Material misalignment detection method and device, electronic equipment and medium
By using deep learning models and image processing technology, the accuracy problem of material misalignment detection during tire molding was solved, achieving efficient and stable misalignment detection, and reducing production costs and raw material waste.
Patent Information
- Application Number
- CN202511716164.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-02-10
AI Technical Summary
Existing technologies are insufficient to accurately detect misalignment issues in materials during tire molding, leading to defective products flowing into subsequent processes, increasing raw material waste and production costs.
By combining deep learning models with image processing technology, images of material transport are acquired through industrial cameras to determine basic dimensions and material regions. The pre-trained deep learning model is then used to extract misalignment edge features to determine whether misalignment has occurred in the material.
It improves the accuracy and stability of misalignment detection, adapts to different lighting conditions and material deformation, reduces false alarm rate and false alarm rate, and improves production efficiency.
Smart Images
Figure CN121504894A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of tire manufacturing technology, and more specifically, to a method, apparatus, electronic device, and medium for detecting misalignment of materials. Background Technology
[0002] In tire forming, the materials used (such as rubber, cord fabric, and tread compound) are typically soft and easily deformable. During the winding process from the conveyor belt to the forming drum, due to mechanical vibration, uneven tension, material misalignment, or insufficient equipment precision, misalignment can easily occur at the material head (i.e., the starting end of the material), meaning the material fails to align precisely according to process requirements. This misalignment not only affects the structural uniformity of the tire but can also lead to quality problems such as uneven tire strength and uneven tread wear, ultimately resulting in defective or even scrap products.
[0003] Since misalignment typically occurs during the initial bonding stage of materials, and the error may be small (e.g., a deviation of a few millimeters), it is difficult to identify accurately and in real time through manual visual inspection. Especially on high-speed production lines, operators are prone to missing inspections due to fatigue or limited field of vision, leading to defective products flowing into subsequent processes. Once misalignment is not detected in time, subsequent vulcanization, molding, and other processes may further amplify the defect, ultimately requiring rework or scrapping. This not only increases raw material waste but also reduces production efficiency and raises overall production costs.
[0004] On tire production lines, the traditional way to detect misalignment is mainly through manual inspection or by using simple algorithms. This method is difficult to reliably identify minute misalignments, which leads to defective products flowing into subsequent processes and ultimately resulting in substandard or even scrap products.
[0005] Therefore, how to accurately detect whether materials have misaligned edges has become an urgent problem to be solved. Summary of the Invention
[0006] This disclosure provides a method, apparatus, electronic device, and medium for detecting misaligned edges of materials. It uses a deep learning model to detect misaligned edges, thereby improving the accuracy of material misalignment detection.
[0007] In a first aspect, embodiments of this disclosure provide a method for detecting misaligned edges of a material, applied to a tire forming machine. The flexible material to be detected is bonded to the belt drum / forming drum of the tire forming machine. The method includes: Acquire the target stitched image corresponding to the flexible material to be tested. The target stitched image is stitched together from the various material transport images captured sequentially by an industrial camera during the material transport process of the flexible material to be tested on the material transport machine. Based on the target stitched image, the basic dimensions of the flexible material to be detected are determined. The basic dimensions include the material head position, material tail position, overlap amount, misalignment area, off-center amount, and material width. Based on the basic dimensions, the location region of the flexible material to be detected in the target stitched image is determined, and based on the location region, an initial material region image containing the flexible material to be detected is extracted from the target stitched image. The initial material region image is further fitted to obtain a target material region image containing the flexible material to be detected; The basic dimensions and the target material region image are input into a pre-trained target deep learning model for feature extraction, resulting in misaligned edge features and misaligned edge morphology features. Based on the characteristics and morphology of misaligned edges, the amount of misalignment of the flexible material to be tested is determined, and the misalignment amount is used to determine whether the flexible material to be tested has a misalignment problem.
[0008] Secondly, embodiments of this disclosure provide a misalignment detection device for materials, applied to a tire forming machine. The flexible material to be detected is bonded to the belt drum / forming drum of the tire forming machine. The device includes: The acquisition module is used to acquire the target stitched image corresponding to the flexible material to be tested. The target stitched image is stitched together from the various material transport images captured sequentially by an industrial camera during the material transport process of the flexible material to be tested on the material transport machine. The first determining module is used to determine the basic dimensions of the flexible material to be detected based on the target stitched image. The basic dimensions include the material head position, material tail position, overlap amount, misalignment area, off-center amount, and material width. The second determining module is used to determine the position region of the flexible material to be detected in the target stitched image based on the basic dimensions, and to extract the initial material region image containing the flexible material to be detected from the target stitched image based on the position region. The first processing module is used to further fit the initial material region image to obtain a target material region image containing the flexible material to be detected. The second processing module is used to input the basic dimensions and the target material area image into the pre-trained target deep learning model for feature extraction, and to obtain the misaligned edge features and misaligned edge morphology features. The third determination module is used to determine the amount of misalignment of the flexible material to be tested based on the misalignment edge features and misalignment edge morphology features, and to determine whether the flexible material to be tested has a misalignment problem based on the amount of misalignment.
[0009] Thirdly, embodiments of this disclosure provide an electronic device including a processor and a memory interconnected thereto; the memory is used to store a computer program; the processor is configured to execute, when the computer program is invoked, the method provided by any possible implementation of the material misalignment detection method.
[0010] Fourthly, embodiments of this disclosure provide a computer-readable storage medium storing a computer program that is executed by a processor to implement the method provided in any possible implementation of the above-described method for detecting misaligned edges of materials.
[0011] Fifthly, embodiments of this disclosure provide a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in any possible implementation of the above-described method for detecting misaligned edges of materials.
