Logistics transfer center conveyor belt deviation detection method and device, medium and equipment

Abstract

The invention provides a logistics transfer center conveyor belt deviation detection method and device, a medium and equipment, and the method comprises the steps: obtaining a logistics transfer center conveyor belt picture, and carrying out the preprocessing of the obtained logistics transfer center conveyor belt picture; constructing a data set based on the preprocessed logistics transfer center conveyor belt pictures; the data set is used for training a detection model; constructing a detection model, and training the constructed detection model by using the data set to obtain a trained detection model; the trained detection model is used for conducting comprehensive detection on images, obtained in real time, of the conveying belt of the logistics transfer center, and a belt deviation result is obtained; according to the detection model, an SAM model and a LaneATT model are fused to achieve deviation detection; wherein the SAM model is used for realizing segmentation of the conveyor belt; the LaneATT model is used for detecting tracks on the left side and the right side of the conveyor belt; deviation detection is carried out based on conveyor belt segmentation and conveyor belt left and right track detection.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics management, and specifically, to a method, device, medium, and equipment for detecting conveyor belt deviation in a logistics transfer center. Background Art

[0002] In modern logistics transfer centers, the conveyor belt system is a key device for realizing efficient transportation and sorting of goods. With the rapid development of technologies such as the Internet of Things and artificial intelligence, how to monitor the operating status of conveyor belts in real time and detect faults is becoming increasingly important. The problem of conveyor belt deviation is widespread. If not discovered and handled in a timely manner, it may lead to goods rolling down, equipment damage, or even more serious safety accidents. Therefore, detecting conveyor belt deviation is of great significance.

[0003] Traditional conveyor belt status monitoring mostly relies on manual inspections. This method not only has a high labor intensity and low work efficiency but is also easily affected by human factors, resulting in missed detections or misjudgments. Although sensor and video monitoring technologies have been introduced in some high-end logistics parks, there is still a lack of real-time analysis capabilities in complex environments and it is unable to efficiently identify and locate deviation problems. These traditional monitoring methods show certain limitations when faced with changing types of goods, transportation conditions, and environmental lighting, and cannot meet the needs of intelligence and automation in modern logistics transfer centers. Summary of the Invention

[0004] The main objective of the present invention is to solve the technical problem in the prior art that the intelligence and automation requirements of modern logistics transfer centers cannot be met.

[0005] The first aspect of the present invention provides a method for detecting conveyor belt deviation in a logistics transfer center, including: Obtain a conveyor belt picture of the logistics transfer center, and preprocess the obtained conveyor belt picture of the logistics transfer center to obtain a preprocessed conveyor belt picture of the logistics transfer center; Construct a data set based on the preprocessed conveyor belt picture of the logistics transfer center; the data set is used to train a detection model; Construct a detection model, and use the data set to train the constructed detection model to obtain a trained detection model; Use the trained detection model to comprehensively detect the conveyor belt picture of the logistics transfer center obtained in real time to obtain a conveyor belt deviation result; The detection model realizes deviation detection by fusing the SAM model and the LaneATT model; wherein, the SAM model is used to realize conveyor belt segmentation; the LaneATT model is used to realize detection of the left and right tracks of the conveyor belt; deviation detection is performed based on conveyor belt segmentation and detection of the left and right tracks of the conveyor belt.

[0006] Optionally, in the first implementation manner of the first aspect of the present invention, the obtaining of the conveyor belt picture of the logistics transfer center and the preprocessing of the obtained conveyor belt picture of the logistics transfer center to obtain the preprocessed conveyor belt picture of the logistics transfer center include: Taking pictures of the conveyor belt from multiple angles through a camera to obtain conveyor belt pictures of the logistics transfer center under different time periods, different cargo carrying states, and different lighting conditions; Performing grayscale processing on the obtained conveyor belt pictures of the logistics transfer center to obtain grayscale pictures of the conveyor belt of the logistics transfer center; Performing denoising processing on the obtained grayscale pictures of the conveyor belt of the logistics transfer center to obtain denoised grayscale pictures of the conveyor belt of the logistics transfer center; Performing image enhancement processing on the obtained denoised grayscale pictures of the conveyor belt of the logistics transfer center to obtain enhanced grayscale pictures of the conveyor belt of the logistics transfer center; Performing image cropping and scaling processing on the obtained enhanced grayscale pictures of the conveyor belt of the logistics transfer center to obtain processed grayscale pictures of the conveyor belt of the logistics transfer center.

