A quality inspection method, system and storage medium for shoes

Abstract

The present invention discloses a quality detection method, system and storage medium for shoes. The method includes: obtaining an appearance image of a shoe component using a first photographing device to obtain a first image; obtaining an image of a label and / or a key position of the shoe component using a second photographing device to obtain a second image; the shoe component includes: a left component and a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole; based on a first preset network model, processing the first image and / or the second image to obtain the type of the shoe component; selecting a second preset network model according to the type; based on the second preset network model, detecting the first image to determine whether there are defects in the shoe component. Through the embodiments of the present invention, artificial intelligence is used to achieve quality detection of shoes, replacing manual detection so as to reduce the detection cost; by deploying an artificial intelligence detection model, it can be called by multiple quality detection workstations, and the quality inspection standards are unified.

Description

Technical Field

[0001] The present invention relates to the technical field of quality inspection, and particularly to a quality inspection method, system and storage medium for shoes. Background Art

[0002] In modern society, people increasingly value the quality of life and pay more attention to the quality of shoes. In the production and circulation of shoes, in order to meet the requirements of high quality, quality inspection is required. However, the existing inspection methods are mostly manual inspections. Manual inspections have their drawbacks. For example, training is required for inspections. Manual inspections take a long time and can cause fatigue. Individual differences among workers can also lead to inconsistent quality inspection standards, resulting in missed and misdetected inspections and low efficiency. At the same time, the cost of manual inspections is also relatively high. Summary of the Invention

[0003] The main object of the present invention is to provide a quality inspection method, system and storage medium for shoes, aiming to solve the problems of high cost and inconsistent quality inspection standards caused by manual quality inspection of shoes in the prior art.

[0004] To achieve the above object, the present invention provides a quality inspection method for shoes, the method comprising:

[0005] Obtaining an appearance image of a shoe component using a first photographing device to obtain a first image; obtaining an image of a label and / or a key position of the shoe component using a second photographing device to obtain a second image; the shoe component comprising: a left component, a right component; the left component comprising: a left shoe body and / or a left insole; the right component comprising: a right shoe body and / or a right insole;

[0006] Processing the first image and / or the second image based on a first preset network model to obtain the type of the shoe component;

[0007] Selecting a second preset network model according to the type;

[0008] Detecting the first image based on the second preset network model to determine whether there is a defect in the shoe component.

[0009] Optionally, the method further comprises the following steps:

[0010] Taking an appearance image of the left component using a third photographing device to obtain a third image; taking an appearance image of the right component using a fourth photographing device to obtain a fourth image;

[0011] Input the third image into a third pre-set network model to perform the left component detection on the image and obtain a first point set of the left component; input the fourth image into the third pre-set network model to perform the right component detection on the image and obtain a second point set of the right component.

[0012] Perform linear fitting on the first point set to obtain a first fitting line; perform linear fitting on the second point set to obtain a second fitting line.

[0013] Judge whether the overlapping ratio of the first fitting line and the second fitting line is greater than a threshold. If it is greater than the threshold, determine that the sizes of the left component and the right component are the same.

[0014] Optionally, the third photographing device is installed above the central position of the left component, and the fourth photographing device is installed above the central position of the right component.

[0015] The first pitch angle of the third photographing device is equal to the second pitch angle of the fourth photographing device.

[0016] The first height from the third photographing device to the left component is equal to the second height from the fourth photographing device to the right component.

[0017] Optionally, the method further includes the following steps:

[0018] Input the first image into a fourth pre-set network model to compare the left component and the right component and judge whether there is a color difference between the left component and the right component.

[0019] Optionally, the number of the second pre-set network models is the same as the number of the types.

[0020] The second pre-set network model is trained using the annotation images corresponding to the matching types.

[0021] Optionally, the method further includes the following steps:

[0022] Before photographing the appearance images of the shoe components, place the left shoe body, the right shoe body, the left insole, and the right insole on a test fixture respectively.

[0023] The test fixture is conveyed to the detection position through a conveying mechanism.

[0024] Optionally, the method further includes the following steps:

[0025] After detecting and judging that the shoe component has a defect based on the second pre-set network model for the first image, manually confirm the defective shoe component.

[0026] When the manual confirmation determines that it is a good product, mark the shoe component as a good product and obtain the first image of the shoe component at the same time;

[0027] Annotate the first image of the shoe component, and the annotated image is used for the training of the second preset network model.