[0012] The beneficial effects of the technical solutions provided in this disclosure are: In this embodiment of the disclosure, during the process of bonding the flexible material to be tested onto the belt drum / forming drum of the tire forming machine, it is necessary to detect in real time whether the flexible material to be tested has misaligned edges. The flexible material to be inspected is sequentially photographed by a pre-installed industrial camera on a material conveyor, resulting in individual material transport images. These images are then stitched together to obtain a target stitched image. Based on this target stitched image, the basic dimensions of the flexible material to be inspected are determined through image processing. Based on these basic dimensions, the location of the flexible material to be inspected within the target stitched image can be determined. Based on this location, an initial material region image containing the flexible material to be inspected, with the background area removed, is extracted from the target stitched image. This initial material region image is further fitted to obtain a target material region image containing the flexible material to be inspected. The determined basic dimensions and the target material region image are then input into a pre-trained target deep learning model to extract features from the target material region image, obtaining misalignment edge features and misalignment edge morphology features. Based on these features, the amount of misalignment of the flexible material to be inspected is determined, and finally, the degree of misalignment is used to determine whether the flexible material to be inspected has a misalignment problem. Through the embodiments of this disclosure, on the one hand, the basic dimensions of the flexible material to be detected and the initial material region image containing the flexible material to be detected are determined in advance through image processing. This preprocessing method provides a basis for subsequent determination of whether misalignment problems occur through the target deep learning model, thereby improving processing efficiency. On the other hand, feature extraction is performed through the trained target deep learning model to obtain misalignment edge features and misalignment edge morphology features. Combined with the basic dimension information, the accurate value of the amount of misalignment edge is calculated. Compared with traditional algorithms, the embodiments of this disclosure can adapt to different lighting conditions, material deformation and noise interference, and greatly improve the stability and accuracy of detection. Attached Figure Description
[0013] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments of this disclosure will be briefly introduced below.
[0014] Figure 1 A schematic flowchart of a method for detecting misaligned edges of materials provided in an embodiment of this disclosure; Figure 2 A schematic diagram of the structure of a misalignment detection device for materials provided in an embodiment of this disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation
[0015] The embodiments of this disclosure are described below with reference to the accompanying drawings. It should be understood that the embodiments described below with reference to the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of this disclosure, and do not constitute a limitation on the technical solutions of the embodiments of this disclosure.
[0016] To make the objectives, technical solutions, and advantages of this disclosure clearer, the embodiments of this disclosure will be described in further detail below with reference to the accompanying drawings.
[0017] The following description of several exemplary embodiments illustrates the technical solutions of this disclosure and the technical effects produced by these solutions. It should be noted that the following embodiments can be referenced, learned from, or combined with each other. Identical terms, similar features, and similar implementation steps in different embodiments will not be repeated.
[0018] First, the technical terms used in this disclosure will be introduced and explained: I. Tire Forming Machine: A tire forming machine is a specialized piece of equipment used to assemble various semi-finished components (such as tread, ply, and bead rings) into a tire blank (green tire). It is one of the core pieces of equipment in the tire manufacturing process. Its technical complexity and precision directly affect the quality, uniformity, and performance of the tire.
[0019] II. Material Conveyor Systems: These refer to automated conveying systems used in tire manufacturing to automatically transport semi-finished tire products (such as rubber compounds, cord fabric, and bead sheets) and the formed green tires and vulcanized finished tires between different processes. The types of conveyors used vary depending on the shape and characteristics of the tires at different stages. The main types include: 1. Application in raw materials / semi-finished components (solid): These materials are usually in rolls or blocks, and are relatively strong. They are mainly transported by belt conveyors and roller conveyors.
[0020] 2. Used for green tires (critical and fragile stage): Green tires are tires that have been formed but not yet vulcanized. They are very soft and easily deformed, requiring special support conveying equipment. They are mainly transported through servo pallet conveying systems, autonomous mobile robots (AMRs), automated guided vehicles (AGVs), and V-belt green tire conveyors.
[0021] 3. Applied to finished tires after vulcanization. After vulcanization, the tires become harder and have a fixed shape, making the conveying methods more diverse and simpler. They are mainly transported by plate chain conveyors, roller conveyors, spiral slides, and tire lifts / lowers.
[0022] III. Area Scan Camera: The core feature of an area scan camera is that its sensor is a two-dimensional matrix (a "plane") composed of pixels. When the shutter is activated, all (or most) pixels on the sensor simultaneously or in a specific order receive light, thereby acquiring a complete two-dimensional image of the observed object in one go. The industrial camera involved in this disclosure is an area scan camera. Through this area scan camera, flexible materials to be inspected that are being transported on the surface of a material conveyor can be photographed to obtain various material transport images.
[0023] IV. Belt Drum: Also known as the belt layer bonding drum, it is the core component of the second stage forming machine in the two-stage tire building machine. It is a precision-engineered metal drum capable of high-speed rotation and complex radial expansion and contraction. Its main task is to combine the pre-cut steel belts and tread into a single unit with extremely high precision, and then transfer and fit it onto the first stage of the formed tire carcass.
[0024] 5. Forming Drum: This is the core component of the tire forming machine, which can be understood as a foldable and expandable "mold skeleton". Its function is to provide a precise fitting platform for all semi-finished tire components (such as carcass ply, belt layer, tread, etc.), and through its complex movement, wrap and shape the flat material into a cylindrical, uncured green tire.
[0025] The misalignment detection method for materials in this embodiment can be executed by a server or a terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server or server cluster providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The aforementioned networks can include, but are not limited to, wired networks and wireless networks. Wired networks include local area networks (LANs), metropolitan area networks (MANs), and wide area networks (WANs). Wireless networks include Bluetooth, Wi-Fi, and other networks that enable wireless communication. Terminal devices can be smartphones (such as Android phones, iOS phones, etc.), tablets, laptops, digital broadcast receivers, MIDs (Mobile Internet Devices), PDAs (Personal Digital Assistants), desktop computers, in-vehicle terminals (such as in-vehicle navigation terminals), smart speakers, smartwatches, etc. Terminal devices and servers can be directly or indirectly connected via wired or wireless communication, but are not limited to these methods. Specific details can be determined based on actual application scenario requirements and are not limited here.
[0026] In tire manufacturing, the rubber composite materials used have significant flexibility, making them highly susceptible to deformation during transfer from the conveyor belt (material transporter) to the forming drum / belt winding drum. Misalignment can easily occur during the winding process from the conveyor belt to the forming drum. This misalignment not only affects the structural uniformity of the tire but can also lead to uneven tire strength, uneven tread wear, and other quality problems, ultimately resulting in defective or even scrap products.
[0027] Currently, two methods are commonly used to detect misaligned edges.
[0028] Method 1: Check for misalignment issues by manual visual inspection. This is especially important on high-speed production lines, where operators may miss issues due to fatigue or limited field of vision, leading to defective products flowing into subsequent processes.