[0007] Optionally, in the second implementation manner of the first aspect of the present invention, the constructing of the data set based on the preprocessed conveyor belt picture of the logistics transfer center includes: Using the LabelImg annotation tool to annotate the goods, the belt in the conveyor belt, and the track in the preprocessed conveyor belt picture of the logistics transfer center; Among them, the annotating of the goods, the belt in the conveyor belt, and the track in the preprocessed conveyor belt picture of the logistics transfer center includes: annotating the edge and wear condition of the belt in the conveyor belt; annotating the shape and position of the goods; annotating the edge and features of the track; Constructing a data set based on the annotated preprocessed conveyor belt picture of the logistics transfer center; Constructing a training set and a validation set based on the data set; The training set is used to train the detection model; the validation set is used to verify the detection model and determine the fusion weights in the detection model.

[0008] Optionally, in the third implementation manner of the first aspect of the present invention, the SAM model is used to implement conveyor belt segmentation, including: Constructing the SAM model based on the attention mechanism in deep learning technology; The SAM model realizes the recognition and segmentation of the belt by focusing on the edge and texture of the belt on the conveyor belt; Train the SAM model using a dataset so that the SAM model learns the shape and position of the goods and the edge and texture features of the belt, obtaining the trained SAM model, so that the trained SAM model can identify the belt and perform segmentation.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present invention, the LaneATT model is used to implement the detection of the tracks on both sides of the conveyor belt, including: Use the Augmentor module to perform data augmentation on the conveyor belt pictures of the logistics transfer center in the dataset, obtaining the dataset after data augmentation processing; Construct the LaneATT model based on a convolutional neural network; Use the dataset after data augmentation processing to train the constructed LaneATT model, obtaining the trained LaneATT model; Use the trained LaneATT model to detect the tracks on both sides of the conveyor belt; The use of the Augmentor module to perform data augmentation on the conveyor belt pictures of the logistics transfer center in the dataset includes: randomly rotating the conveyor belt pictures of the logistics transfer center in the dataset to obtain the rotated conveyor belt pictures of the logistics transfer center; mirror flipping the rotated conveyor belt pictures of the logistics transfer center to obtain the mirror-flipped conveyor belt pictures of the logistics transfer center; scaling the mirror-flipped conveyor belt pictures of the logistics transfer center to obtain the scaled conveyor belt pictures of the logistics transfer center.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present invention, the detection model includes: Use the validation set to verify the SAM model and the LaneATT model respectively; Assign corresponding weights to the verification results of the SAM model and the LaneATT model based on the validation set; Perform weighted fusion based on the SAM model, the LaneATT model, and the corresponding weights to obtain the detection model.

[0011] Optionally, in the sixth implementation manner of the first aspect of the present invention, the detection model further includes: performing belt deviation detection based on the belt segmentation result and the track positioning information; The performing belt deviation detection based on the belt segmentation result and the track positioning information includes: when the belt segmentation result is outside the track range, it is considered that the current belt is deviated; When it is detected that the current belt is deviated, an alarm warning is given and displayed through the visual monitoring interface.

[0012] The second aspect of the present invention provides a device for detecting belt deviation of a logistics transfer center conveyor belt, including: An image acquisition module, configured to acquire images of the conveyor belt in the logistics transfer center, and preprocess the acquired images of the conveyor belt in the logistics transfer center to obtain preprocessed images of the conveyor belt in the logistics transfer center; A dataset construction module, configured to construct a dataset based on the preprocessed images of the conveyor belt in the logistics transfer center; the dataset is used to train a detection model; A model training module, configured to construct a detection model, and use the dataset to train the constructed detection model to obtain a trained detection model; A comprehensive detection module, configured to comprehensively detect the images of the conveyor belt in the logistics transfer center acquired in real time by using the trained detection model to obtain a belt deviation result; The detection model realizes deviation detection by fusing the SAM model and the LaneATT model; wherein, the SAM model is used to realize the segmentation of the conveyor belt; the LaneATT model is used to realize the detection of the left and right tracks of the conveyor belt; deviation detection is performed based on the conveyor belt segmentation and the detection of the left and right tracks of the conveyor belt.

