm is based on multi-scale analysis and adaptive filtering techniques. When light interference (such as reflected light, shadow, etc.) is detected in the image, the image is decomposed into sub-bands of different scales and directions through multi-scale analysis (such as wavelet transform or Contourlet transform) to separate the noise components and detail features in the image, so as to accurately locate and remove the light interference while retaining the details of the wood texture; when surface defects (such as scratches, knots, stains, etc.) are detected in the image, use adaptive filtering techniques (such as NL-Means algorithm or wavelet-domain based adaptive filtering) to dynamically adjust the filtering parameters according to the local statistical information of the image. For example, the NL-Means algorithm adjusts the filter parameters by comparing the similarity between pixels, and can remove noise while retaining texture details; in addition, wavelet-domain adaptive filtering can suppress noise at different scales while retaining the edges and textures of the image; after the above denoising process, a clear wood surface image after denoising is obtained. The image denoising algorithm can effectively remove the light interference and surface defects in the wood surface image while retaining the texture details, significantly improving the image quality.

[0069] Specifically, through high-resolution cameras and image denoising techniques, the quality of wood surface images has been significantly improved, effectively removing light and defect interference, retaining important texture details, and providing accurate data support for wood selection in musical instrument production.

[0070] Furthermore, based on the convolutional neural network model, the texture features of the denoised image are extracted to generate high-precision texture feature vectors, specifically including:

[0071] S11: Use the convolutional neural network model to extract the features of the denoised wood surface image, obtain the texture feature data, generate the texture feature vector according to the extracted texture feature data, and determine the dimension and numerical distribution of the feature vector;

[0072] S12: Match the generated texture feature vector with the preset texture feature standard library to determine whether the feature vector meets the standard;

[0073] Among them, when the feature vector meets the standard, the feature vector data is stored in the texture feature database; when the feature vector does not meet the standard, the feature extraction process is optimized and the texture feature data is re-extracted;

[0074] S13: Through the optimized feature extraction process, generate a new texture feature vector, perform standard matching again, and determine the high-precision texture feature vector according to the matching result, and store it in the texture feature database.

[0075] Use the convolutional neural network model to extract the features of the denoised wood surface image to obtain the texture feature data, specifically including constructing a CNN model for texture feature extraction, including multiple convolutional layers, pooling layers, and fully connected layers.

[0076] Among them, the input layer: receives the denoised wood surface image; the convolutional layer: uses multiple convolutional kernels to perform convolutional operations on the image to extract the local features of the image. The first convolutional layer can use 32 convolutional kernels with a size of 3×3, the second layer uses 64 3×3 convolutional kernels, and the third layer uses 128 3×3 convolutional kernels; the activation function: add a ReLU activation function after each convolutional layer to introduce non-linearity and enhance the expression ability of the model; the pooling layer: uses max pooling operations to downsample the feature map, reduce the size of the feature map, and retain important features at the same time; the fully connected layer: flattens the feature maps extracted by the convolutional layer and the pooling layer into a one-dimensional vector, further extracts high-level features through the fully connected layer, and outputs the texture feature vector.

[0077] In the convolutional layer, the formula for the convolution operation is: C = W * I + b; where W is the convolutional kernel, I is the input image, b is the bias term, and * represents the convolution operation; the convolution operation generates multiple feature maps, each corresponding to a convolutional kernel, capturing different texture features in the image. Apply the ReLU activation function to the convolved feature map: A = ReLU(C); where A represents the feature map after being processed by the ReLU activation function. The ReLU function sets negative values to 0 and retains positive values, enhancing the sparsity of the features; C represents the feature map after the convolution operation. Subsequently, perform downsampling on the feature map through the max pooling operation, retaining the most prominent features: P = MaxPool(A); where P represents the feature map after the max pooling operation, and MaxPool represents the max pooling operation, selecting the maximum value in each pooling window as the output.

