Insole manufacturing method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of insole manufacturing, and provides an insole manufacturing method and system based on artificial intelligence. The method comprises the following steps: defining a defect type label set and a corresponding root cause set of a finished insole, collecting corresponding defect photos, and training a classification model based on a neural network; carrying out continuous shooting, image acquisition and motion blur removal recovery on the production process on the production line; and the restored image is preprocessed. Extracting geometric edge features in the image to form an image feature set; dividing the image feature set into qualified products or defective products based on a Hu moment template matching mechanism; based on a classification model, defect classification is carried out on the image feature set divided into faults, and production adjustment is carried out after reasons causing defects are eliminated in combination with a root cause set; the problems that in the prior art, when the quality of the mold or the injection machine is analyzed through personal experience and visual detection, efficiency is low, and analysis is not comprehensive and thorough can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of insole manufacturing, and particularly to an insole manufacturing method and system based on artificial intelligence. Background Art

[0002] In the production and manufacturing of insoles, after the raw materials of the insole are heated and melted, they are injected into a mold by an injection machine and continue to be heated and foamed to form a blank, and then the finished insole is obtained through stamping, cutting, and trimming. With the improvement of the automation production and quality control standards in the manufacturing industry, higher requirements are put forward for the quality and shape of insoles. The quality of the insole depends on the quality of the mold itself and the quality of the raw materials. If improvement is needed, only the mold or raw materials can be changed according to the quality problems presented by the product.

[0003] In the prior art, engineers generally conduct visual inspection on the cut insoles based on experience, with low efficiency and incomplete and insufficient analysis. Summary of the Invention

[0004] In view of the above technical problems, the present invention provides an insole manufacturing method and system based on artificial intelligence to solve the problems of low efficiency, incomplete and insufficient analysis when relying on personal experience and visual inspection to analyze the quality of insoles in the prior art.

[0005] Other features and advantages of the present disclosure will become apparent through the following detailed description, or be learned in part through the practice of the present disclosure.

[0006] According to one aspect of the present invention, an insole manufacturing method based on artificial intelligence is disclosed, and the method includes:

[0007] Define a set of defect type labels for the finished insole and a corresponding set of root causes for the set of defect type labels, and collect defect photos corresponding to the set of defect type labels to train a classification model based on a neural network;

[0008] Continuously photograph and collect images of the production process on the production line, establish an image degradation model, combine the image degradation model with a blind deconvolution filtering algorithm, and restore the motion-blurred image by minimizing an objective function in the form of a Lagrangian containing a penalty function, and the penalty function uses the H 1 norm;

[0009] Preprocess the restored image to enhance the contrast of the highlight area and reduce the gray value of the low gray value area of the preprocessed image, and reduce the noise and enhance the texture in the image. Based on the edge detection algorithm, extract the geometric edge features in the preprocessed image to form an image feature set;

[0010] Based on the Hu - moment template matching mechanism, calculate the similarity between the image feature set and the standard template. According to the comparison between the matching result and the threshold, divide the image feature set into qualified products or defective products;

[0011] Based on the classification model, classify the defects in the image feature set classified as faults, and combine with the root - cause set to eliminate the causes leading to defects.

[0012] Further, the pre - processing includes:

[0013] Perform gray - scale processing based on gamma transformation on the restored image;

[0014] Perform median filtering on the gray - scale processed image.

[0015] Further, when using the edge - detection algorithm to extract geometric edge features, it includes:

[0016] Calculate the signal - to - noise ratio standard to evaluate the quality of the extracted edge;

[0017] Calculate the edge - location accuracy to make the extracted edge points close to the center of the actual edge;

[0018] Calculate the average distance of the zero - crossing points of the impulse response function.