[0028] In addition, to achieve the above object, the present invention also proposes a quality inspection system for shoes, including a test fixture, a conveying mechanism, a photographing device, and a server;

[0029] The test fixture is used to fix the shoe components, and the shoe components include: a left component and a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole;

[0030] The conveying mechanism is used to convey the test fixture to the test position;

[0031] The photographing device is used to obtain the appearance image of the shoe component to obtain the first image; it is also used to obtain the label of the shoe component and / or the image of the key position to obtain the second image;

[0032] The server is used to process the first image and / or the second image based on the first preset network model to obtain the type of the shoe component; it is also used to select the second preset network model according to the type; it is also used to detect the first image based on the second preset network model to determine whether there is a defect in the shoe component.

[0033] Optionally, the server includes:

[0034] A classification module, which is used to process the first image and / or the second image based on the first preset network model to obtain the type of the shoe component;

[0035] A selection module, which is used to select the second preset network model according to the type;

[0036] A detection module, which is used to detect the first image based on the second preset network model to determine whether there is a defect in the shoe component.

[0037] Optionally, the server further includes:

[0038] An acquisition module, which is used to input the third image into the third preset network model to perform the left component detection on the image and obtain the first point set of the left component; input the fourth image into the third preset network model to perform the right component detection on the image and obtain the second point set of the right component;

[0039] A fitting module, configured to perform linear fitting on the first point set to obtain a first fitting line; and perform linear fitting on the second point set to obtain a second fitting line.

[0040] A judgment module, configured to judge whether the overlapping ratio of the first fitting line and the second fitting line is greater than a threshold. If it is greater than the threshold, it is determined that the sizes of the left component and the right component are consistent.

[0041] In addition, to achieve the above object, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the shoe quality detection method described above are implemented.

[0042] From the above technical methods, it can be seen that the present application has the following advantages:

[0043] In the present application, artificial intelligence is used to implement the quality detection of shoes, replacing manual detection to reduce the detection cost; by deploying an artificial intelligence detection model, it can be called by multiple quality detection workstations, and the quality inspection standards are unified. Description of the Drawings

[0044] Figure 1 It is a schematic flowchart of a shoe quality detection method provided by the present invention.

[0045] Figure 2 It is another schematic flowchart of a shoe quality detection method provided by the present invention.

[0046] Figure 3 It is another schematic flowchart of a shoe quality detection method provided by the present invention.

[0047] Figure 4 It is another schematic flowchart of a shoe quality detection method provided by the present invention.

[0048] Figure 5 It is another schematic flowchart of a shoe quality detection method provided by the present invention.

[0049] Figure 6 It is a structural block diagram of an embodiment of a server of the present invention.

[0050] Figure 7 It is a structural block diagram of another embodiment of a server of the present invention.

[0051] Figure 8 It is a schematic structural diagram of an embodiment of a shoe quality detection system provided by the present invention.

[0052] Figure 9 It is a schematic structural diagram of a hardware operating environment related to the embodiment solution of the present invention.

[0053] The implementation, functional features and advantages of the present invention will be further described in conjunction with embodiments with reference to the accompanying drawings. Detailed implementation manners

[0054] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not used to limit the present invention.

[0055] In subsequent descriptions, suffixes such as "module", "component" or "unit" used to represent elements are only for the convenience of describing the present invention, and they have no specific meaning in themselves. Therefore, "module", "component" or "unit" can be used interchangeably.

[0056] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence.

[0057] It should be understood that this application is applied to a quality inspection system for shoes. Please refer to Figure 8 , Figure 8 which is a schematic structural diagram of an embodiment of the quality inspection system for shoes in an embodiment of the present application. As Figure 8 shown, Figure 8 it includes a test fixture 400, a conveying mechanism 300, a photographing device 200, and a server 100;

[0058] The test fixture 100 is used to fix shoe components. The shoe components include: left components and right components; the left components include: a left shoe body and / or a left insole; the right components include: a right shoe body and / or a right insole;

[0059] The conveying mechanism 300 is used to convey the test fixture 400 to the test position;

[0060] The photographing device 200 is used to obtain an appearance image of the shoe component to obtain a first image; it is also used to obtain an image of a label and / or a key position of the shoe component to obtain a second image;

[0061] The server 100 is used to process the first image and / or the second image based on a first preset network model to obtain the type of the shoe component; it is also used to select a second preset network model according to the type; it is also used to detect the first image based on the second preset network model to determine whether there are defects in the shoe component.

[0062] In one embodiment, as Figure 1 shown, a quality inspection method for shoes provided by the present invention includes:

[0063] Step S101: Use a first imaging device to obtain an appearance image of the shoe component to get a first image; use a second imaging device to obtain an image of the label and / or key positions of the shoe component to get a second image; the shoe component includes: a left component and a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole.

