Gluing quality detection method and device based on deep learning, electronic equipment and storage medium

Abstract

The invention discloses a gluing quality detection method and device based on deep learning, electronic equipment and a storage medium. The method comprises the following steps: acquiring a to-be-detected adhesive tape image; inputting the adhesive tape image into a segmentation model, so that the segmentation model generates a segmentation mask of an adhesive tape region according to the adhesive tape image; removing the background in the adhesive tape image according to the segmentation mask of the adhesive tape region, and generating a region image only containing the adhesive tape; analyzing the regional image to generate geometric feature information of the adhesive tape; comparing the geometric feature information of the adhesive tape with a preset quality standard, and judging whether the gluing quality meets the standard or not; according to the invention, the stability and efficiency of gluing quality detection can be improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, device, electronic device and storage medium for gluing quality detection based on deep learning. Background Art

[0002] In modern industrial production, the quality of glue coating directly affects the performance and reliability of products, especially in the fields of automated assembly and precision manufacturing. The quality inspection of the glue coating process is not only related to the appearance inspection, but also involves aspects such as sealing, strength and durability. With the improvement of production efficiency, the quality control of glue coating is increasingly dependent on real-time monitoring and high-precision detection systems. Automated glue coating detection can ensure the consistency and stability of products, reduce manual intervention, improve production efficiency, and detect glue coating anomalies in a timely manner to prevent product defects from entering the market.

[0003] In the traditional gluing detection process, it is usually necessary to manually set the detection line parameters of each image frame by frame to collect the gluing area. This method becomes extremely cumbersome to set parameters when the gluing path is long or the number of gluing images is large. The manually set detection lines lack flexibility and are prone to errors when faced with complex gluing scenes, affecting the stability and accuracy of the detection. The root cause of these problems is that traditional detection methods rely too much on manual operation and lack intelligent and automated parameter adjustment mechanisms, making it difficult to ensure efficient and accurate detection results when faced with diverse and high-frequency gluing detection tasks. Summary of the invention

[0004] The embodiment of the present invention provides a method, device, electronic device and storage medium for gluing quality detection based on deep learning. The gluing quality detection can be realized by implementing the present invention.

[0005] An embodiment of the present invention provides a method for detecting glue coating quality based on deep learning, comprising:

[0006] Acquire an image of the adhesive strip to be detected;

[0007] Inputting the adhesive strip image into a segmentation model so that the segmentation model generates a segmentation mask of the adhesive strip area according to the adhesive strip image;

[0008] According to the segmentation mask of the rubber strip area, the background in the rubber strip image is removed to generate an area image containing only the rubber strip;

[0009] Analyzing the regional image to generate geometric feature information of the rubber strip;

[0010] Compare the geometric feature information of the rubber strip with the preset quality standard to determine whether the quality of the glue coating meets the standard;

[0011] The training operation of the segmentation model includes:

[0012] Obtaining a rubber strip image and a corresponding label for model training; wherein the label is a segmentation mask of the rubber strip area;

[0013] Constructing a first training set according to the rubber strip images and corresponding labels used for model training;

[0014] Randomly dividing the first training set into a plurality of batches of training samples according to a preset number;

[0015] Input each batch of training samples into the segmentation model in turn, and train the segmentation model until the preset number of training times is reached; wherein, when the prediction model receives each batch of training samples, it outputs the segmentation mask of the rubber strip area corresponding to the training sample; according to the segmentation mask of the rubber strip area and the corresponding label, the loss function value is calculated by the loss function; and the segmentation model is updated by the optimizer according to the loss function value.

[0016] Furthermore, before sequentially inputting the training samples of each batch into the segmentation model, the method further includes:

[0017] For each batch of training samples, data enhancement processing and normalization processing are performed to generate updated training samples; wherein the data enhancement processing includes any one or a combination of random horizontal flipping, random vertical flipping, random rotation, random adjustment of image brightness, random addition of Gaussian noise and random cropping.

[0018] Furthermore, before obtaining the rubber strip images and corresponding labels for model training, the following steps are also included:

[0019] Get a general visual feature dataset;

[0020] Extracting a second training set from the general visual feature dataset;

[0021] The segmentation model is pre-trained according to the second training set.