[0019] Further, when dividing the image feature set into qualified products or defective products, it includes:

[0020] Based on the Hu - moment theory, calculate the first Hu - moment data of the template image, and calculate the second Hu - moment data of the to - be - detected image represented by the image feature set;

[0021] Based on the contour matching of Hu - moments, match the first Hu - moment data with multiple second Hu - moment data to obtain multiple matching results;

[0022] Based on the arbitration function, when the matching result is higher than the threshold but less than its scaled value, perform secondary pre - processing and geometric edge feature extraction on the to - be - detected image, and then perform secondary matching. Determine whether the to - be - detected image is qualified or unqualified according to the result of the secondary matching.

[0023] Further, the defect - type label set includes one or more of the following:

[0024] Dimensions do not conform, deformation, flash, cracking, surface crack, black - dot coking.

[0025] According to a second aspect of the present disclosure, there is provided an insole manufacturing system based on artificial intelligence, including: a model training module for defining a set of defect type labels for finished insoles and a corresponding set of root causes for the set of defect type labels, collecting defect photos corresponding to the set of defect type labels, and training a classification model based on a neural network; an image restoration module for continuously photographing and image collecting the production process on the production line, establishing an image degradation model, combining the image degradation model with a blind deconvolution filtering algorithm, and restoring a motion-blurred image by minimizing an objective function in a Lagrangian form including a penalty function, where the penalty function uses the H 1 -norm; a preprocessing module for preprocessing the restored image to enhance the contrast of the highlight area and reduce the gray value of the low gray value area of the preprocessed image, and to reduce noise and enhance texture in the image, and based on an edge detection algorithm, extracting geometric edge features in the preprocessed image to form an image feature set; a matching module for calculating the similarity between the image feature set and a standard template based on the Hu moment template matching mechanism, and dividing the image feature set into qualified products or defective products according to the comparison between the matching result and a threshold; an analysis module for defect classification of the image feature set classified as faulty based on the classification model, and combining with the set of root causes to eliminate the causes leading to defects.

[0026] The technical solution of the present disclosure has the following beneficial effects:

[0027] By using a matching mechanism to quickly divide defective products and qualified products, it has higher efficiency and accuracy than traditional visual inspection. Then, based on the defined set of defect type labels for finished insoles and the set of root causes, after quickly locating the defect type through the classification model, it is possible to directly locate the corresponding mold defect or injection machine defect, and then quickly solve the root cause of the defect, which in turn feeds back to insole manufacturing and improves the quality of insole manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flowchart of an insole manufacturing method based on artificial intelligence in an embodiment of this specification;

[0029] Figure 2 is a structural block diagram of an insole manufacturing system based on artificial intelligence in an embodiment of this specification;

[0030] Figure 3 is a terminal device for an insole manufacturing method based on artificial intelligence in an embodiment of this specification;

[0031] Figure 4 is a computer-readable storage medium for an insole manufacturing method based on artificial intelligence in an embodiment of this specification. DETAILED DESCRIPTION

[0032] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present disclosure. However, those skilled in the art will recognize that one or more of the specific details may be omitted, or other methods, components, devices, steps, etc. may be used. In other cases, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0033] In addition, the accompanying drawings are only schematic illustrations of the present disclosure. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0034] As Figure 1 shown, an insole manufacturing method based on artificial intelligence is provided in an embodiment of this specification. The execution subject of this method may be a personal computer, a server, etc. This method may specifically include the following steps S101 to S105:

[0035] In step S101, a set of defect type labels for the finished insole and a set of root causes corresponding to the set of defect type labels are defined, and defect photos corresponding to the set of defect type labels are collected to train a classification model based on a neural network.

[0036] Among them, determine the possible defect types of the finished insole, such as: size instability, warping, flash, shrinkage depression, cracks, etc. For each defect type, analyze and define the possible root causes, such as mold design, material problems, improper processing parameters, etc. Collect actual photos of each defect type, ensuring that the photos cover various different defect situations and severities. Preprocess the collected defect photos, including resizing, normalizing brightness and contrast, marking and annotating, etc. Classify the preprocessed photos according to the defect type labels to construct an annotated dataset for training. Use the constructed training dataset to train the selected neural network model to establish an effective neural network-based classification model for identifying and classifying the defect types of the finished insole, thereby improving the quality management and efficiency in the production process.

