A method for detecting musical instrument production defects based on image recognition

By preprocessing the surface image of the instrument case and decomposing it using the EMD algorithm, the problem of low recognition accuracy caused by wood grain interference in traditional methods is solved, and efficient crack detection is achieved.

CN118521583BActive Publication Date: 2024-11-26JIANGYIN GOLDENCUP ANGELS MUSICAL INSTR CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410986004.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-26
Estimated Expiration
2044-07-23

AI Technical Summary

Technical Problem

Traditional image detection and recognition methods in musical instrument production are limited by the interference of wood surface grain, resulting in low accuracy in judging cracks and low recognition precision, especially in the inability to accurately identify tiny cracks.

Method used

By preprocessing the surface image of the instrument case, the direction of the wood grain is determined, a mapping function between pixel coordinates and gray values ​​is constructed, clustering decomposition is performed, intrinsic mode functions (IMFs) are extracted using the EMD algorithm, and significant information entropy is calculated to detect crack defects.

Benefits of technology

It improves the accuracy and reliability of identifying crack defects in musical instrument production, reduces computational load, improves image alignment, reduces noise interference, and enhances the efficiency and accuracy of the inspection process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118521583B_ABST
    Figure CN118521583B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of image processing, and more specifically, to a method for detecting musical instrument production defects based on image recognition, the method comprising: obtaining an accordion case surface image for preprocessing, determining the direction of wood grain, and constructing a mapping function between each column of pixel coordinates and grayscale values; clustering each column of functions, decomposing each cluster cluster, obtaining an intrinsic modal function, calculating the importance of each intrinsic modal function, and in response to the importance being greater than or equal to a preset degree, the importance remains unchanged, otherwise the importance is set to 0; using the importance as an adjustment parameter to regenerate an optimized modal function, and calculating significant information entropy, in response to the significant information entropy being greater than the preset entropy, there is a crack defect at the position of the case surface image corresponding to the cluster cluster. The present invention improves the distinction between cracks and wood grain on the wood surface and improves detection accuracy by decomposing the cluster clusters.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing, and more specifically, to a method for detecting musical instrument production defects based on image recognition. Background Art

[0002] Musical instrument production refers to the process of manufacturing and producing various musical instruments, including various string instruments, keyboard instruments, wind instruments, percussion instruments, and electronic instruments. The musical instrument production process is a comprehensive project that integrates professional knowledge and technology in multiple fields such as design innovation, material science, precision processing, and music art. During the production process of the accordion, the wood used to make the case may crack due to climate change or improper storage; in addition, gaps may be left due to operational errors during processing. The cracks in the case will cause the accordion to leak, thereby affecting the timbre and volume of the accordion and accompanied by noise.

[0003] The existing Chinese patent application document with publication number CN112581434A discloses an image recognition method for product defect detection, which includes collecting images of the product with defects and marking the defects; associating the marked defects with the defect types; preprocessing the defective image; segmenting the defective area using a threshold segmentation method; performing convolution calculation on the segmented image; extracting multi-dimensional regional features of the image and correlation features between the multi-dimensional regional features; collecting images of the product to be detected; finding suspected defective areas; performing image processing on the suspected defective areas, calculating their similarity values, comparing thresholds and similarities, and determining product defects.

[0004] The above application documents solve the limitations of traditional vision and improve the factory's production efficiency. However, the wood grain of the instrument case surface image during the production process is highly irregular and complex. Different wood species, growth conditions and cutting methods will produce different texture patterns, and the textures are highly similar. Traditional image detection and recognition is limited by the interference of the wood surface grain, which may result in the wood grain being identified as noise. The accuracy of crack judgment is not high, the recognition precision is low, and some tiny cracks cannot be accurately identified. Summary of the invention

[0005] In order to solve the problem that traditional image detection and recognition is limited by the interference of wood surface grain in the musical instrument production process, the accuracy of crack judgment is not high, and the recognition precision is low, the present invention provides solutions in the following aspects.