[0013] Optionally, in the first implementation manner of the second aspect of the present invention, the image acquisition module includes: Shoot the conveyor belt from multiple angles through a camera to acquire images of the conveyor belt in the logistics transfer center under different time periods, different cargo carrying states, and different lighting conditions; Perform grayscale processing on the acquired images of the conveyor belt in the logistics transfer center to obtain grayscale images of the conveyor belt in the logistics transfer center; Perform denoising processing on the obtained grayscale images of the conveyor belt in the logistics transfer center to obtain denoised grayscale images of the conveyor belt in the logistics transfer center; Perform image enhancement processing on the obtained denoised grayscale images of the conveyor belt in the logistics transfer center to obtain enhanced grayscale images of the conveyor belt in the logistics transfer center; Perform image cropping and scaling processing on the obtained enhanced grayscale images of the conveyor belt in the logistics transfer center to obtain processed grayscale images of the conveyor belt in the logistics transfer center.

[0014] Optionally, in the second implementation manner of the second aspect of the present invention, the image acquisition module includes: Label the goods, the belt on the conveyor belt, and the tracks in the preprocessed images of the conveyor belt in the logistics transfer center through the LabelImg annotation tool; Wherein, the labeling of the goods, the belt on the conveyor belt, and the tracks in the preprocessed images of the conveyor belt in the logistics transfer center includes: labeling the edge and wear condition of the belt on the conveyor belt; labeling the shape and position of the goods; labeling the edge and features of the tracks; Construct a dataset based on the pre - processed conveyor belt pictures of the logistics transfer center with annotations; Construct a training set and a validation set based on the dataset; The training set is used to train the detection model; the validation set is used to verify the detection model and determine the fusion weights in the detection model.

[0015] Optionally, in the third implementation manner of the second aspect of the present invention, the SAM model is used to implement conveyor belt segmentation, including: Construct the SAM model based on the attention mechanism in deep learning technology; The SAM model realizes the recognition and segmentation of the belt by focusing on the edges and textures of the belt on the conveyor; Use the dataset to train the SAM model so that the SAM model learns the shape and position of the goods and the edge features and texture features of the belt, and obtain the trained SAM model so that the trained SAM model can recognize the belt and perform segmentation.

[0016] Optionally, in the fourth implementation manner of the second aspect of the present invention, the LaneATT model is used to implement the detection of the left and right tracks of the conveyor belt, including: Use the Augmentor module to perform data augmentation on the conveyor belt pictures of the logistics transfer center in the dataset to obtain the dataset after data augmentation processing; Construct the LaneATT model based on the convolutional neural network; Use the dataset after data augmentation processing to train the constructed LaneATT model to obtain the trained LaneATT model; Use the trained LaneATT model to detect the left and right tracks of the conveyor belt; The use of the Augmentor module to perform data augmentation on the conveyor belt pictures of the logistics transfer center in the dataset includes: randomly rotating the conveyor belt pictures of the logistics transfer center in the dataset to obtain the rotated conveyor belt pictures of the logistics transfer center; mirror - flipping the rotated conveyor belt pictures of the logistics transfer center to obtain the mirror - flipped conveyor belt pictures of the logistics transfer center; scaling the mirror - flipped conveyor belt pictures of the logistics transfer center to obtain the scaled conveyor belt pictures of the logistics transfer center.

[0017] Optionally, in the fifth implementation manner of the second aspect of the present invention, the detection model includes: Use the validation set to verify the SAM model and the LaneATT model respectively; Assign corresponding weights to the verification results of the SAM model and the LaneATT model based on the validation set; A detection model is obtained by weighted fusion based on the SAM model, the LaneATT model, and their corresponding weights.

[0018] Optionally, in the sixth implementation manner of the second aspect of the present invention, the detection model further includes: detecting belt deviation based on the belt segmentation result and the track positioning information; The detecting belt deviation based on the belt segmentation result and the track positioning information includes: when the belt segmentation result is outside the track range, it is considered that the current belt is deviated; When it is detected that the current belt is deviated, an alarm warning is given and displayed through a visual monitoring interface.