[0064] Install an industrial camera at the test station and take pictures of the appearance of the shoe components to be detected, such as the left shoe body, the right shoe body, the left insole, and the right insole. The specific number of industrial cameras to be installed and the installation positions of the cameras are determined according to the appearance images to be taken. To completely capture the appearance of the shoe, it is recommended to install 6 industrial cameras at 6 view surfaces of the shoe to be detected to take the appearance images of the 6-sided view of the shoe. A movable industrial camera can also be installed and moved inside the shoe to take pictures of the inside of the shoe. It is also possible to move the jig on which the shoe is installed to complete the images of the 6-sided view of the shoe body and the inside of the shoe body. The shoe component may also include a shoe box.

[0065] The shoe body or the shoe box needs to have the appearance images of the 6-sided view and the images of the inside taken, and the insole needs to have the appearance images of the 6-sided view taken. To facilitate subsequent classification of the shoe body, the shoe box, and the insole, if the 6 sides of the shoe body, the shoe box, and the insole are not all photographed, such as only the upper and lower appearance pictures of the shoe body are taken, then it is necessary to take the appearance picture including the trademark part of the shoe body, the appearance picture including the size of the shoe body, and the appearance picture distinguishing the left and right shoe bodies. The appearance pictures of the insole are processed in the same way as the shoe body.

[0066] The pictures taken are as shown in the following table:

[0067]

[0068]

[0069] Step S102: Based on a first preset network model, process the first image and / or the second image to obtain the type of the shoe component.

[0070] Deploy a first preset network model on the server. The first preset network model can be various types of network models capable of extracting image features. For example, the first preset network model can refer to a U-Net (U-Network) model, a PSPNet (Pyramid Scene Parsing Network) model, a DenseNet (Dense Convolutional Network) model, a ResNet (Residual Network) model, or a MobileNet (Mobile Network) model, etc. The first preset network model itself can have initial parameters, and the initial parameters can be the parameters pre-trained on the ImageNet dataset. During the first iteration of training, the first preset network model is trained based on the initial parameters.

[0071] In the embodiments of the present application, a supervised model training method is adopted, so the sample images have annotations for labeling the true recognition results of the sample images. In the embodiments of the present application, the recognition results of the sample images include classification results. Accordingly, classification annotations are performed.

[0072] Classify the types of shoes and shoe boxes according to the product number of the shoes, and can also be further classified according to the size of the shoes. The classification results can be as shown in the following table:

[0073]

[0074] According to the classification identifier, obtain the appearance images of the shoe body, shoe box, and insole corresponding to the classification identifier, including the appearance pictures with labels and sizes. Annotate these pictures and label the attribution type of each picture. Then use these annotated pictures to train the first preset network model to obtain the trained first preset network model.

[0075] For the appearance pictures of the shoe body taken by the industrial camera, which can also be pictures containing labels or special positions, send them to the trained first preset network model for processing to obtain the types to which these pictures belong. When classifying multiple pictures taken by the industrial camera at a single detection station, there are multiple corresponding types. Then, take the type with the largest number of photos included in that type as the standard. The classification results are as shown in the following table:

[0076]

[0077] Then, determine that the type of the shoe body is: ABPR017-5-A-L-39.

[0078] Step S103: Select a second preset network model according to the type.

[0079] The second preset network model can be a convolutional neural network or a deep learning neural network. This embodiment does not limit the specific type of neural network.

[0080] Obtain the appearance pictures and internal pictures of the shoe body and shoe box of each type, and the appearance picture of the insole. Some of these images are of defective shoes, and some are of normal shoes. By annotating these pictures, and then using the annotated pictures for training the second preset network model. A trained second preset network model is obtained for each classification of shoes. As described in the following table:

[0081] Classification identifier Second preset network model ABPR017-5-A-L-39 ABPR017-5-A-L-39 - Network model ABPR017-5-A-R-39 ABPR017-5-A-R-39 - Network model ABPR017-5-A-L-43 ABPR017-5-A-L-43 - Network model ABPR017-5-A-R-43 ABPR017-5-A-R-43 - Network model ABPR017-5-A-L-47 ABPR017-5-A-L-47 - Network model ABPR017-5-A-R-47 ABPR017-5-A-R-47 - Network model

[0082] After identifying the type of shoe to be detected by the first preset network model, such as: ABPR017-5-A-L-39. Obtain the trained second preset network model corresponding to this type, such as: ABPR017-5-A-L-39-network model.

[0083] Step S104, detect the first image based on the second preset network model, and determine whether there are defects in the shoe components.

[0084] After selecting the trained second preset network model corresponding to the type of shoe, input the appearance picture of the shoe body or shoe box to be detected, the internal picture of the shoe body or shoe box, or the appearance picture of the insole. The specific pictures are determined according to the shoe components to be detected currently. For example, if the left shoe body is to be detected currently. Then input the appearance picture and the internal picture of the shoe body of the left shoe body into the trained second preset network model corresponding to the type of this shoe body for detection. For example, detect using the ABPR017-5-A-L-39-network model. The input pictures are as described in the following table:

[0085]

[0086] After the ABPR017-5-A-L-39-network model detects these pictures and determines that some pictures have defects, such as stains. Then it is determined that the left shoe body has defects, and at the same time, the position with stains is marked by a block diagram on the picture with stains. If it is detected that none of these pictures have defects, then it is determined that the left shoe body has no defects and belongs to good products, and the quality inspection passes.