[0006] A method for detecting defects in musical instrument production based on image recognition comprises: obtaining an image of the surface of an accordion case, preprocessing the image of the surface of the accordion case, determining the direction of wood grain, and constructing a mapping function between the coordinates of each column of pixel points and the grayscale value; clustering each column of functions, taking the average value of the function of the first column of pixel points as the initial cluster center, calculating the possibility that each column of functions belongs to the cluster to which the previous column of functions belongs, updating the cluster center according to the possibility, iterating all column functions to obtain a plurality of cluster clusters, decomposing each cluster cluster to obtain an intrinsic mode function, and calculating the importance of each intrinsic mode function. , where Indicates The importance of the intrinsic mode function, Indicates The first The intrinsic mode function is related to the original The similarity of the class center functions of the clusters, Indicates The first The intrinsic mode function and the The first The similarity of the intrinsic mode functions, Indicates The first The intrinsic mode function and the The first The similarity of the inherent modal functions is calculated. In response to the importance being greater than or equal to the preset importance, the importance remains unchanged, otherwise the importance is set to 0; the importance is used as an adjustment parameter to regenerate the optimized modal function, and the difference between the information entropy of the optimized modal function in each cluster and the average information entropy of all clusters is calculated as the significant information entropy of the corresponding cluster. In response to the significant information entropy being greater than the preset entropy, there is a crack defect at the position of the piano case surface image corresponding to the cluster.

[0007] The effect is: by preprocessing the surface image of the accordion case and determining the direction of the wood grain, the wood texture characteristics can be analyzed more accurately, and the mapping function can be clustered to decompose the image into several regions with similar characteristics. Each cluster is decomposed using EMD to obtain IMFs, which represent the signal characteristics at different time scales. The importance of each IMF is then evaluated to determine the significant information entropy, which can effectively detect defects such as cracks from the image of the accordion case surface, thereby improving the accuracy and reliability of crack defect identification.

[0008] In one embodiment, the preprocessing of the piano case surface image to determine the direction of the wood grain includes:

[0009] Gray-scale the piano case surface image, and perform edge detection on the gray-scaled piano case surface image to obtain wood grain;

[0010] In response to the angle between the wood grain and the horizontal direction of the image being greater than a preset angle, the image is rotated to make the wood grain direction parallel to the horizontal direction of the image, thereby determining the wood grain direction.

[0011] The effect is that by rotating the image so that the wood grain is parallel to the horizontal direction, the alignment of the image can be improved, the direction of the wood grain in the image can be standardized, and the wood grain information can be better analyzed.

[0012] In one embodiment, obtaining the plurality of clusters includes:

[0013] The average value of the function of the first column of pixels is used as the initial cluster center, the number of average minimum values ​​within the category is calculated, and the possibility of each column of functions belonging to the cluster to which the previous column of functions belongs is calculated based on the number of average minimum values;

[0014] In response to the possibility that the column function belongs to the cluster to which the previous column function belongs being greater than or equal to a preset threshold, the column function is added to the cluster to which the previous column function belongs, and the number of cluster centers and the average minimum values ​​within the class are updated; otherwise, it is used as the initial function of the next class, and iterated to all column functions to obtain several clusters.

[0015] The effect is: by identifying the anomalies or outliers between each cluster and other clusters, the data dimension is reduced, the amount of calculation is reduced, and the speed of clustering process analysis is improved.

[0016] In one embodiment, the probability satisfies the following relationship:

[0017] ,

[0018] In the formula, Indicates Ledi The coordinate value function of the pixel point belongs to The probability of clusters, Indicates Ledi The coordinate value function of the pixel point is Indicates The cluster center function of the clusters, Indicates Ledi The pixel function is The covariance of the cluster center functions of the clusters, Indicates The number of minimum values ​​in the column pixel function, Indicates The average number of minima in clusters, represents the natural exponential function.

[0019] In one embodiment, the process of decomposing each cluster includes:

[0020] Traverse the class center function of each cluster, determine the maximum extreme point and the minimum extreme point in the cluster as the local extreme value, and then obtain the upper and lower envelopes by interpolation, iterate until the residual in the decomposition process is zero, and stop the decomposition;

[0021] The class center function satisfies the following relationship:

[0022] ,

[0023] In the formula, Indicates The cluster center function of the clusters, is the number of decompositions, Indicates The first The intrinsic mode functions, Indicates The residual term after the decomposition of the class center function of the clusters.

[0024] The effect is that by obtaining the upper and lower envelopes of the maximum and minimum extreme points in the cluster, the upper and lower envelopes are used as the upper and lower limits in the cluster, thereby removing noise to make the signal clearer and extracting the key features of the wood grain.