[0019] The third aspect of the present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for detecting belt deviation in a logistics transfer center are implemented. The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the above-mentioned method for detecting belt deviation in a logistics transfer center are implemented.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. By combining the SAM model and the LaneATT model, the present invention can comprehensively analyze the belt segmentation and track positioning of the conveyor belt; then detect the deviation phenomenon and reduce the possibilities of missed detection and misjudgment; 2. The conveyor belt images obtained in real time by the present invention can be quickly analyzed through preprocessing and deep learning models, and the deviation situation can be identified in the first time. This near-real-time processing ability enables operation and maintenance personnel to discover problems in time and take measures, thereby ensuring the safety and effectiveness of logistics operations; 3. The present invention collects and processes images under various lighting conditions, different cargo types and states, and can adapt to the complex environments existing in actual work; through multi-angle collection, denoising and enhancement processing of images, the diversity of the data set is ensured, thereby improving the robustness of the model in different situations; 4. The present invention can realize automatic status detection and alarm, which not only improves work efficiency, but also reduces the risk of errors caused by human factors and liberates human resources; 5. Through the visual monitoring interface, the present invention enables operation and maintenance personnel to intuitively monitor the running status of the conveyor belt. Through different color markings, users can quickly identify potential problem areas, improving the response speed and decision-making efficiency. Description of the Drawings

[0021] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments read in conjunction with the accompanying drawings: Figure 1 This is the first flow chart of the conveyor belt deviation detection method provided for the embodiments of the present invention in a logistics transfer center.

[0022] Figure 2 This is the second flow chart of the conveyor belt deviation detection method provided for the embodiments of the present invention in a logistics transfer center.

[0023] Figure 3 This is a schematic structural diagram of the conveyor belt deviation detection device provided for the embodiments of the present invention in a logistics transfer center.

[0024] Figure 4 This is a schematic structural diagram of the electronic device provided for the embodiments of the present invention. Detailed implementation manners

[0025] The embodiments of the present invention provide a conveyor belt deviation detection method, device, medium, and equipment for a logistics transfer center, including: obtaining pictures of the conveyor belt in the logistics transfer center, and preprocessing the obtained pictures of the conveyor belt in the logistics transfer center to obtain preprocessed pictures of the conveyor belt in the logistics transfer center; constructing a data set based on the preprocessed pictures of the conveyor belt in the logistics transfer center; the data set is used to train a detection model; constructing a detection model, and training the constructed detection model using the data set to obtain a trained detection model; using the trained detection model to comprehensively detect the pictures of the conveyor belt in the logistics transfer center obtained in real time to obtain a belt deviation result; the detection model realizes deviation detection by fusing the SAM model and the LaneATT model; wherein, the SAM model is used to realize the segmentation of the conveyor belt; the LaneATT model is used to realize the detection of the left and right tracks of the conveyor belt; deviation detection is performed based on the conveyor belt segmentation and the detection of the left and right tracks of the conveyor belt. The present invention solves the technical problem that the prior art cannot meet the intelligent and automated requirements of modern logistics transfer centers.

[0026] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these process, method, product, or device.

[0027] For ease of understanding, the specific process of the embodiments of the present invention will be described below. Please refer to Figure 1 , the first embodiment of the conveyor belt deviation detection method in the logistics transfer center of the embodiments of the present invention includes: 101. Obtain pictures of the conveyor belt in the logistics transfer center, and preprocess the obtained pictures of the conveyor belt in the logistics transfer center to obtain preprocessed pictures of the conveyor belt in the logistics transfer center; In this embodiment, a high-resolution camera is used to take pictures of the conveyor belt from multiple angles to obtain pictures of the conveyor belt in the logistics transfer center under different time periods, different cargo carrying states, and different lighting conditions; Preprocess the obtained pictures of the conveyor belt in the logistics transfer center to obtain preprocessed pictures of the conveyor belt in the logistics transfer center; In this embodiment, the implementation process of the preprocessing includes: The grayscale operation is the primary link. The grayscale is used to simplify the image color information, reduce the computational complexity, convert the color image into a grayscale image, and discard the redundant computational amount brought by the color information, so that the subsequent processing can focus on the brightness and darkness structure of the image, greatly reducing the computational complexity. For example, for a color RGB image of 1024×768 pixels, each pixel point has 3 color channels, and the data volume is reduced to one-third of the original after grayscale conversion.

[0028] Apply denoising technology to reduce noise interference in the image and improve image clarity. The logistics site is filled with various noise sources, such as machine operation vibration, electrical interference, etc. These noises will blur the image details. The common median filtering denoising method replaces the current pixel value by selecting the median value in the pixel neighborhood, effectively removing salt-and-pepper noise, etc., and restoring the clear outline of the goods and the conveyor belt.