[0087] After the quality inspection is completed, display the quality inspection results in the UI interface. If it is detected that the shoe has defects, then mark the picture with defects by a block diagram, and then display the picture in the UI interface. At the same time, the user can be reminded in other ways, such as by sound and light. For example, if the quality inspection of the shoe shows defects, turn on the red indicator light; if the quality inspection of the shoe is normal, turn on the green indicator light.

[0088] Through the embodiments of the present invention, artificial intelligence is used to achieve quality inspection of shoes, replacing manual inspection to reduce inspection costs; by deploying an artificial intelligence inspection model, it can be called by multiple quality inspection workstations, and the quality inspection standards are unified.

[0089] In one embodiment, Figure 2 provides Figure 1 a process for detecting the dimensional consistency of shoes after completing the quality inspection of the appearance defects of shoes.

[0090] Step S105: Use a third photographing device to photograph the appearance image of the left component to obtain a third image; use a fourth photographing device to photograph the appearance image of the right component to obtain a fourth image.

[0091] After inspecting the appearance of the shoes, it is also necessary to inspect the consistency of the shoes, such as whether the sizes of the left and right shoes are the same.

[0092] Install an industrial camera above the shoes or insoles to be inspected, such as installing an industrial camera A above the shoes or insoles of the left foot to be inspected; install an industrial camera B above the shoes or insoles of the right foot to be inspected. When installing these two cameras, it is necessary to ensure that the first pitch angle of industrial camera A is equal to the second pitch angle of industrial camera B;

[0093] The first height from industrial camera A to the left component (shoes or insoles) of the shoes to be inspected is equal to the second height from industrial camera B to the right component (shoes or insoles) of the shoes to be inspected.

[0094] Industrial camera number Height Pitch angle Industrial camera A 96 cm 89 degrees Industrial camera B 96 cm 89 degrees

[0095] Industrial camera A takes a front view of the left shoe body or left insole to obtain a front image A; industrial camera B takes a front view of the right shoe body or right insole to obtain a front image B. Then, by moving the jig, the back images of the shoes can be taken respectively to obtain back image A and back image B. As described in the following table:

[0096]

[0097] Step S106: Input the third image into a third preset network model to perform detection on the left component of the image to obtain a first point set of the left component; input the fourth image into the third preset network model to perform detection on the right component of the image to obtain a second point set of the right component.

[0098] Input the pictures of the left shoe body or left insole taken by industrial camera A into the trained deep learning model, detect the shoes or insoles in the input images, and obtain the first point set expression of the left shoe body or left insole on the images; input the pictures of the right shoe body or right insole taken by industrial camera B into the trained deep learning model, detect the shoes or insoles in the input images, and obtain the second point set expression of the right shoe body or right insole on the images.

[0099]

[0100] Step S107: Perform linear fitting on the first point set to obtain a first fitting line; perform linear fitting on the second point set to obtain a second fitting line.

[0101] Obtain the X and Z coordinate values of the point set (such as the front point set A) in the camera coordinate system to get the point set W'1(X i ,Z i ). From the characteristics of the point set, it can be known that they satisfy a certain linear relationship X = K×Z + B. Therefore, the least squares method is used for fitting, where the parameters K and B are the parameters to be determined. Find a set of K and B such that all the points in W'1 satisfy X i = K×Z i + B. However, it is obvious that for W'1 which is not all on the same straight line, it is impossible to find a set of K and B such that the left and right sides of the above equation are equal. Therefore, only the gap between X i and K×Z i + B can be minimized as much as possible. If P represents a set of K and B, then f(Z i , P) = K×Z i + B. Then the goal is to find a set of P to minimize the value of the function S in formula (1). When and only when S takes the minimum value, the linear relationship X = K×Z + B corresponding to P at this time is the best solution for fitting the point set.

[0102]

[0103] Apply the least squares method to the point set, and the fitting result line L, that is, the fitting line, can be calculated. The fitting results are shown in the following table:

[0104]

[0105] Step S108: Determine whether the overlapping ratio of the first fitting line and the second fitting line is greater than the threshold. If it is greater than the threshold, determine that the sizes of the left component and the right component are consistent.