[0025] In one embodiment, the similarity between the decomposed intrinsic mode function in the cluster and the original function of the cluster satisfies the following relationship:

[0026] ,

[0027] In the formula, Representation function and function The function used to compare similarity refers to the parameters, represents a constant, Representation function and function The covariance of Representation function The phase of Representation function phase.

[0028] In one embodiment, the regenerated intrinsic mode function satisfies the following relationship:

[0029] ,

[0030] In the formula, Indicates The final intrinsic mode function of the clusters is represents the number of intrinsic mode functions, Indicates The importance of the intrinsic mode function, Indicates The first Intrinsic mode functions.

[0031] In one embodiment, the preset entropy includes:

[0032] The preset entropy is set to twice the standard deviation of the information entropy of all cluster functions.

[0033] The effect is that when the significant information entropy exceeds twice the standard deviation of the information entropy of all cluster functions, it is considered that the significant information entropy exceeds the normal fluctuation range, thereby determining that there are cracks in the surface image of the guitar case.

[0034] The present invention has the following effects:

[0035] 1. The present invention constructs a position-pixel value distribution model in an image through the regular distribution of wood grain, and uses the EMD algorithm to decompose clusters, which can effectively distinguish cracks and wood grain on the wood surface, thereby improving the accuracy of detection.

[0036] 2. The present invention takes into account multiple factors such as the mean value, standard deviation and phase of the function value through the calculation of similarity, making the calculation of similarity more comprehensive and accurate. The accurate calculation of function similarity lays the foundation for subsequent EMD decomposition, IMF screening and synthesis steps, and improves the efficiency and accuracy of the entire detection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0038] Figure 1 It is a method flow chart of steps S1 to S3 in a method for detecting musical instrument production defects based on image recognition in an embodiment of the present invention.

[0039] Figure 2 It is a schematic diagram of a mapping function between the coordinates of each column of pixel points and the grayscale value in a method for detecting musical instrument production defects based on image recognition in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0041] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0042] The specific scenario of the present invention is: during the production process of the accordion case, the cracks on the surface of the case are detected. A high-resolution camera is used to photograph the surface of the case. The camera is located directly above the case, and the light is uniform and sufficient to ensure that the camera can capture a clear image of the wood and the shape of the wood grain on the wood.

[0043] Reference Figure 1 A method for detecting musical instrument production defects based on image recognition includes steps S1 to S3, which are as follows:

[0044] S1: Obtain the surface image of the accordion case, pre-process the surface image of the accordion case, determine the direction of the wood grain, and construct a mapping function between the coordinates of each column of pixel points and the grayscale value.

[0045] Grayscale the surface image of the guitar case, and perform edge detection on the grayscale surface image of the guitar case to obtain wood grain;

[0046] In response to the angle between the wood grain and the horizontal direction of the image being greater than a preset angle, the image is rotated to make the wood grain direction parallel to the horizontal direction of the image, thereby determining the wood grain direction.

[0047] Further explanation, in this embodiment, the preset angle is 15°, which can be adjusted by technical personnel in this field according to actual conditions to obtain wood grain parallel to the horizontal direction of the image, so as to facilitate unified analysis.

[0048] Reference Figure 2 , decompose the processed grayscale image by column, and construct the coordinate-grayscale value model of each column of pixels. In the figure, a represents the crack position, and A and B represent two columns of function curves in the grayscale image, such as: ,in, Indicates Column pixel coordinate-gray value mapping function, Indicates that the column The coordinate value of the pixel point, For the Ledi The gray value of a pixel.

[0049] It should be noted that if Figure 2 As shown in the figure, the wood grain of the guitar case surface image spreads vertically, so by analyzing It can roughly reflect the distribution of wood grain. Since the pixel value of grain and cracks is lower than that of the guitar case surface, Minimum values ​​are more likely to occur at lines and cracks.

[0050] S2: Cluster each column of functions, take the average value of the function of the first column of pixels as the initial cluster center, calculate the possibility that each column of functions belongs to the cluster to which the previous column of functions belongs, update the cluster center according to the possibility, iterate all column functions to obtain several clusters, decompose each cluster to obtain the intrinsic mode function, calculate the importance of each intrinsic mode function, and in response to the importance being greater than or equal to the preset importance, the importance remains unchanged, otherwise the importance is set to 0.