[0029] Improve the image quality through image enhancement technologies such as contrast adjustment and histogram equalization, making the features more obvious. Contrast adjustment can stretch the grayscale value range of the image, making the originally dim and blurred edges of the goods sharp and clear; histogram equalization redistributes the grayscale histogram of the image pixels, making the overall brightness and contrast of the image more balanced and highlighting the key features.

[0030] Crop and scale the image to meet the requirements of the model input. If the model input requires a fixed size of 224×224 pixels, then the original images that are too large or too small need to be adjusted accordingly to ensure the unity of the input data format and avoid errors in model training caused by size differences.

[0031] 102. Construct a data set based on the preprocessed pictures of the conveyor belt in the logistics transfer center; the data set is used to train the detection model; In this embodiment, in the picture annotation process, professional annotation tools are used to accurately identify and mark key information such as goods, conveyor belt edges, and tracks in the image; for example: using annotation tools such as LabelImg, carefully select and mark the goods category and location, and accurately outline the edges of the conveyor belt and the edges and features of the track. For the goods, not only the type is annotated, but also details such as the stacking form and whether there is damage are recorded; for the conveyor belt edge, it is annotated to the millimeter level without missing any subtle arc changes, providing accurate learning targets for subsequent image recognition.

[0032] 103. Build a detection model, and use the data set to train the built detection model to obtain a trained detection model; 104. Use the trained detection model to comprehensively detect the conveyor belt pictures of the logistics transfer center obtained in real time to obtain the belt deviation result; In this embodiment, the detection model realizes deviation detection by fusing the SAM model and the LaneATT model; among them, the SAM model is used to realize the segmentation of the conveyor belt; the LaneATT model is used to detect the tracks on the left and right sides of the conveyor belt; deviation detection is performed based on the conveyor belt segmentation and the detection of the tracks on the left and right sides of the conveyor belt.

[0033] Among them, the SAM model includes: building the SAM model based on the attention mechanism in deep learning technology; The SAM model realizes the recognition and segmentation of the belt by focusing on the edge and texture of the belt on the conveyor belt; Use the data set to train the SAM model so that the SAM model learns the shape and position of the goods and the edge features and texture features of the belt, and obtain a trained SAM model so that the trained SAM model can recognize and segment the belt.

[0034] The LaneATT model includes: using the Augmentor module to perform data augmentation processing on the conveyor belt pictures of the logistics transfer center in the data set to obtain a data set after data augmentation processing; Build the LaneATT model based on the convolutional neural network; Use the data set after data augmentation processing to train the built LaneATT model to obtain a trained LaneATT model; Use the trained LaneATT model to detect the tracks on the left and right sides of the conveyor belt; The use of the Augmentor module to perform data augmentation on the conveyor belt pictures of the logistics transfer center in the dataset includes: randomly rotating the conveyor belt pictures of the logistics transfer center in the dataset to obtain the rotated conveyor belt pictures of the logistics transfer center; mirror-flipping the rotated conveyor belt pictures of the logistics transfer center to obtain the mirror-flipped conveyor belt pictures of the logistics transfer center; and scaling the mirror-flipped conveyor belt pictures of the logistics transfer center to obtain the scaled conveyor belt pictures of the logistics transfer center.

[0035] The detection model includes: using the validation set to verify the SAM model and the LaneATT model respectively; Assign corresponding weights to the verification results of the SAM model and the LaneATT model based on the validation set; Perform weighted fusion based on the SAM model, the LaneATT model, and the corresponding weights to obtain the detection model.

[0036] The detection model further includes: performing belt deviation detection based on the belt segmentation result and the track positioning information; The belt deviation detection based on the belt segmentation result and the track positioning information includes: when the belt segmentation result is outside the track range, it is considered that the current belt is deviated; When it is detected that the current belt is deviated, an alarm warning is issued and displayed through the visual monitoring interface.

[0037] Please refer to Figure 2 , the second embodiment of the belt deviation detection method for the conveyor belt of the logistics transfer center in the embodiment of the present invention includes: 201. Obtain the conveyor belt pictures of the logistics transfer center, and preprocess the obtained conveyor belt pictures of the logistics transfer center to obtain the preprocessed conveyor belt pictures of the logistics transfer center; 202. Construct a dataset based on the preprocessed conveyor belt pictures of the logistics transfer center; the dataset is used to train the detection model; 203. Construct a detection model, and use the dataset to train the constructed detection model to obtain the trained detection model; 204. Use the trained detection model to comprehensively detect the conveyor belt pictures of the logistics transfer center obtained in real time to obtain the belt deviation result; In this embodiment, the detection model realizes deviation detection by fusing the SAM model and the LaneATT model; wherein, the SAM model is used to realize the belt segmentation of the conveyor belt; the LaneATT model is used to realize the detection of the left and right tracks of the conveyor belt; deviation detection is performed based on the conveyor belt segmentation and the detection of the left and right tracks of the conveyor belt.