[0022] Furthermore, the analyzing the regional image to generate geometric feature information of the rubber strip includes:

[0023] The BLOB analysis is performed on the regional image to extract the starting point, the ending point, the area, the length and the width of the rubber strip in the regional image to obtain geometric feature information.

[0024] Furthermore, after determining whether the quality of the glue coating meets the standard, the method further includes:

[0025] When the quality of glue coating does not meet the standards, an abnormal signal is sent to the control system.

[0026] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0027] An embodiment of the present invention provides a glue coating quality detection device based on deep learning, comprising: an image acquisition module to be detected, a model training module, an instance segmentation module, a region image generation module, a geometric feature information generation module and a glue coating quality detection module;

[0028] The to-be-detected image acquisition module is used to acquire the to-be-detected rubber strip image;

[0029] The model training module is used to obtain the rubber strip images and corresponding labels used for model training; wherein the labels are segmentation masks of the rubber strip areas; construct a first training set based on the rubber strip images and corresponding labels used for model training; randomly divide the first training set into a number of batches of training samples according to a preset number; input the training samples of each batch into the segmentation model in turn, and train the segmentation model until a preset number of training times is reached; wherein, when the prediction model receives a batch of training samples, it outputs the segmentation mask of the rubber strip area corresponding to the training sample; calculates the loss function value through the loss function based on the segmentation mask of the rubber strip area and the corresponding label; and updates the segmentation model based on the loss function value using the optimizer;

[0030] The instance segmentation module is used to input the adhesive strip image into the segmentation model so that the segmentation model generates a segmentation mask of the adhesive strip area according to the adhesive strip image;

[0031] The regional image generation module is used to remove the background in the rubber strip image according to the segmentation mask of the rubber strip area, and generate a regional image containing only the rubber strip;

[0032] The geometric feature information generating module is used to analyze the regional image and generate geometric feature information of the rubber strip;

[0033] The glue coating quality detection module is used to compare the geometric feature information of the glue strip with the preset quality standard to determine whether the quality of the glue coating meets the standard.

[0034] Furthermore, the gluing quality detection device based on deep learning further includes: a data preprocessing module;

[0035] The data preprocessing module is used to perform data enhancement and normalization processing on each batch of training samples to generate updated training samples; wherein the data enhancement processing includes any one or a combination of random horizontal flipping, random vertical flipping, random rotation, random adjustment of image brightness, random addition of Gaussian noise and random cropping.

[0036] Furthermore, the deep learning-based gluing quality detection device further includes: an abnormal signal sending module;

[0037] The abnormal signal sending module is used to send an abnormal signal to the control system when the quality of the glue coating does not meet the standard.

[0038] Based on the above method embodiment, the present invention provides a corresponding electronic device embodiment.

[0039] An embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the deep learning-based glue quality detection method described in any one of the above-mentioned method embodiments can be implemented.

[0040] Based on the above method item embodiments, the present invention provides a corresponding storage medium item embodiment.

[0041] An embodiment of the present invention provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the deep learning-based glue quality detection method described in any one of the above method embodiments can be implemented.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The embodiment of the present invention provides a method, device, electronic device and storage medium for gluing quality detection based on deep learning. The method inputs the image of the rubber strip to be detected into a segmentation model, and the model automatically generates a segmentation mask of the rubber strip area. Then, the background in the image is removed according to the segmentation mask, and only the rubber strip area is retained to generate a clear regional image. Then, the regional image is processed using a preset analysis algorithm to extract the geometric feature information of the rubber strip. Finally, the feature information is compared with the preset quality standard to determine whether the gluing meets the quality requirements.

[0044] The present invention adopts a deep learning-based gluing detection method to achieve a detection process that no longer relies on manually setting the detection line parameters of each glue strip image frame by frame. Specifically, the segmentation model can automatically generate a segmentation mask for the glue strip area and remove the background area to generate an area image containing only the glue strip. This automated process effectively eliminates the tedious work of manually adjusting the detection line parameters in the traditional method, significantly reduces manual intervention, and improves detection efficiency and flexibility. In particular, when the gluing path changes, there is no need to reset the parameters, thereby ensuring the stability and efficiency of the detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1It is a flow chart of a method for gluing quality detection based on deep learning provided in one embodiment of the present invention.