[0051] Further explanation, the function of getting each column of pixels Directly decomposing each function will result in too much computation and may cause the problem of generating pseudo vectors due to over-decomposition. Therefore, each column function is clustered first, and adjacent and similar column functions are clustered into the same category. Column functions with high similarity are merged to enhance the features of column functions, reducing the error caused by over-decomposition, which is beneficial to the preliminary extraction of local features of the piano case surface, thereby helping to reduce the amount of calculation.

[0052] The clustering process includes:

[0053] Take the average value of the function of the first column of pixels as the initial cluster center, calculate the number of average minimum values ​​within the category, traverse each column of functions, and calculate the possibility that each column of functions belongs to the cluster to which the previous column of functions belongs;

[0054] Specifically, the probability satisfies the following relationship:

[0055] ,

[0056] In the formula, Indicates Ledi The coordinate value function of the pixel point belongs to The probability of clusters, Indicates Ledi The coordinate value function of the pixel point is Indicates The cluster center function of the clusters, Indicates Ledi The pixel function is The covariance of the cluster center functions of the clusters, Indicates The number of minimum values ​​in the column pixel function, Indicates The average number of minima in clusters, represents the natural exponential function.

[0057] Further explanation: Calculating covariance is used to measure the correlation between two functions. The larger the absolute value of covariance, the greater the correlation between the two functions. If the two variables are not correlated, the covariance is zero, and the covariance of the two functions is positively correlated. The higher the correlation between the two functions, the greater the possibility that the current function belongs to this category; the minimum value can be regarded as a line or crack. Considering that the smaller the difference in the number of minimum values ​​in a column of functions, the greater the possibility that they belong to the same category, the column functions of the same category pass through the same number of lines and cracks.

[0058] In response to the possibility that the column function belongs to the cluster to which the previous column function belongs being greater than or equal to a preset threshold, the column function is added to the cluster to which the previous column function belongs, and the number of cluster centers and the average minimum values ​​within the class are updated; otherwise, it is used as the initial function of the next class, and iterated to all column functions to obtain several clusters.

[0059] To further illustrate, the preset threshold is set to 0.7.

[0060] The process of decomposing each cluster includes:

[0061] Traverse the class center function of each cluster, determine the maximum extreme point and the minimum extreme point in the cluster as the local extreme value, and then obtain the upper and lower envelopes by interpolation, iterate until the residual in the decomposition process is zero, and stop the decomposition;

[0062] Further explanation, in this embodiment, the decomposition adopts the EMD algorithm. Empirical Mode Decomposition (EMD) is an adaptive signal processing method. The EMD algorithm can decompose complex nonlinear and non-stationary signals into a series of intrinsic mode functions (IMFs). Each IMF satisfies specific conditions and can reflect the local characteristics of the signal. The algorithm is well known to those skilled in the art and will not be described in detail. In this embodiment, a series of intrinsic mode functions (IMFs) and residual terms are obtained after decomposition. The mode functions (IMFs) each represent different frequency components in the function, showing texture features of different scales and frequencies in the grayscale image. After decomposition, the minimum value point of the class center function of each cluster is obtained, but the minimum value point corresponds to the wood grain or crack area.

[0063] Specifically, the class center function satisfies the following relationship:

[0064] ,

[0065] In the formula, Indicates The cluster center function of the clusters, is the number of decompositions, Indicates The first The intrinsic mode functions, Indicates The residual term after the decomposition of the class center function of the clusters.

[0066] Further explanation: In this embodiment, the number of splits is set to 5, which can be adjusted by those skilled in the art.

[0067] It should be noted that the traditional EMD algorithm performs noise reduction processing on the signal by removing the first several IMFs and the remainder, and reconstructing the remaining IMFs to obtain the noise-reduced signal. Usually, the high-frequency noise components are concentrated in the first few IMFs, while the main features of the signal may be distributed in the subsequent IMFs. The remaining IMFs after removing the noise IMFs are reconstructed, that is, the signal is reconstructed by superposition of these IMFs. In this embodiment, by selecting the IMF components with texture, crack and piano case surface information, noise interference is eliminated, which is conducive to improving the final defect detection accuracy.