[0106] Judge the overlapping ratio of the fitting lines on the front or back of the left and right shoe bodies or insoles. For example, judge the overlapping ratio of the fitting line on the front of the shoe as shown in the following table:

[0107] Left shoe body Front fitting line A Right shoe body Front fitting line B

[0108] If the overlapping ratio of the front fitting line A and the front fitting line B is greater than or equal to 95%, it indicates that the front dimensions of the left and right shoe bodies are the same. If the overlapping ratio is less than 95%, it indicates that the front dimensions of the left and right shoe bodies are different.

[0109] Then, judge the fitting line on the back of the shoe body:

[0110] Left shoe body Back fitting line A Right shoe body Back fitting line B

[0111] If the overlapping ratio of the back fitting line A and the back fitting line B is greater than or equal to 95%, it indicates that the back dimensions of the left and right shoe bodies are the same. If the overlapping ratio is less than 95%, it indicates that the back dimensions of the left and right shoe bodies are different.

[0112] The treatment method of the insole is the same as that of the shoe body.

[0113] The threshold of the overlapping ratio is not limited and can be adjusted according to the actual test results. For example, the threshold is adjusted to 96%.

[0114] Through the above processing, it can be obtained whether the left and right shoe bodies or insoles are the same. If they are different, the user will be prompted on the UI interface.

[0115] Through the embodiments of the present invention, the size consistency detection of the left and right shoe bodies or insoles of the shoe is realized through artificial intelligence, replacing manual detection so as to reduce the detection cost; by deploying an artificial intelligence detection model, it can be called by multiple quality inspection workstations, and the quality inspection standards are unified.

[0116] In one embodiment, Figure 3 provides Figure 1 The process of detecting the color difference consistency of the shoe after completing the quality inspection of the appearance defects of the shoe in the embodiment.

[0117] Step S109: Input the first image into a fourth preset network model, compare the left part and the right part, and judge whether there is a color difference between the left part and the right part.

[0118] The fourth preset network model can be a convolutional neural network or a deep learning neural network. The specific type of neural network is not limited in this embodiment.

[0119] Obtain the appearance images of the same parts of the left and right shoe bodies or insoles respectively. There are color differences in the same parts of the left and right shoe bodies or insoles of some shoes, and there are no color differences in some. By labeling these pictures, and then using the labeled pictures for training the fourth preset network model.

[0120] Each time, two appearance pictures of the same part of the left and right shoe bodies or insoles are obtained and input into the trained fourth preset network model. The fourth preset network model determines whether there is a color difference between the shoe bodies or insoles in these two pictures. If there is a color difference, it is displayed in the UI interface, and at the same time, the area with the color difference is marked by a frame icon.

[0121] Through the embodiments of the present invention, artificial intelligence is used to realize the color difference consistency detection of the left and right shoe bodies or insoles of shoes, replacing manual detection so as to reduce the detection cost; by deploying an artificial intelligence detection model, it can be called by multiple quality inspection workstations, and the quality inspection standards are unified.

[0122] In one embodiment, Figure 4 provided Figure 1 The embodiment provides a process for conveying shoes before the quality inspection of the appearance defects of the shoes.

[0123] Step S110: Before taking the appearance image of the shoe part, place the left shoe body, the right shoe body, the left insole, and the right insole on a test fixture respectively.

[0124] Manually disassemble the finished shoes into specific parts such as the left and right shoe bodies and the left and right insoles, and place each part on a platform fixture.

[0125] Step S111: The test fixture is conveyed to the detection position through a conveying mechanism.

[0126] The platform fixture on which the shoe body parts are placed is sent to the detection station equipped with an industrial camera through a conveying mechanism, such as a belt conveyor. Then, the industrial camera takes pictures for quality inspection.

[0127] Through the embodiments of the present invention, a fixture is placed for each shoe part, which is convenient for the industrial camera to take pictures.

[0128] In one embodiment, Figure 5 provided Figure 1 The embodiment provides a process for manually inspecting the shoes with defects after the quality inspection of the appearance defects of the shoes.

[0129] Step S112: After detecting and determining that the shoe part has a defect based on the second preset network model for the first image, manually confirm the defective shoe part.

[0130] Step S113: When the manual confirmation is a good product, mark the shoe part as a good product, and at the same time obtain the first image of the shoe part.

[0131] The shoes determined to be defective by artificial intelligence are removed manually in a jig, and then manually inspected for quality to determine whether the shoes are defective. If it is determined manually that the shoes are not defective, it means that the quality inspection by artificial intelligence is incorrect. The shoes need to be manually marked as shoes that pass the quality inspection. At the same time, obtain the appearance pictures taken when the artificial intelligence inspects the shoes, and label these appearance pictures, and the labeling result is that the shoes are normal.

[0132] Step S114: Label the first image of the shoe component, and the labeled image is used for training the second preset network model.

[0133] Input the labeled pictures with the labeling result that the shoes are normal into the second preset network model for training to improve the quality inspection accuracy of the second preset network model.