[0029] Method 2: Some tire manufacturers use misalignment detection systems based on traditional image processing algorithms (such as edge detection and template matching), but these methods have obvious limitations: 1. High false negative rate: Traditional algorithms are sensitive to changes in lighting and interference from material textures, making it difficult to reliably identify minute edge errors in complex industrial environments, resulting in some defects going undetected.
[0030] 2. High false alarm rate: Due to the softness and surface irregularity of tire materials, traditional algorithms are prone to misjudging normal misalignment, shadows or reflections as misalignment, frequently triggering false alarms and affecting production efficiency.
[0031] 3. Poor adaptability: Different specifications of tire materials, such as rubber materials of different widths and colors, require frequent adjustment of algorithm parameters, making it difficult to achieve intelligent adaptive detection and increasing maintenance costs.
[0032] In summary, the current technical solution has obvious limitations, and the system cannot accurately detect misaligned edges.
[0033] To address the aforementioned problems, this disclosure provides a method for detecting misaligned edges of materials, applied to a tire forming machine. The flexible material to be tested is bonded to the belt drum / forming drum of the tire forming machine, such as... Figure 1 As shown, the method includes the following steps: Step S101: Obtain the target stitched image corresponding to the flexible material to be detected. The target stitched image is stitched together from the various material transport images captured sequentially by an industrial camera during the process of the flexible material to be detected being transported on the material transport machine. Step S102: Based on the target stitched image, determine the basic dimensions of the flexible material to be detected. The basic dimensions include the material head position, material tail position, overlap amount, misalignment area, off-center amount, and material width. Step S103: Based on the basic dimensions, determine the location region of the flexible material to be detected in the target stitched image, and based on the location region, extract the initial material region image containing the flexible material to be detected from the target stitched image. Step S104: Further fit the initial material region image to obtain a target material region image containing the flexible material to be detected. Step S105: Input the basic dimensions and the target material area image into the pre-trained target deep learning model for feature extraction to obtain misaligned edge features and misaligned edge morphological features; Step S106: Based on the misalignment edge features and misalignment edge morphology features, determine the misalignment amount of the flexible material to be tested, and determine whether the flexible material to be tested has a misalignment problem based on the misalignment amount.
[0034] Optionally, the embodiments of this disclosure can be applied to tire manufacturing scenarios. The flexible material to be tested can be understood as a material with a certain degree of flexibility and deformability, such as rubber materials, composite materials (composed of multiple flexible materials), elastomer materials, etc. For example, the flexible material to be tested includes a long strip of rubber with triangular ends used in tire production. Specifically, the flexible material to be tested may include a belt layer rubber sheet used in tire production. The belt layer rubber sheet is one of the components of the tire. It is a rubber coating on brass-plated steel cord, calendered or extruded onto the steel cord in the form of a thin sheet, cut at a certain angle and width, and then overlapped to form a continuous roll for tire forming. The belt layer formulation mainly targets fatigue resistance and tear resistance. It is understood that the embodiments of this disclosure do not limit the flexible material to be tested.
[0035] To detect whether misalignment occurs in the flexible material under test during transport, a detection system provided in this embodiment can be used. This system comprises a material conveyor, a tire forming machine, an industrial camera, and a processing device. The material conveyor transports the flexible material under test, the tire forming machine adheres the material to the surface, and the processing device processes the stitched image of the material transport images captured by the industrial camera. Based on this stitched image, and using a target deep learning model, it detects whether misalignment occurs in the flexible material under test.
[0036] First, it is necessary to acquire a target stitched image corresponding to the flexible material to be tested. In one optional embodiment, acquiring the target stitched image of the flexible material to be tested includes: receiving a start transmission signal forwarded by the tire forming machine controller, wherein, upon receiving the start transmission signal sent by the tire forming machine controller, the material conveyor controller controls the material conveyor to start transmitting the flexible material to be tested; sending a start transmission signal to an industrial camera, wherein, upon receiving the start transmission signal, the industrial camera acquires material transmission images at a preset image acquisition frequency within a preset material transmission time period; acquiring each material transmission image captured sequentially by the industrial camera; and stitching the various material transmission images together to obtain the target stitched image.
[0037] Optionally, the start and stop of the conveying of the flexible material to be tested are determined by signals sent by the tire forming machine controller. When the flexible material to be tested needs to be conveyed, the tire forming machine controller sends a start conveying signal to the material conveyor controller. Upon receiving the start conveying signal, the material conveyor controller controls the material conveyor to perform the material conveying operation. Both the tire forming machine controller and the material conveyor controller are programmable logic controllers (PLCs).
[0038] When the tire forming machine controller sends a start transmission signal to the material conveyor controller, it simultaneously sends the same start transmission signal to the processing equipment. Upon receiving the start transmission signal, the processing equipment forwards it to the industrial camera. Once the industrial camera receives the start transmission signal, it will capture images of the flexible material to be inspected within a preset material transmission time period, according to a preset image acquisition frequency, obtaining various material transmission images, and then transmit these images to the processing equipment.
[0039] The material transfer time period is the time required to transfer one piece of material, which is the time required to transfer the head and tail of the flexible material to be tested.
[0040] Then, the processing equipment receives the material transport images transmitted by the industrial camera and stitches them together sequentially according to the image acquisition time of each material transport image to obtain the target stitched image.
[0041] Image processing is performed on the target stitched image to determine its edges. These edges include the material edges. Based on the material edges and the definition of basic dimensions, the basic dimensions of the flexible material to be inspected can be determined. These basic dimensions include the head position, tail position, overlap, misalignment area, off-center amount, and material width. The overlap, also known as the joint amount, refers to the length of the overlapping portion between the two rolls when they are joined. The material width refers to the actual width of the material. The off-center amount refers to the offset distance between the centerline of the entire roll of material and the theoretical centerline of the production line or equipment during transmission. The head position is the starting end of the material, and its position in the image is the head position. The tail position is the ending end of the material, and its position in the image is the tail position.
[0042] Then, based on the determined basic dimensions, the location region of the flexible material to be detected in the target stitched image can be determined. Based on this location region, image cropping can be performed to extract the image containing only the region of the flexible material to be detected, thus obtaining the initial material region image and removing the interference region that does not contain the flexible material to be detected.
[0043] By further fitting the initial material region image, a target material region image containing the flexible material to be detected can be obtained.