[0037] In step S102, continuously capture and collect images of the production process on the production line, establish an image degradation model, combine the image degradation model with a blind deconvolution filtering algorithm, and restore the motion-blurred image by minimizing an objective function in the form of a Lagrangian that includes a penalty function, and the penalty function uses the H 1 norm.

[0038] Among them, due to the inertial motion generated by the insole being driven during production, when collecting images while producing, within a very short exposure time, there is relative motion between the camera and the target image, and this motion blur causes the pixels of the captured image to drift. On the other hand, the environment in the production workshop will also affect the motion blur of the image, and the interference of noise, such as light and dark changes, dust, etc., will also reduce the quality of the image. Therefore, after the image is collected, recovery calculation is required to make the features of the image obvious.

[0039] Specifically, the degradation model is defined as:

[0040] g(x,y) = H(x,y) * f(x,y) + n(x,y);

[0041] g(x,y) is the degraded image, H(x,y) is the degradation function, f(x,y) is the original image, that is, it can be understood as a pure image without degradation and noise interference, and n(x,y) is the noise.

[0042] Due to problems such as inconsistent lighting and poor camera hardware performance in real life, it is difficult to reliably obtain the point spread function on the production site. Therefore, when the exact cause of some image degradation cannot be determined, a blind deconvolution filtering algorithm can be used to restore the original image as much as possible. When restoring the blurred image, construct the objective function of the blind deconvolution filter, make E(∫n 2 dx) = σ 2 , and introduce E(x) to calculate the expected value of the noise, and this objective function is expressed as:

[0043] ||h × f - g|| 2 = E[∫(h × f - g) 2 dx] = E(∫n 2 dx) = σ 2 E(x);

[0044] ||h × f - g|| 2 is the sum of squared residuals, which is the square of the Euclidean distance between the result of the convolution of the original image and the degradation function and the observed degraded image. E[∫(h × f - g) 2 dx] is the expected value, which is the average value of the integral of the sum of squared residuals over the entire image domain x. E(∫n 2 dx) is the expected value of the noise.

[0045] The Lagrangian form for minimizing the objective function is:

[0046] min L(f, h) = min[||h × f - g|| 2 + α 1 r(f) + α 2 r(h)];

[0047] L(f, h) is the Lagrangian function, r(f) and r(h) are penalty functions for the original image and the degradation function, and α 1 and α 2 are positive weighting coefficients, that is, the core calculation method is to minimize L(f, h).

[0048] According to the H 1 norm, the penalty function is defined. According to the rule such that the Lagrangian form is transformed into:

[0049]

[0050] Taking the partial derivatives of f and h and setting them to zero, the update rules in the frequency domain are obtained:

[0051]

[0052] F is the Fourier transform of f, H is the Fourier transform of h, G is the Fourier transform of g, and R(u, v) is an empirical formula estimated from the product normality of the corresponding production line.

[0053]

[0054] u and v are frequency variables in the Fourier transform, and M and N represent the dimensions of the image in the horizontal and vertical directions.

[0055] When restoring a blurred image, F and H are alternately updated until a preset convergence condition is met, and a restored clear image is obtained.

[0056] In step S103, the restored image is preprocessed so that the contrast of the highlight area of the preprocessed image is enhanced, the gray value of the low gray value area is reduced, and the noise in the image is reduced and the texture is enhanced. Based on the edge detection algorithm, the geometric edge features in the preprocessed image are extracted to form an image feature set.