[0134] Through the embodiments of the present invention, the pictures corresponding to the shoes with incorrect quality inspection judgments by artificial intelligence are labeled, and then the labeled pictures are used to train the artificial intelligence, improving the quality inspection accuracy of the artificial intelligence.

[0135] In addition, the embodiments of the present invention also propose a server. Refer to Figure 6 , the server 100 includes:

[0136] A classification module 101, configured to process the first image and / or the second image based on a first preset network model to obtain the type of the shoe component;

[0137] A selection module 102, configured to select a second preset network model according to the type;

[0138] A detection module 103, configured to detect the first image based on the second preset network model to determine whether the shoe component is defective.

[0139] Through the embodiments of the present invention, artificial intelligence is used to implement quality inspection of shoes, replacing manual inspection to reduce inspection costs; by deploying an artificial intelligence detection model, it can be called by multiple quality inspection workstations, and the quality inspection standards are unified.

[0140] In addition, the embodiments of the present invention also propose a server. Refer to Figure 7 , on the basis of Figure 6 described above, it further includes:

[0141] An acquisition module 104, configured to input the third image into a third preset network model to perform the left component detection on the image to obtain a first point set of the left component; input the fourth image into the third preset network model to perform the right component detection on the image to obtain a second point set of the right component;

[0142] A fitting module 105 is configured to perform linear fitting on the first point set to obtain a first fitting line, and perform linear fitting on the second point set to obtain a second fitting line.

[0143] A judging module 106 is configured to judge whether an overlapping ratio of the first fitting line and the second fitting line is greater than a threshold. If it is greater than the threshold, it is determined that the sizes of the left component and the right component are consistent.

[0144] Through the embodiments of the present invention, artificial intelligence is used to realize the size consistency detection of the left and right shoe bodies or shoe pads of shoes, replacing manual detection so as to reduce the detection cost. By deploying an artificial intelligence detection model, it can be called by multiple quality inspection stations, and the quality inspection standards are unified.

[0145] In addition, an embodiment of the present invention also provides a quality inspection system for shoes. Refer to Figure 8 , the system includes: a test fixture 400, a conveying mechanism 300, a photographing device 200, and a server 100;

[0146] The test fixture 400 is configured to fix shoe components, and the shoe components include: a left component and a right component; the left component includes: a left shoe body and / or a left shoe pad; the right component includes: a right shoe body and / or a right shoe pad;

[0147] The conveying mechanism 300 is configured to convey the test fixture to a test position;

[0148] The photographing device 200 is configured to acquire an appearance image of the shoe component to obtain a first image, and is further configured to acquire an image of a label and / or a key position of the shoe component to obtain a second image;

[0149] The server 100 is configured to process the first image and / or the second image based on a first preset network model to obtain the type of the shoe component, and is further configured to select a second preset network model according to the type, and is further configured to detect the first image based on the second preset network model to judge whether there is a defect in the shoe component.

[0150] Through the embodiments of the present invention, artificial intelligence is used to realize the quality inspection of shoes, replacing manual inspection so as to reduce the inspection cost. By deploying an artificial intelligence inspection model, it can be called by multiple quality inspection stations, and the quality inspection standards are unified.

[0151] Refer to Figure 9 , Figure 9 It is a schematic structural diagram of a hardware operating environment of the server 100 involved in the solution of the embodiment of the present invention.

[0152] As Figure 9As shown in the figure, the hardware operating environment may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to implement connection communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as WI-FI, 4G, 5G interfaces). The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0153] Those skilled in the art can understand that Figure 9 the structure shown in does not constitute a limitation on the server 100, and it may include more or fewer components than shown in the figure, or combine some components, or have a different component layout.

[0154] As Figure 9 shown, the memory 1005, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a shoe quality detection program.

[0155] In Figure 9 the hardware operating environment shown, the network interface 1004 is mainly used for data communication with an external network; the user interface 1003 is mainly used to receive input instructions from users; the hardware operating environment calls the shoe quality detection program stored in the memory 1005 through the processor 1001 and performs the following operations:

[0156] Use a first photographing device to obtain an appearance image of the shoe component to obtain a first image; use a second photographing device to obtain an image of the label and / or key position of the shoe component to obtain a second image; the shoe component includes: a left component and a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole;

[0157] Based on a first preset network model, process the first image and / or the second image to obtain the type of the shoe component;

[0158] Select a second preset network model according to the type;

[0159] Based on the second preset network model, detect the first image to determine whether there are defects in the shoe component.