[0044] Finally, the target material region image and basic dimensional information are input into a pre-trained target deep learning model. This model is specifically trained for detecting misaligned edges in material images from the tire manufacturing process, and has the ability to accurately detect whether misaligned edges exist in the material region image. It can combine basic dimensional information to extract features from the target material region image, obtaining the misaligned edge features and misaligned edge morphology features of the flexible material to be detected. Based on these features and morphology, and according to the definition of misalignment amount, the misalignment amount of the flexible material to be detected can be determined. Based on the misalignment amount, it can be determined whether the flexible material to be detected has a misaligned edge problem.
[0045] Through the embodiments of this disclosure, on the one hand, the basic dimensions of the flexible material to be detected and the image of the material region containing the flexible material to be detected are determined in advance through image processing. This preprocessing method provides a basis for subsequent determination of whether misaligned edges occur through the target deep learning model, thereby improving processing efficiency. On the other hand, the misaligned edge detection is performed by the trained target deep learning model, which combines information from both the basic dimensions and the material region image, greatly improving the accuracy of material misaligned edge detection.
[0046] The following details how to determine the basic dimensions of the flexible material to be detected. In one optional embodiment, the basic dimensions of the flexible material to be detected are determined based on the target stitched image, including: performing image sharpening, image denoising, edge extraction, and morphological filtering operations on the target stitched image to obtain the image edges of the target stitched image; filling the discontinuities of the image edges of the target stitched image through morphological closing operations, and determining the head position of the flexible material to be detected based on Hough transform line detection; determining the tail position of the flexible material to be detected based on the head position and standard material length information; determining the overlap amount of the flexible material to be detected based on the head position and tail position; using adaptive threshold segmentation and sub-pixel edge detection, determining the left and right width edges, as well as the top and bottom edges of the flexible material to be detected in the target stitched image; determining the material width based on the left and right width edges; determining the off-center amount based on the top and bottom edges and the centerline of the material conveyor; establishing a two-dimensional coordinate system mapping model, determining the geometric relationship of the edges in the target stitched image through the two-dimensional coordinate system mapping model, and determining the misalignment area by combining the region growing method.
[0047] Optionally, image edges of the target stitched image can be obtained by performing operations such as image sharpening, image denoising, edge extraction, and morphological filtering on the target stitched image. Then, discontinuities in the image edges of the target stitched image are filled using morphological closing operations, and the position of the head of the flexible material to be detected is determined based on line detection using Hough Transform. Closing is one of the fundamental operations in mathematical morphology, defined as a combination of dilation and erosion operations on an image. This operation belongs to the basic theoretical system of mathematical morphology image processing and can fill small holes inside objects, close narrow gaps, and eliminate discrete noise while maintaining the basic geometric features of the target object. Hough Transform is a method for detecting specific shapes in an image, most commonly for detecting lines. The core idea of Hough Transform is to map edge points in image space to points in parameter space, and then find the concentrated regions of these points in parameter space to detect lines or other shapes in the image.
[0048] Then, based on the material head position and the standard material length corresponding to the standard material length information, the material head position is used as a reference and the standard material length is used to determine a limited range. Within this limited range, the material tail position of the target splicing image can be determined.
[0049] The overlap of the flexible material to be tested can be determined based on the positions of the head and tail of the material and the definition of overlap.
[0050] By employing adaptive threshold segmentation and sub-pixel edge detection, the left and right width edges, as well as the top and right edges, of the flexible material to be detected in the target stitched image can be determined. Based on the distance between the left and right width edges, the width of the flexible material to be detected can be determined. Based on the top and bottom edges, the centerline of the flexible material to be detected in the target stitched image can be determined. The distance between the centerline of the flexible material to be detected and the centerline of the material conveyor is the off-center amount of the flexible material to be detected.
[0051] Adaptive thresholding is a method that segments images by dynamically calculating thresholds for local regions, suitable for scenes with uneven lighting or complex color differences. Its core principle is to divide the image into small regions and adaptively adjust the threshold based on the local brightness distribution, thereby improving segmentation accuracy. Subpixel edge detection's core idea is to use the grayscale information (or gradient information) of pixels near the edge to fit a continuous function model, and then calculate the precise edge location by finding the extrema, zeros, or inflection points of this function.
[0052] A two-dimensional coordinate system mapping model is established. This model determines the geometric relationships of edges in the target stitched image. Then, combined with the region growing method, the misalignment region of the flexible material to be inspected is determined. The misalignment region is used to determine the amount of misalignment. The amount of misalignment refers to the offset distance caused when the new and old materials are not completely aligned in the width direction during the jointing of materials. The misalignment region is the area where misalignment occurs during the jointing of materials.
[0053] In essence, two-dimensional coordinate system mapping is a function or transformation. It maps points in one two-dimensional coordinate system (source coordinate system) to another two-dimensional coordinate system (target coordinate system) according to certain rules. Region growing is an image segmentation algorithm based on similarity criteria. It forms regions by progressively merging adjacent pixels starting from a seed point. The basic principle is: starting from a selected seed point (a single pixel or a small region), adjacent pixels are merged into the current region according to preset similarity criteria (such as grayscale, color, texture, etc.). The newly merged pixels can be used as new seed points to continue growing until no pixels meet the criteria.
[0054] The above methods can be used to determine the basic dimensions of the flexible material to be tested, namely the head position, tail position, overlap, misalignment area, off-center amount, and material width.
[0055] By pre-determining key dimensions such as the discharge head position, discharge tail position, material width, overlap amount, misalignment area, and off-center amount through the embodiments of this disclosure, the position and boundary range of the material in the image can be accurately determined, ensuring accurate spatial positioning of the material during the transmission process. This lays the foundation for feature extraction of the subsequent deep learning model, provides reference information for the subsequent target deep learning model, reduces the computational load of the target deep learning model, and thus improves the detection efficiency of the target deep learning model.
[0056] In an optional embodiment, after determining whether the flexible material to be tested has a misalignment problem based on the misalignment amount, the method further includes: if it is determined that a misalignment problem has occurred, sending a control signal to the tire forming machine controller, so that the tire forming machine controller controls the tire forming machine to perform stepping or stopping operations according to the control signal.
[0057] Optionally, if the target deep learning model determines that the flexible material to be tested does not have a misalignment problem, the material conveyor and the tire forming machine will both maintain their original working models and continue to operate.