[0068] Specifically, the importance satisfies the following relationship:

[0069] ,

[0070] In the formula, Indicates The importance of the intrinsic mode function, Indicates The first The intrinsic mode function is related to the original The similarity of the class center functions of the clusters, Indicates The first The intrinsic mode function and the The first The similarity of the intrinsic mode functions, Indicates The first The intrinsic mode function and the The first The similarity of the intrinsic mode functions.

[0071] In response to the importance being greater than or equal to the preset level, the importance remains unchanged, otherwise the importance is set to 0.

[0072] To further illustrate, the default level is set to 0.7.

[0073] Specifically, the update formula is as follows:

[0074] ,in, Indicates the preset level.

[0075] Among them, the similarity between the decomposed intrinsic mode function in the cluster and the original function of the cluster satisfies the following relationship:

[0076] ,

[0077] In the formula, Representation function and function The function used to compare similarity refers to the parameters, represents a constant, Representation function and function The covariance of Representation function The phase of Representation function phase.

[0078] Further explanation: Representation function and function The phase distance, the phase reflects to a certain extent the distribution of the minimum value of the function. In this embodiment, the minimum value in the function is generally at the texture or crack. If the phases of the two functions are close, the more likely it is. Representation function and function The covariance is used to measure the correlation between two variables. The larger the absolute value of the covariance, the greater the correlation between the two functions. If the two variables have no correlation, the covariance is zero. Set it to 1.5 for normalization and can be adjusted according to actual conditions.

[0079] It should be noted that the similarity between the IMF of the current column function and the IMF of the clusters before and after it, as well as the similarity between the IMF of the current column function and the original cluster, is analyzed separately, which further shows that if an IMF is very similar to the original cluster, then it can be considered that this IMF contains important features of the original signal, so its importance is relatively high. At the same time, the IMF that is not much different from the same level of intrinsic mode function (IMF) decomposed from the surrounding functions is regarded as noise. This is because noise usually manifests as random fluctuations, which may appear in similar forms in different IMFs, rather than as a meaningful component of the signal. Therefore, if an IMF is not much different from the adjacent IMFs, then its importance is relatively low.

[0080] It should be noted that cracks usually cause discontinuity in the wood grain on the surface of the guitar case, which will cause changes in information entropy. Therefore, it is possible to determine whether there are cracks on the surface of the guitar case by analyzing the information entropy.

[0081] S3: Regenerate the optimized modal function using the importance as an adjustment parameter, and take the difference between the information entropy of the optimized modal function in each cluster and the average information entropy of all clusters as the significant information entropy of the corresponding cluster. In response to the significant information entropy being greater than the preset entropy, there is a crack defect at the position of the piano case surface image corresponding to the cluster.

[0082] Specifically, the regenerated intrinsic mode function satisfies the following relationship:

[0083] ,

[0084] In the formula, Indicates The final intrinsic mode function of the clusters is represents the number of intrinsic mode functions, Indicates The importance of the intrinsic mode function, Indicates The first Intrinsic mode functions.

[0085] To further illustrate, in the above steps, the importance of each IMF is calculated, and the IMFs that contribute more to the final result are screened out, so that higher weights are given when synthesizing the final IMF, which helps to highlight the meaningful features in the signal while suppressing noise or unimportant components.

[0086] In this embodiment, the information entropy of each type of function is calculated, which is a well-known technology in the art and will not be described in detail. The information entropy is a measure of the uncertainty of a random variable. The smaller the value, the smaller the uncertainty of the variable, that is, the more certain the information is; conversely, the larger the information entropy, the more uncertain the information is.

[0087] Further explanation: if the information entropy of a class function increases significantly relative to other class functions, this may indicate the presence of cracks in the region, because cracks introduce additional texture variations, thereby increasing the uncertainty of the information.

[0088] Specifically, the information entropy of each cluster satisfies the following formula:

[0089] ,

[0090] In the formula, Indicates The significant information entropy of clusters, Indicates The information entropy of clusters, Represents the average value of the information entropy of all clustering functions.

[0091] The preset entropy is set to twice the standard deviation of the information entropy of all clustering functions.

[0092] This is the end of the present invention.