[0160] Optionally, the method further includes the following steps:

[0161] Use a third imaging device to capture an appearance image of the left component to obtain a third image; use a fourth imaging device to capture an appearance image of the right component to obtain a fourth image;

[0162] Input the third image into a third preset network model to perform detection of the left component on the image, and obtain a first point set of the left component; input the fourth image into the third preset network model to perform detection of the right component on the image, and obtain a second point set of the right component;

[0163] Perform linear fitting on the first point set to obtain a first fitting line; perform linear fitting on the second point set to obtain a second fitting line;

[0164] Determine whether the overlap ratio between the first fitting line and the second fitting line is greater than a threshold. If it is greater than the threshold, determine that the sizes of the left component and the right component are the same.

[0165] Optionally, the third imaging device is installed above the central position of the left component, and the fourth imaging device is installed above the central position of the right component;

[0166] The first pitch angle of the third imaging device is equal to the second pitch angle of the fourth imaging device;

[0167] The first height from the third imaging device to the left component is equal to the second height from the fourth imaging device to the right component.

[0168] Optionally, the method further includes the following steps:

[0169] Input the first image into a fourth preset network model to compare the left component and the right component, and determine whether there is a color difference between the left component and the right component.

[0170] Optionally, the number of the second preset network models is the same as the number of the types;

[0171] The second preset network model is trained using the labeled images corresponding to the matching types.

[0172] Optionally, the method further includes the following steps:

[0173] Before capturing the appearance images of the shoe components, place the left shoe body, the right shoe body, the left insole, and the right insole on a test fixture respectively;

[0174] The test fixture is conveyed to the detection position through a conveying mechanism.

[0175] Optionally, the method further includes the following steps:

[0176] After detecting the first image based on the second preset network model and determining that there is a defect in the shoe component, manually confirm the defective shoe component.

[0177] When the manual confirmation determines it is a good product, mark the shoe component as a good product, and at the same time obtain the first image of the shoe component.

[0178] Annotate the first image of the shoe component, and the annotated image is used for training the second preset network model.

[0179] Through the embodiments of the present invention, artificial intelligence is used to achieve quality inspection of shoes, replacing manual inspection to reduce inspection costs; by deploying an artificial intelligence detection model, it can be called by multiple quality inspection workstations, and the quality inspection standards are unified.

[0180] In addition, the embodiments of the present invention also propose a computer-readable storage medium, on which a quality inspection program for shoes is stored. When the quality inspection program for shoes is executed by a processor, the following operations are implemented:

[0181] Use a first photographing device to obtain an appearance image of a shoe component to obtain a first image; use a second photographing device to obtain an image of a label and / or a key position of the shoe component to obtain a second image; the shoe component includes: a left component and a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole;

[0182] Based on a first preset network model, process the first image and / or the second image to obtain the type of the shoe component.

[0183] Select a second preset network model according to the type.

[0184] Based on the second preset network model, detect the first image to determine whether there is a defect in the shoe component.

[0185] Optionally, the method further includes the following steps:

[0186] Use a third photographing device to photograph the appearance image of the left component to obtain a third image; use a fourth photographing device to photograph the appearance image of the right component to obtain a fourth image;

[0187] Input the third image into a third preset network model to perform detection on the left component of the image to obtain a first point set of the left component; input the fourth image into the third preset network model to perform detection on the right component of the image to obtain a second point set of the right component;

[0188] Perform a linear fit on the first point set to obtain a first fitting line; perform a linear fit on the second point set to obtain a second fitting line;

[0189] Determine whether the overlap ratio of the first fitting line and the second fitting line is greater than a threshold. If it is greater than the threshold, determine that the sizes of the left component and the right component are the same.

[0190] Optionally, the third photographing device is installed above the central position of the left component, and the fourth photographing device is installed above the central position of the right component;

[0191] The first pitch angle of the third photographing device is equal to the second pitch angle of the fourth photographing device;

[0192] The first height from the third photographing device to the left component is equal to the second height from the fourth photographing device to the right component.

[0193] Optionally, the method further includes the following steps:

[0194] Input the first image into a fourth preset network model, compare the left component and the right component, and determine whether there is a color difference between the left component and the right component.

[0195] Optionally, the number of the second preset network models is the same as the number of the types;

[0196] The second preset network model is trained using the labeled images corresponding to the matching types.

[0197] Optionally, the method further includes the following steps:

[0198] Before photographing the appearance image of the shoe component, place the left shoe body, the right shoe body, the left insole, and the right insole on a test fixture respectively;

[0199] The test fixture is transported to the detection position through a conveying mechanism.

[0200] Optionally, the method further includes the following steps:

[0201] After detecting and determining that the shoe component has a defect based on the second preset network model for the first image, manually confirm the defective shoe component;

[0202] When the manual confirmation is a good product, mark the shoe component as a good product, and at the same time obtain the first image of the shoe component;

[0203] Annotate the first image of the shoe component, and the annotated image is used for training the second preset network model.