[0046] Figure 2 It is a flowchart of the training operation of the segmentation model provided by one embodiment of the present invention.

[0047] Figure 3 It is a structural schematic diagram of a glue coating quality detection device based on deep learning provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, an embodiment of the present invention provides a method for gluing quality detection based on deep learning, which at least includes the following steps:

[0050] Step S1, obtaining an image of a rubber strip to be detected;

[0051] In a preferred embodiment, the image of the rubber strip to be detected is obtained by a gluing robot, which includes a robot glue gun for gluing and three cameras for taking images of the rubber strip. In order to obtain the image of the rubber strip to be detected, first, it is necessary to calibrate the relationship between the robot and the camera coordinate system so as to accurately calculate the motion trajectory of the glue gun in the image. Secondly, it is necessary to intelligently select the detection camera so that the camera corresponding to the image with the most centered trajectory is selected as the detection camera. After the calibration is completed, the actual gluing operation can be carried out. In the process of the robot controlling the glue gun to apply glue, the camera will continuously obtain the current image of the rubber strip, and detect and segment the glued area through the deep learning model.

[0052] Specifically, a two-dimensional code calibration plate is used to establish a conversion matrix between the robot tool coordinate system and the camera coordinate system. The calibration adopts the nine-point calibration method, and the high-precision two-dimensional code calibration plate is fixed in the robot workspace to ensure that it is stably placed within the camera field of view, avoid light interference, and ensure that the posture remains unchanged. Start the camera system, shoot the two-dimensional code calibration plate, extract the corner coordinates through the image processing algorithm, and use the PnP algorithm to calculate the posture of the calibration plate relative to the camera in combination with the actual calibration plate parameters to obtain the conversion matrix from the camera coordinate system to the calibration plate coordinate system. Control the robot to move the glue gun to align with the characteristic points of the calibration plate and record the posture data of the tool coordinate system. Through multiple calibrations under different postures, the average value is calculated to reduce the error, and the conversion matrix from the tool coordinate system to the calibration plate coordinate system is obtained. The conversion matrix from the tool coordinate system to the calibration plate coordinate system and the conversion matrix from the calibration plate to the camera coordinate system are multiplied to construct the conversion relationship between the camera coordinate system and the tool coordinate system, which is used to describe the relative position and posture of the two. The calibration accuracy is ensured through verification experiments, and the converted position data is compared with the actual detection results of the camera to ensure that the error is within an acceptable range.

[0053] Secondly, at each photo location, the system uses three cameras to take images of the gluing track. The position of the gluing track in each image is calculated through the calibrated coordinate transformation relationship, and the camera corresponding to the image with the track most centered is selected as the detection camera. This method avoids the problem of the colloid not being able to be fully photographed due to viewing angle limitations, thereby ensuring the integrity and accuracy of the detection.

[0054] After calibration, the actual gluing operation can be carried out. The camera takes pictures at set time intervals during the gluing process to collect standard images of the complete gluing process. These standard images are used to build a detection model for the glue strip, which serves as the standard for subsequent detection. The acquisition of standard images is the key link to ensure that the detection system can accurately compare and judge the gluing quality.

[0055] Step S2: input the adhesive strip image into a segmentation model, so that the segmentation model generates a segmentation mask of the adhesive strip area according to the adhesive strip image.

[0056] like Figure 2 As shown, in a preferred embodiment of the present invention, the training operation of the segmentation model includes:

[0057] Step S2.1, obtaining a rubber strip image and a corresponding label for model training; wherein the label is a segmentation mask of the rubber strip area;

[0058] Step S2.2: construct a first training set based on the rubber strip images and corresponding labels used for model training;

[0059] Step S2.3, randomly dividing the first training set into a plurality of batches of training samples according to a preset number;

[0060] Step S2.4, input each batch of training samples into the segmentation model in sequence, and train the segmentation model until the preset number of training times is reached; wherein, when the prediction model receives each batch of training samples, it outputs the segmentation mask of the rubber strip area corresponding to the training sample; according to the segmentation mask of the rubber strip area and the corresponding label, the loss function value is calculated by the loss function; and the segmentation model is updated according to the loss function value by using the optimizer. The loss function is a binary cross entropy loss and an intersection-over-union loss. The optimizer is an AdamW optimizer.