[0078] After multiple convolutional and pooling operations, the resulting feature map is flattened into a one-dimensional vector and input into the fully connected layer. The fully connected layer further extracts high-level features and outputs a texture feature vector, and the texture features are represented by the statistical information of the feature response map; by calculating the average value of the feature response map corresponding to each convolutional kernel: where, is the average value of the i-th feature response map.

[0079] Specifically, use a convolutional neural network (CNN) to automatically extract texture features from the denoised wood image, generate an accurate feature vector. By constructing a CNN model containing multiple convolutional layers, pooling layers, and fully connected layers, effectively capture the local and high-level texture features of the wood. After feature extraction, matching a preset standard library, and optimization processing, finally obtain a high-precision texture feature vector and store it in the database, providing reliable data support for wood quality control and musical instrument manufacturing to ensure that the final product meets strict quality requirements.

[0080] S2. Input the texture feature vector into a preset texture classification model to obtain the classification result of the wood texture type; use a laser scanner to obtain the three-dimensional data of the body surface, and remove the influence of the reflectivity and ambient light through a filtering algorithm;

[0081] Furthermore, input the texture feature vector into a preset texture classification model to obtain the classification result of the wood texture type, which specifically includes:

[0082] S21. Use the texture feature vector as input data, process it through a preset texture classification model to obtain the classification result of the wood texture type. According to the classification result, judge whether the wood texture type matches the preset standard. When it matches, store the classification result in the texture type database;

[0083] S22. When the classification result does not match the preset standard, adjust the parameters of the texture classification model, re - classify the texture type, and through the adjusted texture classification model, obtain the classification result of the wood texture type again, and judge whether it matches the preset standard;

[0084] S23. According to the matching result, determine the final classification data of the wood texture type, store it in the texture type database, generate the statistical distribution information of the wood texture type, and determine the characteristic law of the wood texture type according to the statistical distribution information.

[0085] Among them, when inputting the texture feature vector into the preset texture classification model and obtaining the classification result of the wood texture type, the support vector machine algorithm is used for implementation. SVM is a supervised learning algorithm. By finding an optimal hyperplane in the feature space to separate data of different categories, for non - linearly separable data, SVM can map the data to a high - dimensional space through a kernel function to achieve classification.

[0086] Specific process: Take the extracted texture feature vector as the input feature and the wood texture type as the label, and select a suitable kernel function according to the data characteristics, such as linear kernel, polynomial kernel or radial basis function (RBF) kernel. For complex texture classification tasks, select the RBF kernel;

[0087] Input the feature vector and the corresponding texture type label into the SVM model, use the training data to fit the SVM model, find the optimal hyperplane through an optimization algorithm (such as the sequential minimal optimization algorithm SMO), and adjust the hyperparameters (such as the penalty parameter and the kernel function parameter) to optimize the model performance;

[0088] Use cross - validation (such as k - fold cross - validation) to evaluate the classification accuracy, recall rate and F1 score of the model, and then input the new texture feature vector into the trained SVM model to obtain the texture type classification result.

[0089] For linear SVM, the goal is to find a hyperplane that maximizes the margin between different category data; for non - linear SVM, through the kernel function K(x i , x j ) map the data to a high - dimensional space, and the optimization objective is: Among them, w is the normal vector of the hyperplane, representing the classification weight vector, a represents the distance between the hyperplane and the origin, C is the penalty parameter, used to balance the relationship between maximizing the margin and minimizing the classification error, ξ i is the slack variable, allowing some data points to be within the margin or on the wrong side, so as to handle non - linearly separable data.

[0090] The support vector machine (SVM) is used as the main classification algorithm. Since the data after extracting the musical instrument texture features usually has a high dimension, SVM can effectively process such data. SVM can handle complex non-linear classification problems through kernel functions and is suitable for texture classification. SVM has a certain robustness to noise data and can maintain a stable classification effect in the actual production environment.