[0057] Among them, the preprocessing includes: performing gray processing on the restored image based on gamma transformation; performing median filtering on the gray-processed image. Gamma transformation performs a non-linear operation on the gray value of the input restored image. After setting the gamma value according to the normal state of the field environment, the gray value of the output image has an exponential relationship with the gray value of the input image, improving the visibility and contrast of the image. Median filtering includes: selecting a pixel point in the image as the center of the convolution kernel, determining the neighborhood centered on the current pixel, which is usually square or circular, collecting the gray values of all pixels in the neighborhood, sorting the collected pixel gray values, finding the median among them, replacing the original pixel value at the center of the convolution kernel with the calculated median, repeating the operation, traversing each pixel point of the entire image matrix, and obtaining an image with noise removed.

[0058] When using the edge detection algorithm to extract geometric edge features, it includes: calculating the signal-to-noise ratio standard to evaluate the quality of the extracted edges; calculating the edge localization accuracy to make the extracted edge points close to the center of the actual edge; calculating the average distance of the zero-crossing points of the impulse response function. The edge detection algorithm can be the Canny algorithm.

[0059] Among them, in the edge detection process, correct edge information should not be ignored, and noise information should not be misinterpreted as an edge. The larger the signal-to-noise ratio, the higher the quality of the extracted edges. The signal-to-noise ratio is defined as:

[0060]

[0061] G(-x) is the edge function, and h(x) represents the impulse response of a filter with a width of ω.

[0062] In calculating the edge localization accuracy, the edge points of the algorithm should be as close as possible to the center of the actual edge, then calculate:

[0063]

[0064] Among them, L is the edge localization accuracy, G'(-x) and h'(x) respectively represent the derivative of the edge function and the impulse response. The larger L is, the higher the localization accuracy.

[0065] When different choices are given for the same edge and only one of the choices can be proven accurate, the average distance of the zero-crossing points of the impulse response function of the detection operator should satisfy:

[0066]

[0067] In step S104, based on the Hu-moment template matching mechanism, the similarity between the image feature set and the standard template is calculated. According to the comparison between the matching result and the threshold, the image feature set is classified as a qualified product or a defective product.

[0068] Among them, classifying the image feature set as a qualified product or a defective product includes: based on the Hu-moment theory, calculating the first Hu-moment data of the template image, and calculating the second Hu-moment data of the to-be-detected image represented by the image feature set; based on the contour matching of Hu-moments, matching the first Hu-moment data with multiple pieces of the second Hu-moment data to obtain multiple matching results; based on the arbitration function, when the matching result is higher than the threshold but lower than its scaling value, performing secondary preprocessing and geometric edge feature extraction on the to-be-detected image, and then performing secondary matching, and determining whether the to-be-detected image is qualified or unqualified according to the result of the secondary matching.

[0069] Specifically, the edge contour of an image has multiple features, which form the basis of the template matching mechanism. Moments are important operators for describing image features. Moments are features obtained by integrating all points on the contour. The description of pixel distribution by image moments can reflect the shape features of the image, and the relationship between pixels and the origin moments and centroid moments of the image is used to reflect the features of the image. Hu-moments are descriptions of the geometric features of an image and are invariant to rotation, translation, and scaling of the image. Moments can be calculated as:

[0070] M pq =∫∫x p y q I(x,y)dxdy(p,q=0,1,2…);

[0071] Among them, M pq represents the origin moment of order (p + q), p and q are the orders in the x and y directions respectively, I(x,y) is the intensity function of the image, representing the pixel value at the position (x,y), and dxdy represents the infinitesimal element for integrating x and y.

[0072] The contour matching based on Hu-moments is as follows:

[0073]

[0074] Among them, ψ i j is the result of template matching, where i corresponds to the image sequence and j represents the number of matches for the same image. It represents the summation of 7 Hu moments. m i k A and m i k B respectively represent the first Hu moment data of the template image and the second Hu moment data of the image to be detected (the image feature set obtained by edge detection). The matching result ψ i The smaller the value of j, the more similar the template image and the image to be detected are, and the higher the matching degree is.