[0061] Optionally, before sequentially inputting each batch of training samples into the segmentation model, the method further includes:

[0062] For each batch of training samples, data enhancement processing and normalization processing are performed to generate updated training samples; wherein the data enhancement processing includes any one or a combination of random horizontal flipping, random vertical flipping, random rotation, random adjustment of image brightness, random addition of Gaussian noise and random cropping.

[0063] In an optional embodiment, before obtaining the rubber strip image and the corresponding label for model training, the method further includes:

[0064] Get a general visual feature dataset;

[0065] Extracting a second training set from the general visual feature dataset;

[0066] The segmentation model is pre-trained according to the second training set.

[0067] Specifically, the parameters of the segmentation model are initialized through a pre-training strategy. The pre-training is based on a general visual feature dataset and learns a rich distribution of image features, so that the model has good feature extraction capabilities. On this basis, the segmentation model can quickly adapt to the rubber strip segmentation task, shorten the training time and improve the task performance, which is particularly suitable for industrial scenarios with small data volumes.

[0068] Step S3, removing the background in the rubber strip image according to the segmentation mask of the rubber strip area, and generating a regional image containing only the rubber strip;

[0069] First, use the rubber strip area segmentation mask, which is a binary image of the same size as the original rubber strip image, where the white area represents the location of the rubber strip and the black area represents the background. Perform a pixel-by-pixel logical operation on the segmentation mask and the original rubber strip image, and set the pixel values ​​of the background part in the mask to zero, thereby removing the background and retaining only the pixel values ​​of the rubber strip area defined by the segmentation mask.

[0070] Subsequently, the initially generated image containing only the rubber strip area is post-processed, such as edge smoothing, isolated pixel removal, and color adjustment, to further optimize the visual effect and data quality of the rubber strip area. These post-processing steps help reduce the impact of segmentation errors on subsequent analysis and ensure that the generated image can accurately reflect the actual shape and characteristics of the rubber strip.

[0071] The generated image containing only the adhesive strip area can be used for subsequent adhesive strip quality analysis, including the measurement of geometric features such as adhesive strip width, consistency, and straightness, as well as the detection of adhesive strip coating integrity and surface defects such as bubbles, fractures, etc. By removing background interference, the accuracy and efficiency of the analysis process will be greatly improved.

[0072] Step S4, analyzing the regional image to generate geometric feature information of the rubber strip;

[0073] In a preferred embodiment, the analyzing the regional image to generate geometric feature information of the rubber strip includes:

[0074] The BLOB analysis is performed on the regional image to extract the starting point, the ending point, the area, the length and the width of the rubber strip in the regional image to obtain geometric feature information.

[0075] Specifically, the image of the area containing only the rubber strip is subjected to BLOB (Binary Large Object) analysis, which aims to identify continuous areas with similar features in the image. The BLOB analysis method scans the area image and identifies independent connected areas, i.e., the areas where the rubber strip is located. Each connected area can be regarded as a BLOB object, which contains multiple geometric information. Extracting the geometric feature information of each BLOB area mainly includes the following aspects:

[0076] Start and end points: By analyzing the outline of the rubber strip, determine the start and end points of the rubber strip in the image. These two points are usually the leftmost and rightmost ends (or the top and bottom) of the rubber strip in the image, which are very important for identifying the spatial position of the rubber strip. By connecting these two points, the overall direction of the rubber strip in the image can be roughly estimated.

[0077] Area: Calculate the area of ​​the adhesive strip area, that is, the number of all pixels in the BLOB area. Area information helps to evaluate the amount of adhesive strip applied and can be used to compare with standard images or adhesive coating standards to determine whether the adhesive strip meets the predetermined requirements.

[0078] Length: The length of the adhesive strip is obtained by calculating the Euclidean distance between the two farthest points in the adhesive strip area in the image. Length is an important indicator to measure the uniformity and integrity of adhesive coating, and can reflect whether the adhesive strip is coated according to the set path.

[0079] Width: By analyzing the horizontal pixel distribution of the adhesive strip, the widest part of the adhesive strip in the image is calculated to obtain the width of the adhesive strip. The width information can be used to detect whether the adhesive strip is evenly coated and whether there is any problem such as missing coating or the adhesive strip is too wide.