[0091] Specifically, the support vector machine (SVM) algorithm is used to classify the extracted wood texture feature vectors. This method can accurately distinguish different wood texture types and optimize the classification model to improve the accuracy and robustness of texture matching in musical instrument production.

[0092] Furthermore, a laser scanner is used to obtain the three-dimensional data of the instrument body surface, and the influence of reflectivity and ambient light is removed through a filtering algorithm. Specifically, it includes:

[0093] Use a laser scanner to perform high-precision scanning on the musical instrument body to obtain the three-dimensional point cloud data of the surface. The scanner emits a laser beam and captures the reflected light, and calculates the three-dimensional coordinates of each point by accurately measuring the flight time or phase difference of the light.

[0094] Perform preliminary processing on the collected original point cloud data, including removing obvious noise points and outliers, and performing normalization processing to ensure the consistency and accuracy of the data.

[0095] Apply a filtering algorithm to reduce the influence of ambient light and the reflectivity of the instrument body surface on the data. Eliminate the influence of reflectivity through region segmentation and reflectivity correction, and at the same time use light intensity compensation and time filtering techniques to reduce the interference of ambient light.

[0096] Among them, the surface is divided into multiple regions, the reflectivity of each region is analyzed, and the measured data is corrected according to the known reflectivity characteristics; according to the intensity of the ambient light, the measured data is compensated to eliminate the influence of light intensity changes, and multiple scans are performed at different time points. By comparing the scan results at different time points, the influence of ambient light is removed.

[0097] Use the processed point cloud data for surface curvature analysis. By calculating the curvature of the surface points, analyze the geometric features of the surface, which are used to identify the unevenness and texture features of the surface, so as to extract the key features of the instrument body surface.

[0098] Based on the features extracted from different perspectives, through data registration technology, align and fuse the data from different perspectives in a unified coordinate system to eliminate the influence brought by the perspective difference; generate a triangular mesh model of the instrument body surface according to the fused point cloud data.

[0099] Perform surface smoothing on the triangular mesh model of the instrument body surface. By analyzing the reconstructed triangular mesh model, detect defects on the instrument body surface, such as scratches, pits, etc., and evaluate the surface consistency to check for uneven or abnormal areas.

[0100] Specifically, through the use of a laser scanner and filtering algorithms, this method achieves high-precision acquisition and analysis of the three-dimensional data of the instrument body surface, effectively removing the influence of reflectivity and ambient light, thereby accurately identifying and quantifying the geometric features and defects of the instrument body surface, and improving the quality control level in the instrument manufacturing process.

[0101] S3. Analyze the filtered three-dimensional data, extract the depth and distribution information of the grinding marks, and calculate the adjustment value of the grinding parameters;

[0102] Furthermore, analyze the filtered three-dimensional data, extract the depth and distribution information of the grinding marks, and calculate the adjustment value of the grinding parameters, specifically including:

[0103] According to the filtered three-dimensional data, use a depth analysis method to extract the depth information of the grinding marks. For the extracted depth information, analyze the distribution characteristics of the grinding marks, including the density, direction, and area of the marks, etc., to obtain the distribution information;

[0104] Conduct statistical analysis on the extracted distribution characteristics, such as calculating the average density of the marks, the skewness and kurtosis of the direction distribution, etc., to quantify the distribution characteristics of the marks;

[0105] Based on the depth information and distribution information, calculate the adjustment value of the grinding parameters. According to the adjustment value, determine the specific values of the grinding parameters. When the adjustment value exceeds the preset threshold, recalculate the parameter adjustment value, and finally generate an adjustment plan for the grinding parameters.