[0058] If the target deep learning model determines that the flexible material to be inspected has a misalignment problem, a control signal needs to be sent to the tire forming machine controller. Upon receiving this control signal, the tire forming machine controller can, according to the signal's indication, control the tire forming machine to continue stepping or execute a shutdown operation, sending appropriate control signals as needed. At this time, the staff will receive a corresponding warning and inspect the equipment.
[0059] In addition, the industrial camera is mounted in front of the flexible material to be inspected and can extend and retract. When a misalignment problem occurs, the industrial camera receives a stop shooting signal from the processing equipment. Upon receiving this stop shooting signal, the industrial camera will stop shooting and retract via the telescopic structure, providing working space for the operator.
[0060] Through the embodiments disclosed herein, when a misalignment problem occurs, the tire forming machine can be promptly controlled to step forward or stop, thus avoiding greater losses and reducing costs.
[0061] In one optional embodiment, the target deep learning model is trained on the initial deep learning model in the following manner: A target image dataset is obtained, wherein each image in the target image dataset is labeled to indicate whether a misaligned edge problem exists; the target image dataset includes a training set; the input layer, convolutional layer, pooling layer, fully connected layer, activation function, loss function, and optimizer of the initial deep learning model are set to obtain a deep learning model with the set model architecture; based on the training set and the loss function, the deep learning model with the set model architecture is trained, and the weights are adjusted using the backpropagation algorithm until the loss function converges; the deep learning model with the optimized model architecture at the point of convergence is taken as the target deep learning model.
[0062] Optionally, the target deep learning model is obtained by training the initial deep learning model. The specific training process is as follows: Step 1: Data collection and preprocessing.
[0063] In one optional embodiment, obtaining the target image dataset includes: collecting a set of tire misalignment images with misalignment issues; labeling each image in the tire misalignment image set to indicate whether a misalignment issue occurs; performing image resizing, pixel value normalization, rotation, scaling, and flipping operations on the tire misalignment image set to obtain the target image dataset; and dividing the target image dataset into a training set, a validation set, and a test set according to different functions.
[0064] Optionally, the training dataset should be as comprehensive as possible and consist of images of flexible materials in the tire manufacturing field. It should include images of misaligned edges of various types. A database of millions of images should be built according to this standard, containing hundreds of typical misaligned edge samples of different specifications. Data augmentation schemes should be used: random rotation (±15°), brightness jitter (±30%), Gaussian noise injection (σ=0.05), and elastic deformation simulation to obtain a set of tire misaligned edge images. The images should be labeled to indicate whether misalignment occurs in each image.
[0065] Preprocessing the images in the tire misalignment image set can increase the diversity of the data and reduce overfitting by adjusting the image size, normalizing pixel values, rotating, scaling, and flipping. After these preprocessing operations, the target image dataset can be obtained.
[0066] Then, according to different functions, the target image dataset is divided into a training set, a validation set, and a test set. For example, the proportion of the training set can be 70%, the proportion of the validation set can be 15%, and the proportion of the test set can be 15%.
[0067] Step 2: Set the model architecture of the target deep learning model.
[0068] Optionally, the backbone network is an improved ResNet-50 (i.e., a 50-layer residual network), the feature extraction layer is a 3×3 deformable convolution, and the classification head is a 2-layer MLP + Dropout (0.5). The Multilayer Perceptron (MLP) is a classic feedforward artificial neural network, and Dropout is a powerful regularization technique designed to prevent overfitting in neural networks, especially complex networks like large MLPs. A ratio of 0.5 means that each neuron has a 50% chance of being turned off during training on any given sample.
[0069] Step 3: Train the model using the training set.
[0070] Using the training set and loss function, a deep learning model with the configured model architecture is trained, and the weights are adjusted using the backpropagation algorithm to minimize the loss function. The loss value and accuracy are monitored during training, and hyperparameters such as learning rate and batch size are adjusted. When the loss function converges, the deep learning model with the converged model architecture is used as the trained target deep learning model.
[0071] Through the embodiments disclosed herein, training can be performed specifically on a set of tire misalignment images based on the field of tire manufacturing technology, which greatly improves the recognition rate of misalignment.
[0072] In one optional embodiment, obtaining the target image dataset includes: collecting a set of tire misalignment images with misalignment issues; labeling each image in the tire misalignment image set to indicate whether a misalignment issue occurs; performing image resizing, pixel value normalization, rotation, scaling, and flipping operations on the tire misalignment image set to obtain the target image dataset; and dividing the target image dataset into a training set, a validation set, and a test set according to different functions.
[0073] Optionally, the model's performance can be evaluated on a validation set to avoid overfitting. The model is evaluated using metrics such as accuracy and loss rate on the validation set. If the validation results are unsatisfactory, the model is retrained; if they are satisfactory, the final performance is evaluated on the test set to ensure good generalization ability. Metrics such as accuracy and confusion matrix on the test set are recorded to evaluate the model's performance.
[0074] Through the embodiments of this disclosure, the model can be evaluated using a validation set and a test set, thus ensuring the model's performance.
[0075] The following describes in detail the misalignment detection method for materials in this disclosure with reference to an embodiment.
[0076] This disclosure employs advanced deep learning technology to perform high-precision feature extraction and fusion analysis on the material surface during the tire molding process, enabling intelligent calculation of misalignment values (i.e., misalignment amount). The implementation principle is as follows: 1. Material Position and Range Detection: After the detection system completes the detection of key dimensions such as the material head, tail, and width, it can accurately determine the position and boundary range of the material in the image. This ensures accurate spatial positioning of the material during the conveying process, laying the foundation for subsequent feature extraction.
[0077] 2. Multi-feature extraction and fusion calculation: Within the detected material range, the system further fits a high-precision detection area (i.e., the target material area image) and uses a deep learning model to extract a variety of key features (i.e., misaligned edge features and misaligned edge morphology features).
[0078] 3. After these features are fused and calculated, combined with other detected data such as overlap amount, material width, off-center amount, and misalignment area, the precise value of the misalignment is calculated, i.e., the misalignment amount. Based on the misalignment amount, it can be determined whether the material has a misalignment problem.
[0079] Compared with traditional algorithms, the method in this embodiment can adapt to different lighting conditions, material deformation and noise interference, and greatly improve the stability and accuracy of detection.
[0080] The following details the specific steps of the material misalignment detection method in the embodiments of this disclosure.