[0080] The obtained geometric feature information (starting point, end point, area, length, width, etc.) can be used for subsequent quality assessment and testing, including whether the adhesive strip is applied according to the expected path, size and shape. By comparing with the standard adhesive coating image, it is also possible to find defects or anomalies that may occur during the adhesive coating process, providing a basis for quality control and optimization in the production process.

[0081] Step S5, comparing the geometric feature information of the rubber strip with the preset quality standard to determine whether the quality of the glue coating meets the standard;

[0082] It should be noted that the range of various geometric characteristic parameters of the rubber strip is determined according to the preset quality standards. For example, the length between the starting point and the end point of the rubber strip should be within a certain tolerance range, the width of the rubber strip should meet the design requirements, and the area should match the predetermined coating amount. In addition, it is also necessary to set the geometric shape standards during the glue coating process. For example, the rubber strip should have a certain degree of straightness, the width should be kept uniform, and there should be no areas that are too narrow or too wide. For specific application scenarios, there may also be requirements for the surface integrity of the rubber strip (such as no bubbles, no breaks, etc.) and the consistency of the coating. Next, the geometric characteristic information of the rubber strip (such as starting point, end point, length, width, area, etc.) is compared item by item to determine whether it meets the preset standards. For example:

[0083] Length comparison: Compare the calculated length of the glue strip with the standard length. If the difference exceeds the allowable error range, it will be judged as unqualified. It may be that there is a gap in the glue coating process or the glue coating path is abnormal.

[0084] Width comparison: By comparing the actual width of the strip with the standard width range, check whether it is too wide or too narrow. If the strip width deviation is large, it may cause uneven coating, affecting the bonding effect or appearance quality of the strip.

[0085] Area comparison: Compare the area of ​​the strip area with the preset coating area to ensure that the amount of glue applied meets the standard. If the area is insufficient, it may mean that the glue is not applied enough, affecting the sealing or strength of the strip; if the area is too large, it may cause the coating of the strip to be too thick, thus affecting its functionality or appearance.

[0086] Shape and Flatness: The shape of the strip is evaluated to ensure the straightness and uniformity of the strip coating. For example, the strip coating should avoid obvious waves, twists or gaps to ensure its adhesion and sealing.

[0087] In addition, in some advanced applications, the integrity of the adhesive surface may also need to be considered, that is, whether the adhesive strip has defects such as bubbles, cracks or impurities. In these cases, image processing algorithms such as edge detection and texture analysis can be combined to further check whether the adhesive strip surface is smooth and meets the predetermined appearance standards.

[0088] By comparing the calculated geometric feature information with the preset standards, the system can automatically determine whether each gluing process meets the quality requirements. If all parameters are within the standard range, the gluing quality is considered qualified; if any parameter exceeds the standard range, the system will mark it as unqualified and may trigger an alarm or record abnormal information. This process helps to achieve real-time monitoring of gluing quality, ensure the accuracy and consistency of each gluing operation, thereby improving the stability of the production process and the reliability of the product.

[0089] In practical applications, automatic detection of glue coating quality not only improves production efficiency and reduces manual detection errors, but also provides data support for the optimization of the production process. By analyzing unqualified cases, companies can adjust production processes or equipment in a timely manner to further improve product quality and production efficiency.

[0090] In a preferred embodiment, after judging whether the quality of the glue coating meets the standard, the method further includes:

[0091] When the quality of glue coating does not meet the standards, an abnormal signal is sent to the control system.

[0092] After completing the characteristic analysis of the rubber strip, the system will perform further processing based on the detected results. If a quality problem is detected in the rubber strip (such as abnormal width, broken glue or uneven glue coating), the system will immediately evaluate the type and location of the abnormality and generate a feedback signal. This feedback signal will be transmitted to the PLC (Programmable Logic Controller). The PLC is responsible for controlling the robot's actions based on the detection results to ensure the smooth progress of the production process. Specifically, the PLC can adjust the production process by pausing the operation, adjusting the robot's action parameters, modifying the glue gun's glue coating speed or pressure, etc. to ensure that the glue coating quality meets the predetermined standards.

[0093] In addition, the PLC can also trigger an alarm when an abnormality is detected, notifying the operator to intervene in time to avoid the production of defective products. For example, the system will prompt the operator to check the equipment through screen display or sound alarm, quickly locate the problem and take necessary measures, such as cleaning the glue gun, adjusting the equipment, etc., to avoid the production of defective products.