[0106] Among them, extract the depth and distribution information of the grinding marks from the filtered three-dimensional data, and calculate the adjustment value of the grinding parameters. First, use 3D modeling software or professional algorithms to compare the difference between the actual surface and the ideal model to accurately extract the depth information, and then by identifying and analyzing the distribution characteristics such as the density, direction, and area of the grinding marks, conduct statistical analysis to quantify these characteristics. Then, based on these depth and distribution information, construct a prediction model, such as a regression analysis or machine learning model, to estimate the adjustment value of the grinding parameters (including grinding pressure, speed, and tool selection). This model will consider the statistical characteristics of the marks and predict the effect of parameter adjustment on mark improvement.

[0107] Specifically, by analyzing the filtered three-dimensional data, accurately extract and quantify the depth and distribution information of the grinding marks, and then calculate the adjustment value of the grinding parameters to optimize the grinding process and ensure that the instrument surface reaches the ideal smoothness and uniformity.

[0108] S4. An optical interferometer is used to obtain the interference image of the coating surface. The interference of surface curvature and light scattering is reduced through an image enhancement algorithm. Based on the enhanced interference image, an edge detection algorithm is used to identify the coating defect area and calculate the defect distribution;

[0109] Further, an optical interferometer is used to obtain the interference image of the coating surface. The interference of surface curvature and light scattering is reduced through an image enhancement algorithm, specifically including:

[0110] Obtain the optical interference image of the coating surface, extract the optical information data in the image, and use an image enhancement algorithm to process the optical information data to reduce the influence of the curvature value and scattering degree on the image value;

[0111] According to the processed image value, analyze the interference amount distribution characteristics of the coating surface. When the interference amount exceeds the preset threshold, the image value is reprocessed by the enhancement method;

[0112] Based on the processing result, determine a suitable algorithm type and data source matching scheme. Through the matching of the algorithm type and data source, generate the final image data of the coating surface.

[0113] Among them, extracting the optical information data in the image and using an image enhancement algorithm to reduce the influence of the curvature value and scattering degree on the image value includes using the image processing toolbox in MATLAB, such as using the imadjust function to adjust the contrast, or using fspecial to create a Gaussian filter and then using the imfilter function to filter the image; it also includes implementing denoising processing of the image through an adaptive median filtering algorithm, and implementing skew correction processing on the image to obtain the filtering result, which is used to improve the image quality, so as to more accurately analyze the interference amount distribution characteristics of the coating surface.

[0114] Specifically, by using the interference image of the coating surface obtained by an optical interferometer and applying an image enhancement algorithm, significant improvement effects have been achieved: the optical information data in the image has been effectively extracted and processed, greatly reducing the interference caused by surface curvature and light scattering; by using the image processing toolbox of MATLAB, through technical means such as adjusting the contrast, applying a Gaussian filter, and adaptive median filtering, not only the clarity and contrast of the image have been improved, but also the noise has been successfully removed and the skew of the image has been corrected, so that the interference amount distribution characteristics of the coating surface can be more accurately analyzed. These improved image data provide a solid foundation for the accurate evaluation of the coating quality, making the detection and analysis of coating defects more reliable, and ultimately improving the accuracy and efficiency of musical instrument coating quality control.

[0115] Further, based on the enhanced interference image, an edge detection algorithm is used to identify the coating defect area and calculate the defect distribution, specifically including:

[0116] An edge detection algorithm is used to extract edge values from the enhanced interference image, determine the boundary range of the defect area, segment the defect area according to the edge values, obtain the image values of the defect area, and calculate the area and position of each defect area;

[0117] A spatial distribution analysis algorithm is used to process the image values of the defect area, generate the distribution values of the defects, judge the aggregation characteristics of the defects. When the distribution value exceeds the preset threshold, a clustering algorithm is used to reclassify the defect area to determine the type and quantity of the defects;

[0118] According to the type and quantity of the defects, a feature extraction algorithm is used to extract the feature values of the defects from the interference image, generate a detailed description of the defects. According to the detailed description of the defects, a classification algorithm is used to classify the defects into grades to determine the severity of the defects.

[0119] Based on the severity of the defects, a spatial distribution map of the coating defects is generated to judge the overall quality status of the coating.