[0081] The detection system of this disclosure mainly consists of three parts: a controller (i.e., processing equipment), an industrial camera, and a precision mechanical motion mechanism, forming a complete intelligent detection system. The system adopts a modular design, and the components work together to achieve high-precision detection.
[0082] Among them, the precision mechanical motion mechanism adopts a high-precision linear module driven by a servo motor, including a tire forming machine and a material conveyor.
[0083] The controller, or processing device, adopts an industrial-grade dual-processor architecture, combining real-time control and high-performance computing capabilities, and connects the tire forming machine controller with an industrial camera.
[0084] The industrial camera employs a high-resolution industrial-grade area array camera, equipped with a professional optical lens and laser to ensure image quality. The camera is mounted within a precision mechanical motion mechanism, positioned in front of the object being measured. This industrial camera can scan multiple images of the object to be inspected. These images are then stitched together to obtain the target stitched image. The molding machine controller sends a start transmission signal, responding to acquisition and calculation operations. Based on the target stitched image, the controller calculates the corresponding results and feeds them back to the molding machine controller, which then controls the molding machine to perform stepping or stopping operations.
[0085] The workflow is as follows: Industrial cameras are precisely positioned in front of the object being measured through a mechanical motion mechanism. The field of view, exposure, aperture, and ISO can be automatically adjusted according to image quality.
[0086] After the molding machine controller sends a trigger signal, the system begins image acquisition. The image processing algorithm completes the calculation and analysis within 50ms, and the detection results are fed back to the molding machine controller in real time via industrial Ethernet.
[0087] Misalignment detection algorithm flow: Phase 1: Locating the misalignment detection area: 1. For material head area detection, multi-scale Gaussian filtering is used for image denoising. The Canny operator combined with the improved Sobel edge detection algorithm is used to extract edges. The edge discontinuities are filled by morphological closing operation. The material head position is determined by line detection based on Hough transform. Combined with the standard material length information in the process parameter database, the joint area is accurately calculated.
[0088] 2. Width edge detection: Adaptive threshold segmentation algorithm is applied, and sub-pixel edge detection technology is used with an accuracy of 0.1 pixels. The optimal edge line is fitted by Random Sample Consensus (RANSAC) algorithm.
[0089] 3. Misalignment region determination: A two-dimensional coordinate system mapping model is established, and potential misalignment regions are calculated through geometric relationships. The region growing method is used to determine the ROI (Region of Interest) of the image to be detected, i.e., the misalignment region.
[0090] Phase Two: Accurately extracting the misaligned edges and calculating the misalignment amount: 1. Traditional algorithm preprocessing: Apply local contrast enhancement within the ROI, extract features using Histogram of Oriented Gradient (HOG), and preliminarily screen candidate edges based on K-means clustering algorithm.
[0091] 2. Deep Learning-Based Fine Screening: A multi-scale Feature Pyramid Network (FPN) is constructed, and an attention mechanism is applied to enhance key features. The output is optimized using Non-Maximum Suppression (NMS) algorithm. Implementation details of the deep learning model include building a database of millions of images containing hundreds of typical misaligned edge samples of different sizes, and employing data augmentation schemes: random rotation (±15°), brightness jitter (±30%), Gaussian noise injection (σ=0.05), and elastic deformation simulation.
[0092] The model architecture consists of a modified ResNet-50 backbone network, a 3×3 deformable convolutional layer for feature extraction, a 2-layer MLP with Dropout (0.5) for classification, and Focal Loss (γ=2, α=0.25) as the loss function.
[0093] Training optimization: initial learning rate: 0.001, cosine annealing decay batch size: 32, and early stopping mechanism: validation set loss does not decrease for 3 consecutive rounds, regularization: L2 weight decay (λ=0.0001).
[0094] Model deployment uses TensorRT (TensorRuntime) to accelerate inference.
[0095] The embodiments disclosed herein have the following technical effects: On the tire production line, by detecting misaligned edges of materials that exceed the process standards, manufacturers can promptly identify potential problems and prevent scrapped tires from flowing into the next process, which helps to improve the overall quality of tires.
[0096] Well-trained deep learning models have the following advantages: 1. Detection accuracy: The accuracy rate of misaligned edge recognition is ≥99.7%.
[0097] 2. Stability: Under continuous operation 24 / 7, the mean time between failures (MTBF) is >5000 hours.
[0098] 3. Adaptability: Supports automatic switching of 100+ tire specifications.
[0099] This disclosure embodiment achieves an optimized upgrade of misalignment detection during tire forming through an innovative "traditional algorithm + deep learning" fusion architecture, providing reliable quality assurance for intelligent manufacturing.
[0100] The beneficial effects of the embodiments disclosed herein are as follows: 1. Significantly improves quality control level. On the tire production line, high-precision inspection can identify minute misalignments that are difficult to identify reliably by traditional manual inspection or simple algorithms, thereby improving the product qualification rate.
[0101] 2. Significantly improves production efficiency. Traditional manual inspection relies on operator experience and is difficult to adapt to the pace of high-speed production lines, easily leading to missed inspections or misjudgments. The automatic inspection system in this embodiment can operate continuously 24 / 7, maintaining stable inspection capabilities even in high-speed production environments, reducing the need for manual intervention. Simultaneously, the system supports real-time feedback and can be linked with automated equipment to achieve rapid correction, further optimizing production cycle time, reducing downtime caused by misalignment issues, and improving overall production efficiency.
[0102] 3. Data-driven process optimization and equipment maintenance: This disclosure not only provides real-time detection functions, but also records historical data of misalignment detection, including the frequency and trend analysis of misalignment occurrence, to help identify anomalies in specific time periods or equipment states.
[0103] 4. Analyze the correlation between raw material quality and trace the edge misalignment rate of different batches of materials to optimize supplier selection.
[0104] 5. Monitor the operating status of the molding equipment. If misalignment issues are found to occur frequently, it indicates mechanical wear or abnormal tension control.
[0105] This data can be further used for process optimization, such as adjusting material tension, optimizing bonding parameters, or improving equipment calibration strategies, thereby achieving long-term quality improvement and cost control. In addition, the data logging function supports production traceability, facilitating quality review and accountability, meeting the management needs of modern intelligent manufacturing.