[0094] This feedback mechanism enables the system to achieve closed-loop control during the gluing process, ensuring product quality through real-time monitoring and automatic adjustment. During the gluing process, the system can not only accurately identify and analyze the quality of the glue strips through the deep learning model, but also combine with traditional automated control systems (such as PLC) to control the production process in real time. Through the collaborative work of deep learning and PLC, the entire production line achieves efficient adaptive control, can automatically adjust the production process according to real-time data, reduce manual intervention, and improve the level of automation, thereby greatly improving production efficiency and quality stability.

[0095] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0096] like Figure 3 As shown, an embodiment of the present invention provides a glue coating quality detection device based on deep learning, including: an image acquisition module to be detected, a model training module, an instance segmentation module, a regional image generation module, a geometric feature information generation module and a glue coating quality detection module;

[0097] The to-be-detected image acquisition module is used to acquire the to-be-detected rubber strip image;

[0098] The model training module is used to obtain the rubber strip images and corresponding labels used for model training; wherein the labels are segmentation masks of the rubber strip areas; construct a first training set based on the rubber strip images and corresponding labels used for model training; randomly divide the first training set into a number of batches of training samples according to a preset number; input the training samples of each batch into the segmentation model in turn, and train the segmentation model until a preset number of training times is reached; wherein, when the prediction model receives a batch of training samples, it outputs the segmentation mask of the rubber strip area corresponding to the training sample; calculates the loss function value through the loss function based on the segmentation mask of the rubber strip area and the corresponding label; and updates the segmentation model based on the loss function value using the optimizer;

[0099] The instance segmentation module is used to input the adhesive strip image into the segmentation model so that the segmentation model generates a segmentation mask of the adhesive strip area according to the adhesive strip image;

[0100] The regional image generation module is used to remove the background in the rubber strip image according to the segmentation mask of the rubber strip area, and generate a regional image containing only the rubber strip;

[0101] The geometric feature information generating module is used to analyze the regional image and generate geometric feature information of the rubber strip;

[0102] The glue coating quality detection module is used to compare the geometric feature information of the glue strip with the preset quality standard to determine whether the quality of the glue coating meets the standard.

[0103] In a preferred embodiment, the deep learning-based gluing quality detection device further includes: an abnormal signal sending module;

[0104] The abnormal signal sending module is used to send an abnormal signal to the control system when the quality of the glue coating does not meet the standard.

[0105] In another preferred embodiment, the gluing quality detection device based on deep learning further includes: a data preprocessing module;

[0106] The data preprocessing module is used to perform data enhancement and normalization processing on each batch of training samples to generate updated training samples; wherein the data enhancement processing includes any one or a combination of random horizontal flipping, random vertical flipping, random rotation, random adjustment of image brightness, random addition of Gaussian noise and random cropping.

[0107] It should be noted that the embodiments of the device described above correspond to the above embodiments of the present invention, and can implement any of the methods described above in the present invention. In addition, the embodiments of the above-mentioned device are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0108] Based on the above method embodiment of the present invention, a corresponding electronic device embodiment is provided.

[0109] An embodiment of the present invention provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the deep learning-based glue quality detection method described in any one of the present invention is implemented, or, when the processor executes the computer program, the functions of each module in the above-mentioned device embodiments are implemented.

[0110] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0111] The terminal device may be a computing device such as a desktop computer, a notebook, a palmtop computer, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0112] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0113] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0114] Based on the above method embodiment, the present invention provides a storage medium embodiment;

[0115] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute any one of the above-mentioned deep learning-based glue quality detection methods of the present invention.