[0075] The arbitration function can be defined as:

[0076] J(ψ ij ) = 1 (ψ ij ≤Ω);

[0077]

[0078] Ω is the threshold. When the value of J(ψ ij ) is 0, the above-mentioned secondary preprocessing and feature extraction are performed on the image, and then secondary matching is performed. If the result of the secondary matching is still higher than the threshold, the product of this image can be regarded as a defective product; if the value of J(ψ ij ) is 1, the product represented by this image is a qualified product; if the value of J(ψ ij ) is -1, the product of this image can be directly classified into defective products, and the image will be further classified in detail for defect types in the follow-up.

[0079] In step S105, based on the classification model, defect classification is performed on the image feature set classified as a fault, and combined with the root cause set, the causes leading to defects are eliminated.

[0080] In one embodiment, the defect type label set includes one or more of the following:

[0081] Size instability, material shortage, deformation, flash, shrinkage depression, cracking, surface crack, black spot coking.

[0082] As a supplement, the root cause set corresponding to the size instability includes: insufficient heating capacity, unstable feeding, unstable screw speed, uncontrolled temperature, malfunction of the proportional valve, malfunction of the main pressure valve, unstable back pressure, lack of strength and stiffness of the mold, non-wear-resistant core and cavity materials, unreasonable system, and too large temperature difference of the mold caused by the cooling system.

[0083] The root cause set corresponding to the material shortage includes: thermometer failure, improper selection of the inner hole diameter of the injection nozzle, too short time interval between two injections, and unreasonable pouring of the mold. The root cause set corresponding to the deformation includes:

[0084] The set of root causes of the shrinkage depression includes: the nozzle size of the injection machine is inappropriate, the clamping force is insufficient, the wall thickness of the mold is too thick, the cooling system causes inconsistent temperatures inside and outside the cavity, the heating system causes inconsistent temperatures inside and outside the cavity, and the sizes of the main runner, runner, and gate are inappropriate;

[0085] The set of root causes of the cracking includes: gas interference exists during the forming process of the plastic part.

[0086] The set of root causes of the surface crack includes: the ejection mechanism of the mold is unbalanced; the set of root causes of the black spot coking includes: the barrel of the injection machine is overheated or stuck, the mold exhaust is not smooth or the lubricant is not properly matched.

[0087] Based on the same idea, as Figure 2 shown, an exemplary embodiment of the present disclosure also provides an insole manufacturing system based on artificial intelligence, including: a model training module, configured to define a set of defect type labels for the finished insole and a set of corresponding root causes for the set of defect type labels, and collect defect photos corresponding to the set of defect type labels to train a classification model based on a neural network; an image restoration module, configured to continuously capture and collect images of the production process on the production line, establish an image degradation model, combine the image degradation model with a blind deconvolution filtering algorithm, and restore a motion-blurred image by minimizing an objective function in the form of a Lagrangian containing a penalty function, where the penalty function uses an H 1 norm; a preprocessing module, configured to preprocess the restored image to enhance the contrast of the highlight area and reduce the gray value of the low gray value area of the preprocessed image, and to reduce noise and enhance texture in the image, and extract geometric edge features in the preprocessed image based on an edge detection algorithm to form an image feature set; a matching module, configured to perform a similarity matching calculation on the image feature set and a standard template based on the Hu moment template matching mechanism, and divide the image feature set into qualified products or defective products according to the comparison of the matching result with a threshold; an analysis module, configured to perform defect classification on the image feature set classified as faulty based on the classification model, and combine with the set of root causes to eliminate the causes leading to the defects.

[0088] In this device, defective products and qualified products are quickly divided through a matching mechanism, which is more efficient and accurate than traditional visual inspection. Then, based on the defined set of defect type labels for the finished insole and the set of root causes, after quickly locating the defect type through the classification model, it is possible to directly locate the corresponding mold defect or injection machine defect, thereby quickly solving the root cause of the defect and improving the quality of the mold.

[0089] The specific details of each module in the above device have been described in detail in the embodiments of the method section. For the details not disclosed, please refer to the embodiments in the method section, so they will not be elaborated here.