[0106] This disclosure provides a misalignment detection device for materials, applied to a tire forming machine. The flexible material to be detected is bonded to the belt drum / forming drum of the tire forming machine, such as... Figure 2As shown, the device 20 may include: an acquisition module 201, a first determination module 202, a second determination module 203, a first processing module 204, a second processing module 205, and a third determination module 206, wherein: The acquisition module 201 is used to acquire the target stitched image corresponding to the flexible material to be detected. The target stitched image is stitched together from the various material transport images captured sequentially by an industrial camera during the process of the flexible material to be detected being transported on the material transport machine. The first determining module 202 is used to determine the basic dimensions of the flexible material to be detected based on the target stitched image. The basic dimensions include the material head position, material tail position, overlap amount, misalignment area, off-center amount, and material width. The second determining module 203 is used to determine the position region of the flexible material to be detected in the target stitched image based on the basic dimensions, and to extract the initial material region image containing the flexible material to be detected from the target stitched image based on the position region. The first processing module 204 is used to further fit the initial material region image to obtain a target material region image containing the flexible material to be detected. The second processing module 205 is used to input the basic dimensions and the target material area image into a pre-trained target deep learning model for feature extraction, and to obtain misaligned edge features and misaligned edge morphological features. The third determining module 206 is used to determine the amount of misalignment of the flexible material to be tested based on the misalignment edge features and misalignment edge morphology features, and to determine whether the flexible material to be tested has a misalignment problem based on the amount of misalignment.
[0107] In an optional embodiment, the device further includes a control module configured to: if a misalignment problem is determined to occur, send a control signal to the tire forming machine controller, so that the tire forming machine controller controls the tire forming machine to perform stepping or stopping operations according to the control signal.
[0108] In an optional embodiment, the device further includes a model training module for: training an initial deep learning model to obtain a target deep learning model by: acquiring a target image dataset, wherein each image in the target image dataset is labeled to indicate whether a misaligned edge problem occurs, and the target image dataset includes a training set; setting the input layer, convolutional layer, pooling layer, fully connected layer, activation function, loss function, and optimizer of the initial deep learning model to obtain a deep learning model with the set model architecture; training the deep learning model with the set model architecture based on the training set and the loss function, and adjusting the weights through a backpropagation algorithm until the loss function converges, and using the deep learning model with the optimized model architecture at the time of convergence as the target deep learning model.
[0109] In one optional embodiment, the target image dataset further includes a validation set and a test set. The model training module is used to: validate the target deep learning model through the validation set to obtain the validation evaluation index corresponding to the validation set; determine whether the target deep learning model is qualified based on the validation evaluation index; if it is unqualified, retrain it; if it is qualified, test the target deep learning model through the test set to obtain the test evaluation index corresponding to the test set and record it; and determine the performance of the target deep learning model through the test evaluation index.
[0110] In one optional embodiment, the model training module is used to: collect a set of tire misalignment images with misalignment issues; label each image in the tire misalignment image set to indicate whether a misalignment issue occurs; perform image resizing, pixel value normalization, rotation, scaling, and flipping operations on the tire misalignment image set to obtain a target image dataset; and divide the target image dataset into a training set, a validation set, and a test set according to different functions.
[0111] In one optional embodiment, the first determining module 202 is specifically used for: performing image sharpening, image denoising, edge extraction, and morphological filtering operations on the target stitched image to obtain the image edges of the target stitched image; filling the discontinuities of the image edges of the target stitched image through morphological closing operations, and determining the head position of the flexible material to be detected based on Hough transform line detection; determining the tail position of the flexible material to be detected based on the head position and standard material length information; determining the overlap amount of the flexible material to be detected based on the head position and tail position; determining the left and right width edges, as well as the top and bottom edges of the flexible material to be detected in the target stitched image using an adaptive threshold segmentation method and sub-pixel edge detection; determining the material width based on the left and right width edges; determining the off-center amount based on the top and bottom edges and the centerline of the material conveyor; establishing a two-dimensional coordinate system mapping model, determining the geometric relationship of the edges in the target stitched image through the two-dimensional coordinate system mapping model, and determining the misalignment area by combining the region growing method.
[0112] In an optional embodiment, the acquisition module 201 is specifically configured to: receive a start transmission signal forwarded by the tire forming machine controller, wherein, upon receiving the start transmission signal sent by the tire forming machine controller, the material conveyor controller controls the material conveyor to start transmitting the flexible material to be tested; send a start transmission signal to the industrial camera, wherein, upon receiving the start transmission signal, the industrial camera acquires material transmission images at a preset image acquisition frequency within a preset material transmission time period; acquire each material transmission image captured sequentially by the industrial camera; and stitch the various material transmission images together to obtain a target stitched image.
[0113] Through the embodiments of this disclosure, on the one hand, the basic dimensions of the flexible material to be detected and the initial material region image containing the flexible material to be detected are determined in advance through image processing. This preprocessing method provides a basis for subsequent determination of whether misalignment problems occur through the target deep learning model, thereby improving processing efficiency. On the other hand, feature extraction is performed through the trained target deep learning model to obtain misalignment edge features and misalignment edge morphology features. Combined with the basic dimension information, the accurate value of the amount of misalignment edge is calculated. Compared with traditional algorithms, the embodiments of this disclosure can adapt to different lighting conditions, material deformation and noise interference, and greatly improve the stability and accuracy of detection.
[0114] The apparatus of this disclosure embodiment can execute the method provided in this disclosure embodiment, and its implementation principle is similar, and it has corresponding technical effects. The actions performed by each module in the apparatus of each embodiment of this disclosure correspond to the steps in the method of each embodiment of this disclosure. For a detailed functional description of each module of the apparatus, please refer to the description in the corresponding method shown above, and it will not be repeated here.
[0115] This disclosure provides an electronic device (computer apparatus / device / system) including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method provided in any optional embodiment of this disclosure.
[0116] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of this disclosure.
[0117] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with this disclosure. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.
[0118] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0119] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.
[0120] The memory 4003 is used to store computer programs that execute embodiments of the present disclosure, and is controlled by the processor 4001 to execute them. The processor 4001 is used to execute the computer programs stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.
[0121] This disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.
[0122] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.
[0123] It should be understood that although arrows indicate various operation steps in the flowcharts of the embodiments of this disclosure, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of the embodiments of this disclosure, the implementation steps in each flowchart can be executed in other orders as required. Furthermore, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage can also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured as required, and the embodiments of this disclosure do not limit this.