[0204] Through the embodiments of the present invention, the quality inspection of shoes is realized through artificial intelligence, replacing manual inspection so as to reduce the inspection cost; by deploying an artificial intelligence inspection model, it can be called by multiple quality inspection workstations, and the quality inspection standards are unified.

[0205] It should be noted that in this article, the terms "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or system. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or system including the element.

[0206] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0207] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions to enable a terminal device (which can be a mobile phone, computer, server, controller, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0208] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied to other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A method for quality inspection of shoes, characterized in that, the method includes: Using a first photographing device to obtain an appearance image of a shoe component, obtaining a first image; using a second photographing device to obtain an image of a label and / or a key position of the shoe component, obtaining a second image; the shoe component includes: a left component, a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole; Based on a first preset network model, processing the first image and / or the second image to obtain the type of the shoe component; Selecting a second preset network model according to the type; Based on the second preset network model, detecting the first image to determine whether there are defects in the shoe component; Wherein, the method further includes the following steps: Using a third photographing device to photograph the appearance image of the left component, obtaining a third image; using a fourth photographing device to photograph the appearance image of the right component, obtaining a fourth image; Inputting the third image into a third preset network model to perform detection on the left component for the image, obtaining a first point set of the left component; inputting the fourth image into the third preset network model to perform detection on the right component for the image, obtaining a second point set of the right component; Performing linear fitting on the first point set to obtain a first fitting line; performing linear fitting on the second point set to obtain a second fitting line; Judging whether the overlapping ratio of the first fitting line and the second fitting line is greater than a threshold value. If it is greater than the threshold value, it is determined that the sizes of the left component and the right component are consistent.

2. The method according to claim 1, characterized in that, The third photographing device is installed above the central position of the left component, and the fourth photographing device is installed above the central position of the right component; The first pitch angle of the third photographing device is equal to the second pitch angle of the fourth photographing device; The first height from the third photographing device to the left component is equal to the second height from the fourth photographing device to the right component.

3. The method according to claim 1, characterized in that, The method further includes the following steps: Inputting the first image into a fourth preset network model to compare the left component and the right component to judge whether there is a color difference between the left component and the right component.

4. The method according to claim 1, characterized in that, The number of the second preset network models is the same as the number of the types; The second preset network model is trained using the corresponding labeled image of the matching type.

5. The method according to claim 1, characterized in that, The method further includes the following steps: Before photographing the appearance image of the shoe component, placing the left shoe body, the right shoe body, the left insole, and the right insole on a test fixture respectively; The test fixture is conveyed to the detection position through a conveying mechanism.

6. The method according to claim 1, characterized in that, The method further includes the following steps: After detecting and judging that there are defects in the shoe component based on the second preset network model, manually confirming the defective shoe component. When the manual confirmation determines it as a good product, mark the shoe component as a good product and obtain the first image of the shoe component at the same time. Annotate the first image of the shoe component, and the annotated image is used for the training of the second preset network model.

7. A quality inspection system for shoes Characterized in that it includes a test fixture, a conveying mechanism, a photographing device, and a server; The test fixture is used to fix the shoe components, and the shoe components include: a left component and a right component; the left component includes: a left shoe body and / or a left insole; the right component includes: a right shoe body and / or a right insole; The conveying mechanism is used to convey the test fixture to the test position; The photographing device is used to obtain the appearance image of the shoe component to obtain the first image; it is also used to obtain the label of the shoe component and / or the image of the key position to obtain the second image; The server is used to process the first image and / or the second image based on the first preset network model to obtain the type of the shoe component; it is also used to select the second preset network model according to the type; it is also used to detect the first image based on the second preset network model to determine whether there are defects in the shoe component; The photographing device is also used to photograph the appearance image of the left component to obtain the third image; photograph the appearance image of the right component to obtain the fourth image; The server is also used to input the third image into the third preset network model to perform the left component detection on the image to obtain the first point set of the left component; input the fourth image into the third preset network model to perform the right component detection on the image to obtain the second point set of the right component; perform linear fitting on the first point set to obtain the first fitting line; perform linear fitting on the second point set to obtain the second fitting line; determine whether the overlapping ratio of the first fitting line and the second fitting line is greater than the threshold, and if it is greater than the threshold, determine that the sizes of the left component and the right component are consistent.

8. The system according to claim 7 Characterized in that The server includes: A classification module for processing the first image and / or the second image based on the first preset network model to obtain the type of the shoe component; A selection module for selecting the second preset network model according to the type; A detection module for detecting the first image based on the second preset network model to determine whether there are defects in the shoe component.

9. A computer-readable storage medium, on which a computer program is stored Characterized in that When the computer program is executed by a processor, it implements the quality inspection method for shoes according to any one of claims 1-6.

Images