[0090] Based on the same idea, the embodiments of this specification also provide an insole manufacturing device based on artificial intelligence, as Figure 3 shown.

[0091] The insole manufacturing device based on artificial intelligence can be the terminal device or server provided in the above embodiments.

[0092] The insole manufacturing device based on artificial intelligence may vary greatly due to different configurations or performances. It may include one or more processors 301 and a memory 302. One or more application programs or data may be stored in the memory 302. Among them, the memory 502 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) and / or a cache storage unit, and may further include a read-only storage unit. The application programs stored in the memory 302 may include one or more program modules (not shown in the figure). Such program modules include but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. Further, the processor 301 may be configured to communicate with the memory 302 and execute a series of computer-executable instructions in the memory 302 on the video privacy protection device. The video privacy protection device may also include one or more power supplies 303, one or more wired or wireless network interfaces 304, one or more I / O interfaces (input / output interfaces) 305, and communicate with one or more external devices 306 (such as a keyboard, pointing device, Bluetooth device, etc.), and may also communicate with one or more devices that enable a user to interact with the device, and / or communicate with any device that enables the device to communicate with one or more other computing devices (such as a router, modem, etc.). Such communication may be carried out through the I / O interface 305. And the device may also communicate with one or more networks (such as a local area network (LAN)) through the wired or wireless interface 304.

[0093] Specifically in this embodiment, the video privacy protection device includes a memory and one or more programs. One or more of the programs are stored in the memory, and one or more of the programs may include one or more modules. Each module may include a series of computer-executable instructions for the video privacy protection device and is configured to be executed by one or more processors. The one or more programs include the following computer-executable instructions:

[0094] Define a set of defect type labels for finished insoles and a corresponding set of root causes for the defect type labels, collect defect photos corresponding to the defect type label set, and train a classification model based on a neural network;

[0095] Continuously photograph and collect images of the production process on the production line, establish an image degradation model, combine the image degradation model with a blind deconvolution filtering algorithm, and restore the motion-blurred image by minimizing an objective function in the form of a Lagrangian that includes a penalty function. The penalty function uses the H 1 norm;

[0096] Preprocess the restored image to enhance the contrast of the highlight area and reduce the gray value of the low gray value area of the preprocessed image, and to reduce noise and enhance texture in the image. Based on an edge detection algorithm, extract the geometric edge features in the preprocessed image to form an image feature set;

[0097] Based on the Hu moment template matching mechanism, calculate the similarity between the image feature set and a standard template, and divide the image feature set into qualified products or defective products according to the comparison between the matching result and a threshold;

[0098] Based on the classification model, classify the defects in the image feature set classified as faulty, and combine the root cause set to eliminate the causes leading to the defects.

[0099] Based on the same idea, the exemplary embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon a program product capable of implementing the above-described method of this specification. In some possible embodiments, various aspects of the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Method" section of this specification.

[0100] Refer to Figure 4 As shown, a program product 400 for implementing the above method according to the exemplary embodiments of the present disclosure is described. It may be in the form of a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0101] The program product may employ any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0102] The computer readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which the readable program code is carried. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0103] The program code contained on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0104] The program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user's computing device, partially on the user's device, as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0105] Those skilled in the art can easily understand from the description of the above embodiments that the example embodiments described herein can be implemented by software or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, a USB flash drive, a portable hard drive, etc.) or on a network, including several instructions to enable a computing device (such as a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the exemplary embodiments of the present disclosure.

[0106] In addition, the above drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, rather than for limiting purposes. It is easy to understand that the processes shown in the above drawings do not indicate or limit the chronological order of these processes. Additionally, it is also easy to understand that these processes can be executed synchronously or asynchronously in, for example, multiple modules.

[0107] It should be noted that although several modules or units of devices for action execution are mentioned in the above detailed description, such a division is not mandatory. In fact, according to the exemplary embodiments of the present disclosure, the features and functions of two or more of the above-described modules or units can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0108] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure, which follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present disclosure are pointed out by the claims.