[0120] Among them, in the process of using the edge detection algorithm to identify the coating defect area and calculate the defect distribution based on the enhanced interference image, the edge values in the image are extracted by the edge detection algorithm, the boundary range of the defect area is determined, and then the defect area is accurately segmented according to these edge values, so as to obtain the image values of each defect area and calculate the area and position of each defect area;

[0121] Subsequently, a spatial distribution analysis algorithm is used to process the image values of these defect areas, generate the distribution values of the defects, judge the aggregation characteristics of the defects on the coating. If the distribution value of the defects exceeds the preset threshold, indicating that the defects have a specific aggregation pattern, a clustering algorithm is used to reclassify the defect area to determine the type and quantity of the defects. According to the identified type and quantity of the defects, a feature extraction algorithm is further used to extract the feature values of the defects from the interference image, generating a detailed description of the defects. These feature values may include the shape, size, edge clarity, etc. of the defects. Then, according to the detailed description of the defects, a classification algorithm is used to classify the defects into grades to determine the severity of the defects. Finally, based on the severity of the defects, a spatial distribution map of the coating defects is generated.

[0122] Specifically, the edge detection and clustering algorithms are used to accurately identify and classify the coating defects from the enhanced interference image, quantify the defect distribution, so as to provide detailed defect analysis and overall quality assessment for coating quality control.

[0123] S5. Combine the defect distribution with the preset coating thickness threshold to judge the qualification of the coating quality. According to the adjustment value of the grinding parameters and the judgment result of the coating quality, generate the optimized parameters for the body processing.

[0124] Further, combining the defect distribution with a preset coating thickness threshold, the qualification of the coating quality is judged, and based on the adjustment value of the polishing parameters and the judgment result of the coating quality, the optimized parameters for the body processing are generated, specifically including:

[0125] Obtain the defect distribution data, combine it with the preset coating thickness threshold, and judge the qualification of the coating quality. When the coating quality is unqualified, determine the adjustment value of the polishing parameters according to the defect distribution data and the thickness distribution data;

[0126] Based on the adjustment value and the judgment result of the coating quality, generate the optimized parameters for the body processing, and then use the clustering algorithm to classify the defect areas to determine the aggregation characteristics and distribution rules of the defects;

[0127] According to the aggregation characteristics and distribution rules of the defects, adjust the adjustment value of the polishing parameters, optimize the body processing parameters, generate a coating quality improvement plan based on the optimized parameters and the judgment result of the coating quality, and generate the final optimized parameters for the body processing according to the improvement plan and the processing parameters.

[0128] Specifically, through comprehensive analysis of the defect distribution, the coating thickness threshold, and the polishing parameters, the optimization of the body processing process of the musical instrument is realized, ensuring that the coating quality meets the standards and improving the overall quality of the final product.

[0129] Example 2

[0130] Please refer to Figure 4 , this embodiment provides a production quality control system for musical instruments to implement a production quality control method for musical instruments, including:

[0131] An image acquisition and preprocessing module uses a high-resolution camera to obtain the original image of the wood surface and applies an image denoising algorithm to eliminate the interference of surface defects and light conditions. Through multi-scale analysis and adaptive filtering techniques, such as wavelet transform and NL-Means algorithm, the noise components in the image are accurately located and removed, while the details of the wood texture are retained, thereby improving the image quality and preparing for subsequent feature extraction;

[0132] A texture feature extraction and classification module uses a convolutional neural network (CNN) model to extract features from the denoised wood surface image, generating high-precision texture feature vectors. By constructing a CNN model containing multiple convolutional layers, pooling layers, and fully connected layers, the local and high-level features of the image are extracted, and then the generated texture feature vectors are matched with a preset texture feature standard library to determine whether the feature vectors meet the standards, thereby classifying the wood texture types;