[0116] The storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0117] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0118] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for gluing quality detection based on deep learning, characterized in that: include: Acquire an image of the adhesive strip to be detected; Inputting the adhesive strip image into a segmentation model so that the segmentation model generates a segmentation mask of the adhesive strip area according to the adhesive strip image; According to the segmentation mask of the rubber strip area, the background in the rubber strip image is removed to generate an area image containing only the rubber strip; Analyzing the regional image to generate geometric feature information of the rubber strip; Compare the geometric feature information of the rubber strip with the preset quality standard to determine whether the quality of the glue coating meets the standard; The training operation of the segmentation model includes: Obtaining a rubber strip image and a corresponding label for model training; wherein the label is a segmentation mask of the rubber strip area; Constructing a first training set according to the rubber strip images and corresponding labels used for model training; Randomly dividing the first training set into a plurality of batches of training samples according to a preset number; Input each batch of training samples into the segmentation model in turn, and train the segmentation model until the preset number of training times is reached; wherein, when the prediction model receives each batch of training samples, it outputs the segmentation mask of the rubber strip area corresponding to the training sample; according to the segmentation mask of the rubber strip area and the corresponding label, the loss function value is calculated by the loss function; and the segmentation model is updated by the optimizer according to the loss function value.

2. The method for detecting glue quality based on deep learning according to claim 1, characterized in that: Before inputting the training samples of each batch into the segmentation model in sequence, the method further includes: For each batch of training samples, data enhancement processing and normalization processing are performed to generate updated training samples; wherein the data enhancement processing includes any one or a combination of random horizontal flipping, random vertical flipping, random rotation, random adjustment of image brightness, random addition of Gaussian noise and random cropping.

3. The method for detecting glue quality based on deep learning according to claim 1, characterized in that: Before obtaining the rubber strip images and corresponding labels for model training, it also includes: Get a general visual feature dataset; Extracting a second training set from the general visual feature dataset; The segmentation model is pre-trained according to the second training set.

4. The method for detecting glue coating quality based on deep learning according to claim 1, characterized in that: The step of analyzing the regional image to generate geometric feature information of the rubber strip includes: The BLOB analysis is performed on the regional image to extract the starting point, the ending point, the area, the length and the width of the rubber strip in the regional image to obtain geometric feature information.

5. The method for detecting glue coating quality based on deep learning according to claim 1, characterized in that: After judging whether the quality of the glue coating meets the standard, the method further includes: When the quality of glue coating does not meet the standards, an abnormal signal is sent to the control system.

6. A glue coating quality detection device based on deep learning, characterized in that: include: An image acquisition module to be detected, a model training module, an instance segmentation module, a region image generation module, a geometric feature information generation module and a glue coating quality detection module; The to-be-detected image acquisition module is used to acquire the to-be-detected rubber strip image; The model training module is used to obtain the rubber strip images and corresponding labels used for model training; wherein the labels are segmentation masks of the rubber strip areas; construct a first training set based on the rubber strip images and corresponding labels used for model training; randomly divide the first training set into a number of batches of training samples according to a preset number; input the training samples of each batch into the segmentation model in turn, and train the segmentation model until a preset number of training times is reached; wherein, when the prediction model receives a batch of training samples, it outputs the segmentation mask of the rubber strip area corresponding to the training sample; calculates the loss function value through the loss function based on the segmentation mask of the rubber strip area and the corresponding label; and updates the segmentation model based on the loss function value using the optimizer; The instance segmentation module is used to input the adhesive strip image into the segmentation model so that the segmentation model generates a segmentation mask of the adhesive strip area according to the adhesive strip image; The regional image generation module is used to remove the background in the rubber strip image according to the segmentation mask of the rubber strip area, and generate a regional image containing only the rubber strip; The geometric feature information generating module is used to analyze the regional image and generate geometric feature information of the rubber strip; The glue coating quality detection module is used to compare the geometric feature information of the glue strip with the preset quality standard to determine whether the quality of the glue coating meets the standard.

7. The deep learning-based gluing quality detection device according to claim 6, characterized in that: Also includes: Data preprocessing module; The data preprocessing module is used to perform data enhancement and normalization processing on each batch of training samples to generate updated training samples; wherein the data enhancement processing includes any one or a combination of random horizontal flipping, random vertical flipping, random rotation, random adjustment of image brightness, random addition of Gaussian noise and random cropping.

8. The deep learning-based gluing quality detection device according to claim 6, characterized in that: Also includes: Abnormal signal sending module; The abnormal signal sending module is used to send an abnormal signal to the control system when the quality of the glue coating does not meet the standard.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, it can implement the deep learning-based glue quality detection method described in any one of claims 1 to 5.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can implement the deep learning-based glue coating quality detection method described in any one of claims 1 to 5.

Images