[0093] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0094] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for detecting musical instrument production defects based on image recognition, characterized in that: include: Obtain the surface image of the accordion case, pre-process the surface image of the accordion case, determine the direction of the wood grain, and construct a mapping function between the coordinates of each column of pixel points and the grayscale value; Cluster each column of functions, take the average value of the function of the first column of pixels as the initial cluster center, calculate the possibility that each column of functions belongs to the cluster to which the previous column of functions belongs, update the cluster center according to the possibility, iterate all column functions to obtain several clusters, decompose each cluster to obtain the intrinsic mode function, and calculate the importance of each intrinsic mode function. , where Indicates The importance of the intrinsic mode function, Indicates The first The intrinsic mode function is related to the original The similarity of the class center functions of the clusters, Indicates The first The intrinsic mode function and the The first The similarity of the intrinsic mode functions, Indicates The first The intrinsic mode function and the The first The similarity of the intrinsic mode functions is calculated. If the importance is greater than or equal to the preset importance, the importance remains unchanged. Otherwise, the importance is set to 0. The importance is used as a regulating parameter to regenerate the optimized modal function to satisfy the following relationship: , where Indicates The final intrinsic mode function of the clusters is represents the number of intrinsic mode functions, Indicates The importance of the intrinsic mode function, Indicates The first Intrinsic modal functions; the difference between the information entropy of the optimized modal function in each cluster and the average information entropy of all clusters is taken as the significant information entropy of the corresponding cluster, and in response to the significant information entropy being greater than the preset entropy, there is a crack defect at the position of the piano case surface image corresponding to the cluster.

2. The method for detecting musical instrument production defects based on image recognition according to claim 1, characterized in that: The preprocessing of the surface image of the guitar case to determine the direction of the wood grain includes: Gray-scale the piano case surface image, and perform edge detection on the gray-scaled piano case surface image to obtain wood grain; In response to the angle between the wood grain and the horizontal direction of the image being greater than a preset angle, the image is rotated to make the wood grain direction parallel to the horizontal direction of the image, thereby determining the wood grain direction.

3. The method for detecting musical instrument production defects based on image recognition according to claim 1, characterized in that: The plurality of clusters are obtained, including: The average value of the function of the first column of pixels is used as the initial cluster center, the number of average minimum values ​​within the category is calculated, and the possibility of each column of functions belonging to the cluster to which the previous column of functions belongs is calculated based on the number of average minimum values, satisfying the following relationship: , where Indicates Ledi The coordinate value function of the pixel point belongs to The probability of clusters, Indicates Ledi The coordinate value function of the pixel point is Indicates The cluster center function of the clusters, Indicates Ledi The pixel function is The covariance of the cluster center functions of the clusters, Indicates The number of minimum values ​​in the column pixel function, Indicates The average number of minima in clusters, represents the natural exponential function; In response to the possibility that the column function belongs to the cluster to which the previous column function belongs being greater than or equal to a preset threshold, the column function is added to the cluster to which the previous column function belongs, and the number of cluster centers and the average minimum values ​​within the class are updated; otherwise, it is used as the initial function of the next class, and iterated to all column functions to obtain several clusters.

4. The method for detecting musical instrument production defects based on image recognition according to claim 1, characterized in that: The process of decomposing each cluster includes: Traverse the class center function of each cluster, determine the maximum extreme point and the minimum extreme point in the cluster as the local extreme value, and then obtain the upper and lower envelopes by interpolation, iterate until the residual in the decomposition process is zero, and stop the decomposition; The class center function satisfies the following relationship: , In the formula, Indicates The cluster center function of the clusters, is the number of decompositions, Indicates The first The intrinsic mode functions, Indicates The residual term after the decomposition of the class center function of the clusters.

5. The method for detecting musical instrument production defects based on image recognition according to claim 1, characterized in that: in, The similarity between the decomposed intrinsic mode function in the cluster and the original function of the cluster satisfies the following relationship: , In the formula, Representation function and function The function used to compare similarity refers to the parameters, represents a constant, Representation function and function The covariance of Representation function The phase of Representation function phase.

6. The method for detecting musical instrument production defects based on image recognition according to claim 1, characterized in that: in, The preset entropy includes: The preset entropy is set to twice the standard deviation of the information entropy of all cluster functions.

Citation Information

Patent Citations

  • Image recognition method for product defect detection

    CN112581434A

  • Building wood defect detection method based on gray level transformation

    CN114897896A

  • Titanium ingot production quality detection method and system based on image processing

    CN118334029A

  • Surface defect detection method, apparatus, system, storage medium, and program product

    WO2022127919A1