[0124] The above description is only an optional implementation method for some implementation scenarios of this disclosure. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this disclosure without departing from the technical concept of this disclosure also fall within the protection scope of the embodiments of this disclosure.
Claims
1. A method for detecting misaligned edges of materials, applied to a tire forming machine, wherein the flexible material to be tested is bonded to the belt drum / forming drum of the tire forming machine, characterized in that... The method includes: Obtain a target stitched image corresponding to the flexible material to be tested, wherein the target stitched image is stitched together from various material transport images captured sequentially by an industrial camera during the material transport process of the flexible material to be tested on the material transport machine; Based on the target stitched image, the basic dimensions of the flexible material to be detected are determined, wherein the basic dimensions include the material head position, the material tail position, the overlap amount, the misalignment area, the off-center amount, and the material width. Based on the aforementioned basic dimensions, the location region of the flexible material to be detected in the target stitched image is determined, and based on the location region, an initial material region image containing the flexible material to be detected is extracted from the target stitched image. The initial material region image is further fitted to obtain a target material region image containing the flexible material to be detected; The basic dimensions and the target material region image are input into a pre-trained target deep learning model for feature extraction to obtain misaligned edge features and misaligned edge morphological features. Based on the misaligned edge features and the misaligned edge morphology features, the amount of misalignment of the flexible material to be tested is determined, and based on the amount of misalignment, it is determined whether the flexible material to be tested has a misaligned edge problem.
2. The method according to claim 1, characterized in that, After determining whether the flexible material to be tested has a misalignment problem based on the misalignment amount, the method further includes: If a misalignment problem is identified, a control signal is sent to the tire forming machine controller, so that the tire forming machine controller can control the tire forming machine to perform stepping or stopping operations according to the control signal.
3. The method according to claim 1, characterized in that, The target deep learning model is obtained by training the initial deep learning model in the following way: Obtain a target image dataset, wherein each image in the target image dataset is labeled to indicate whether a misaligned edge problem occurs, and the target image dataset includes a training set; The input layer, convolutional layer, pooling layer, fully connected layer, activation function, loss function, and optimizer of the initial deep learning model are set to obtain the deep learning model with the set model architecture. Based on the training set and the loss function, a deep learning model with the configured model architecture is trained, and the weights are adjusted through the backpropagation algorithm until the loss function converges. The deep learning model with the optimized model architecture at the time of convergence is taken as the target deep learning model.
4. The method according to claim 3, characterized in that, The target image dataset also includes a validation set and a test set, and the method further includes: The target deep learning model is validated using the validation set, and the validation evaluation index corresponding to the validation set is obtained. The validity of the target deep learning model is determined based on the verification and evaluation metrics. If the test is unsuccessful, the training must be repeated. If it passes, the target deep learning model is tested using the test set, and the test evaluation index corresponding to the test set is obtained and recorded. The performance of the target deep learning model is determined by the test evaluation metrics.
5. The method according to claim 3, characterized in that, The acquisition of the target image dataset includes: Collect a set of images of tires exhibiting misalignment issues; For each image in the tire misalignment image set, label whether a misalignment problem occurs; The target image dataset is obtained by performing image resizing, pixel value normalization, rotation, scaling, and flipping operations on the tire misalignment image set. The target image dataset is divided into training set, validation set and test set according to different functions.
6. The method according to claim 1, characterized in that, The step of determining the basic dimensions of the flexible material to be detected based on the target stitched image includes: The target stitched image is subjected to image sharpening, image denoising, edge extraction, and morphological filtering operations to obtain the image edges of the target stitched image; The discontinuities at the edges of the target stitched image are filled by morphological closing operations, and the position of the material head of the flexible material to be detected is determined based on line detection using Hough transform. Based on the head position and standard length information, the tail position of the flexible material to be tested is determined; The overlap amount of the flexible material to be tested is determined based on the material head position and the material tail position; An adaptive threshold segmentation method and sub-pixel edge detection are used to determine the left and right width edges, as well as the top and bottom edges of the flexible material to be detected in the target stitched image. The width of the material is determined based on the left width edge and the right width edge; The off-center amount is determined based on the upper edge and the lower edge, as well as the centerline of the material conveyor. A two-dimensional coordinate system mapping model is established. The geometric relationship of the edges in the target stitched image is determined through the two-dimensional coordinate system mapping model, and the misalignment region is determined by combining the region growing method.
7. The method according to claim 1, characterized in that, The process of acquiring the target stitched image corresponding to the flexible material to be detected includes: The material conveyor controller receives a start transmission signal forwarded by the tire forming machine controller, and upon receiving the start transmission signal sent by the tire forming machine controller, controls the material conveyor to start transmitting the flexible material to be tested. The start transmission signal is sent to the industrial camera, wherein, when the industrial camera receives the start transmission signal, the material transmission image is acquired at a preset image acquisition frequency within a preset material transmission time period. Acquire images of each material transfer sequentially captured by the industrial camera; The images of each material transport are stitched together to obtain the target stitched image.
8. A misalignment detection device for materials, applied to a tire forming machine, wherein the flexible material to be detected is bonded to the belt drum / forming drum of the tire forming machine, characterized in that... The device includes: The acquisition module is used to acquire a target stitched image corresponding to the flexible material to be detected, wherein the target stitched image is stitched together from various material transport images captured sequentially by an industrial camera during the process of the flexible material to be detected being transported on the material transport machine; The first determining module is used to determine the basic dimensions of the flexible material to be detected based on the target stitched image, wherein the basic dimensions include the material head position, the material tail position, the overlap amount, the misalignment area, the off-center amount, and the material width. The second determining module is used to determine the position region of the flexible material to be detected in the target stitched image based on the basic dimension, and to extract an initial material region image containing the flexible material to be detected from the target stitched image based on the position region. The first processing module is used to further fit the initial material region image to obtain a target material region image containing the flexible material to be detected. The second processing module is used to input the basic dimension and the target material area image into a pre-trained target deep learning model for feature extraction, and to obtain misaligned edge features and misaligned edge morphological features. The third determining module is used to determine the amount of misalignment of the flexible material to be tested based on the misalignment edge features and the misalignment edge morphology features, and to determine whether the flexible material to be tested has a misalignment problem based on the amount of misalignment.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-7.