[0133] 3D scanning and surface analysis module, which uses a laser scanner to obtain 3D data of the piano body surface, removes the influence of reflectivity and ambient light through a filtering algorithm, reduces interference through region segmentation, reflectivity correction, light intensity compensation and time filtering techniques, performs surface curvature analysis using the processed point cloud data, extracts key features of the piano body surface, and generates a triangular mesh model for surface smoothing and defect detection;

[0134] Coating analysis and defect identification module, which uses an optical interferometer to obtain the interference image of the coating surface, reduces the interference of surface curvature and light scattering through an image enhancement algorithm, uses an edge detection algorithm to extract edge values from the enhanced interference image, determines the boundary range of the defect area, calculates the area and position of each defect area, then uses a spatial distribution analysis algorithm to process the image values of the defect area, generates the distribution value of the defect, judges the aggregation characteristics of the defect, and uses a clustering algorithm to reclassify the defect area;

[0135] Quality assessment and parameter optimization module, which combines the defect distribution with the preset coating thickness threshold to judge the qualification of the coating quality, generates the optimized parameters for piano body processing according to the adjustment value of the grinding parameters and the coating quality judgment result, classifies the defect areas through a clustering algorithm, determines the aggregation characteristics and distribution rules of the defects, adjusts the grinding parameters according to this information, optimizes the piano body processing parameters, generates a coating quality improvement plan, and generates the final optimized parameters for piano body processing according to the improvement plan and the processing parameters.

[0136] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the above-disclosed technical content without departing from the technical solution scope of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for controlling the production quality of a musical instrument, characterized in that: The steps include: A high-resolution camera is used to obtain wood surface images, and the image denoising algorithm is used to remove surface defects and interference from light conditions. The texture features of the denoised image are extracted based on the convolutional neural network model to generate a high-precision texture feature vector. Input the texture feature vector into the preset texture classification model to obtain the classification result of the wood texture type; use a laser scanner to obtain the three-dimensional data of the surface of the guitar body, and use a filtering algorithm to remove the influence of reflectivity and ambient light; Analyze the filtered 3D data, extract the depth and distribution information of the grinding marks, and calculate the adjustment value of the grinding parameters; An optical interferometer is used to obtain the interference image of the coating surface. The image enhancement algorithm is used to reduce the interference between the surface curvature and light scattering. Based on the enhanced interference image, an edge detection algorithm is used to identify the coating defect area and calculate the defect distribution. The defect distribution and the preset coating thickness threshold are combined to judge the quality of the coating. The optimized parameters for guitar body processing are generated based on the polishing parameter adjustment value and the coating quality judgment result.

2. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: A high-resolution camera is used to obtain images of the wood surface, and an image denoising algorithm is used to remove surface defects and interference from light conditions, including: The original image of the wood surface is obtained by a high-resolution camera, and the original image is input into a preset image denoising algorithm. When light interference exists in the detected image, a filter is used to eliminate the light noise. When surface defects are detected, the defect area is marked by a defect recognition module, and the marked area is optimized in combination with a denoising algorithm to obtain a clear wood surface image. The optimized image is compared with the preset quality standard to determine whether the image meets the requirements, and the image that meets the standard is stored in the database; Among them, the image denoising algorithm is used to remove surface defects and light conditions interference based on multi-scale analysis and adaptive filtering technology. When light interference is detected in the image, the image is decomposed into sub-bands of different scales and directions through multi-scale analysis to separate the noise components and detail features in the image; when surface defects are detected in the image, the adaptive filtering technology is used to dynamically adjust the filtering parameters according to the local statistical information of the image.

3. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: The texture features of the denoised image are extracted based on the convolutional neural network model to generate high-precision texture feature vectors, including: A convolutional neural network model is used to extract features from the denoised wood surface image to obtain texture feature data. A texture feature vector is generated based on the extracted texture feature data, and the dimension and value distribution of the feature vector are determined. Matching the generated texture feature vector with a preset texture feature standard library to determine whether the feature vector meets the standard; When the feature vector meets the standard, the feature vector data is stored in the texture feature database, and when the feature vector does not meet the standard, the feature extraction process is optimized and the texture feature data is re-extracted; Through the optimized feature extraction process, a new texture feature vector is generated, and standard matching is performed again. According to the matching results, a high-precision texture feature vector is determined and stored in the texture feature database.

4. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: Input the texture feature vector into the preset texture classification model to obtain the classification results of wood texture types, including: Using texture feature vectors as input data, processing through a preset texture classification model, obtaining a classification result of wood texture types, judging whether the wood texture type matches a preset standard based on the classification result, and storing the classification result in a texture type database if it matches; When the classification result does not match the preset standard, the parameters of the texture classification model are adjusted, and the texture type classification is re-performed. The classification result of the wood texture type is obtained again through the adjusted texture classification model to determine whether it matches the preset standard; According to the matching results, the final wood texture type classification data is determined and stored in the texture type database, and the statistical distribution information of the wood texture type is generated. According to the statistical distribution information, the characteristic law of the wood texture type is determined.

5. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: A laser scanner is used to obtain the three-dimensional data of the surface of the guitar body, and a filtering algorithm is used to remove the influence of reflectivity and ambient light, including: Use a laser scanner to scan the instrument body with high precision to obtain 3D point cloud data of the surface. The scanner emits a laser beam and captures the reflected light, and calculates the 3D coordinates of each point by accurately measuring the flight time or phase difference of the light. Then, we apply filtering algorithms to reduce the impact of ambient light and the reflectivity of the guitar surface on the data. We eliminate the impact of reflectivity through area segmentation and reflectivity correction. We also use light intensity compensation and time filtering techniques to reduce the interference of ambient light. The processed point cloud data is used to perform surface curvature analysis. By calculating the curvature of the surface points, the geometric features of the surface are analyzed to identify the surface unevenness and texture features, and to extract the key features of the guitar body surface. Based on the features extracted from different perspectives, the data from different perspectives are aligned and fused in a unified coordinate system through data registration technology to eliminate the impact of perspective differences; based on the fused point cloud data, a triangular mesh model of the guitar body surface is generated; The triangular mesh model of the piano body surface is smoothed, and the defects on the piano body surface are detected by analyzing the reconstructed triangular mesh model, and the surface consistency is evaluated to check whether there are uneven or abnormal areas.

6. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: Analyze the filtered 3D data, extract the depth and distribution information of the grinding marks, and calculate the adjustment value of the grinding parameters, including: According to the filtered three-dimensional data, the depth information of the grinding marks is extracted using a depth analysis method, the distribution characteristics of the grinding marks are analyzed for the extracted depth information to obtain distribution information, and the extracted distribution characteristics are statistically analyzed to quantify the distribution characteristics of the marks; Based on the depth information and distribution information, the adjustment value of the polishing parameter is calculated, and the specific value of the polishing parameter is determined according to the adjustment value. When the adjustment value exceeds the preset threshold, the parameter adjustment value is recalculated, and finally an adjustment plan for the polishing parameter is generated.

7. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: An optical interferometer is used to obtain the interference image of the coating surface, and the interference of surface curvature and light scattering is reduced through image enhancement algorithms, including: Obtain the optical interference image of the coating surface, extract the light information data in the image, and use the image enhancement algorithm to process the light information data to reduce the influence of curvature value and scattering degree on the image value; According to the processed image value, the distribution characteristics of the interference amount on the coating surface are analyzed. When the interference amount exceeds the preset threshold, the image value is reprocessed by the enhancement method; Based on the processing volume results, the appropriate algorithm and data source matching scheme is determined, and the final image data of the coating surface is generated by matching the algorithm and the data source.

8. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: Based on the enhanced interference image, the edge detection algorithm is used to identify the coating defect area and calculate the defect distribution, including: The edge detection algorithm is used to extract edge values ​​from the enhanced interference image, determine the boundary range of the defect area, segment the defect area according to the edge value, obtain the image value of the defect area, and calculate the area and position of each defect area; The spatial distribution analysis algorithm is used to process the image value of the defect area, generate the distribution value of the defect, and judge the clustering characteristics of the defect. When the distribution value exceeds the preset threshold, the clustering algorithm is used to reclassify the defect area to determine the type and number of defects; According to the type and quantity of defects, a feature extraction algorithm is used to extract the characteristic values ​​of the defects from the interference image to generate a detailed description of the defects. Based on the detailed description of the defects, a classification algorithm is used to grade the defects and determine the severity of the defects. According to the severity of the defects, a spatial distribution map of coating defects is generated to judge the overall quality status of the coating.

9. The method for controlling the production quality of a musical instrument according to claim 1, characterized in that: Combine the defect distribution with the preset coating thickness threshold to judge the quality of the coating. According to the polishing parameter adjustment value and the coating quality judgment result, generate the optimized parameters for the guitar body processing, including: Obtain defect distribution data, and judge the quality of coating in combination with the preset coating thickness threshold. If the coating quality is unqualified, determine the adjustment value of grinding parameters based on the defect distribution data and thickness distribution data; The optimized parameters for guitar body processing are generated by adjusting the values ​​and the coating quality judgment results. Then, the clustering algorithm is used to classify the defect areas and determine the clustering characteristics and distribution patterns of the defects. According to the defect aggregation characteristics and distribution patterns, the polishing parameter adjustment values ​​are adjusted to optimize the guitar body processing parameters. Through the optimization parameters and coating quality judgment results, a coating quality improvement plan is generated. Based on the improvement plan and processing parameters, the final guitar body processing optimization parameters are generated.

10. A production quality control system for a musical instrument, applied to a production quality control method for a musical instrument as claimed in any one of claims 1 to 9, characterized in that: include: Image acquisition and preprocessing module, which uses a high-resolution camera to obtain the original image of the wood surface, and applies an image denoising algorithm to eliminate the interference of surface defects and light conditions. It locates and removes noise components in the image through multi-scale analysis and adaptive filtering technology; The texture feature extraction and classification module uses a convolutional neural network model to extract features from the denoised wood surface image and generate a high-precision texture feature vector. By building a CNN model, it extracts local and high-level features of the image, and then matches the generated texture feature vector with a preset texture feature standard library to determine whether the feature vector meets the standard and classify the wood texture type. The 3D scanning and surface analysis module uses a laser scanner to obtain 3D data of the piano body surface, and uses a filtering algorithm to remove the influence of reflectivity and ambient light. It uses region segmentation, reflectivity correction, light intensity compensation and time filtering technology to reduce interference. It uses the processed point cloud data to perform surface curvature analysis, extract key features of the piano body surface, and generate a triangular mesh model for surface smoothing and defect detection. The coating analysis and defect recognition module uses an optical interferometer to obtain the interference image of the coating surface, and uses an image enhancement algorithm to reduce the interference of surface curvature and light scattering. It uses an edge detection algorithm to extract edge values ​​from the enhanced interference image, determine the boundary range of the defect area, and calculate the area and position of each defect area. It then uses a spatial distribution analysis algorithm to process the image value of the defect area, generate the distribution value of the defect, determine the clustering characteristics of the defect, and use a clustering algorithm to reclassify the defect area; The quality assessment and parameter optimization module combines the defect distribution with the preset coating thickness threshold to judge the quality of the coating. Based on the polishing parameter adjustment value and the coating quality judgment result, it generates the optimization parameters for the guitar body processing. It classifies the defect areas through the clustering algorithm, determines the clustering characteristics and distribution patterns of the defects, adjusts the polishing parameters based on the information, optimizes the guitar body processing parameters, generates a coating quality improvement plan, and generates the final guitar body processing optimization parameters based on the improvement plan and processing parameters.

Images