Claims

1. A method for manufacturing insoles based on artificial intelligence, characterized in that: The method comprises: Define a defect type label set of finished insoles and a root cause set corresponding to the defect type label set, collect defect photos corresponding to the defect type label set, and train a classification model based on a neural network; The production process on the production line is continuously photographed and imaged, an image degradation model is established, the image degradation model is combined with a blind deconvolution filtering algorithm, and the dynamic blurred image is restored by minimizing an objective function in the Lagrangian form containing a penalty function. The penalty function adopts H 1 norm; Preprocessing the restored image to enhance the contrast of the highlight area of ​​the preprocessed image and reduce the gray value of the low gray value area, as well as reduce the noise and enhance the texture in the image. Based on the edge detection algorithm, extract the geometric edge features in the preprocessed image to form an image feature set; Based on the Hu moment template matching mechanism, the image feature set is matched with the standard template for similarity calculation, and the image feature set is divided into qualified products or defective products according to the comparison between the matching result and the threshold value; Based on the classification model, the image features classified as faults are collectively classified into defect classification, and combined with the root cause set, the causes leading to the defects are eliminated, and then production adjustments are made.

2. The artificial intelligence-based insole manufacturing method according to claim 1, characterized in that: The pre-processing comprises: Performing grayscale processing based on gamma transformation on the restored image; Perform median filtering on the grayscale processed image.

3. The artificial intelligence-based insole manufacturing method according to claim 1, characterized in that: When using the edge detection algorithm to extract geometric edge features, it includes: Calculate the signal-to-noise ratio criterion to assess the quality of the extracted edges; Calculate edge positioning accuracy so that the extracted edge points are close to the center of the actual edge; Computes the average distance between the zero crossings of the impulse response function.

4. The artificial intelligence-based insole manufacturing method according to claim 1, characterized in that: Classifying the image feature set into qualified products or defective products includes: Based on the Hu moment theory, first Hu moment data of the template image is calculated, and second Hu moment data of the image to be detected represented by the image feature set is calculated; Based on Hu moment contour matching, matching the first Hu moment data with a plurality of the second Hu moment data to obtain a plurality of matching results; Based on the arbitration function, when the matching result is higher than the threshold but lower than its scaling value, the image to be detected is subjected to secondary preprocessing and geometric edge feature extraction, and then secondary matching is performed, and whether the image to be detected is qualified or unqualified is determined according to the result of the secondary matching.

5. The method for manufacturing an insole based on artificial intelligence according to claim 1, characterized in that: The defect type label set includes one or more of the following: Dimensional discrepancy, deformation, flash, cracking, surface cracks, black spots and carbonization.

6. An artificial intelligence-based insole manufacturing system, characterized in that: include: A model training module, used to define a defect type label set of a finished insole and a root cause set corresponding to the defect type label set, collect defect photos corresponding to the defect type label set, and train a classification model based on a neural network; The image restoration module is used to continuously shoot and collect images of the production process on the production line, establish an image degradation model, combine the image degradation model with the blind deconvolution filtering algorithm, and restore the dynamic blurred image by minimizing an objective function in the Lagrangian form containing a penalty function. The penalty function adopts H 1 norm; A preprocessing module is used to preprocess the restored image, so that the contrast of the highlight area of ​​the preprocessed image is enhanced and the gray value of the low gray value area is reduced, and the noise in the image is reduced and the texture is enhanced. Based on the edge detection algorithm, the geometric edge features in the preprocessed image are extracted to form an image feature set; A matching module, used for performing matching calculation of the similarity between the image feature set and the standard template based on the Hu moment template matching mechanism, and classifying the image feature set into qualified products or defective products according to the comparison between the matching result and the threshold value; The analysis module is used to classify defects in the image features classified as faults based on the classification model, and make production adjustments after eliminating the causes of the defects in combination with the root cause set.

Images