[0038] Among them, the SAM model includes: During the process of deploying the large model SAM (Structural Attention Model) to segment the conveyor belt, the key lies in using deep learning techniques, especially the attention mechanism, to accurately identify and segment the belt area on the conveyor belt. The model training begins by inputting a large amount of image data containing conveyor belts to learn the unique texture features of the belt, such as the fine lines characteristic of rubber materials, the traces after wear, and the special light and shadow changes at the connection with the metal rollers. This model can automatically focus on key information such as the edges and textures of the belt by learning the features and structures in the image, thus achieving efficient segmentation of the belt.

[0039] During actual deployment, it is first necessary to configure the model parameters according to the computing resources of the logistics transfer center. If there is a high-performance GPU cluster on-site, the depth and width of the model can be appropriately increased to improve the segmentation accuracy; otherwise, the model structure needs to be streamlined to ensure real-time performance. After the model loads the image, the attention mechanism quickly comes into play, locking onto the belt area like a searchlight and automatically filtering out interference factors such as goods and surrounding equipment. After introducing the SAM model, the original method of relying on manual inspections to judge the running status and wear conditions of the belt has been revolutionized. The model can process dozens of high-definition images per second and output the belt segmentation results in real time. Once abnormalities such as belt tearing or deviation are detected, an alarm is immediately triggered, raising the level of automated monitoring to a new height and greatly optimizing the accuracy and response speed of the conveyor belt monitoring system.

[0040] The LaneATT model includes: constructing the LaneATT model to detect the left and right tracks of the conveyor belt, and using the Augmentor module for data augmentation during training; The purpose of constructing the LaneATT model is to accurately capture the information of the left and right tracks of the conveyor belt to ensure the accuracy of the goods transportation route. In the early stage of training, the data quality and diversity are crucial, and the Augmentor module takes on this important task. The random rotation operation simulates the slight deviation of the camera shooting angle, randomly rotating the image within the range of -30° to 30° to make the model adapt to the track forms under different installation angles; the mirror flipping copies and flips the image in the left-right and up-down directions to expand the data diversity, especially suitable for track scenarios with symmetric structures; the scaling operation randomly changes the image size in the ratio of 0.8 to 1.2, and then combines it with cropping to a fixed size, enabling the model to recognize the track features at different distances and resolutions. The Augmentor constructs an augmentation pipeline through simple Python code. After being processed by the Augmentor, the limited original data set is expanded several times or even dozens of times, and the model encounters more diverse track scenarios during training, significantly improving its robustness. After being put into actual use, whether it is direct strong light, shadow occlusion, or slight wear on the track surface, the LaneATT model can accurately locate the track to ensure the smooth and orderly transportation of goods.

[0041] The detection model includes: adopting a weighted fusion strategy to integrate the SAM model and the LaneATT model. SAM focuses on the belt, and LaneATT focuses on the track. For example, according to the performance on the previous validation set, a weight of 0.6 is assigned to the belt segmentation result of the SAM model, and a weight of 0.4 is assigned to the track detection result of the LaneATT model. The fused output contains both the accurate belt area and the stable track positioning. In the model prediction stage, when inputting the real-time conveyor belt image stream, the fusion model quickly gives the comprehensive prediction result. Using the existing edge computing devices in the logistics transfer center, such as on-site servers or intelligent gateways with a certain computing power, the model is deployed on them to achieve near-real-time processing. At the software level, a supporting visual monitoring interface is developed, which uses different colors to identify the belt, track, and cargo status, and intuitively displays it to the operation and maintenance personnel. Once the model predicts an anomaly, such as the deviation of the belt from the track or the risk of cargo dropping, it not only pops up a warning on the interface but also notifies relevant personnel in multiple channels such as text messages and in-site broadcasts for timely intervention, comprehensively ensuring the efficient and safe operation of logistics transfer.

[0042] The above describes the conveyor belt deviation detection method in the logistics transfer center in the embodiment of the present invention. Next, the conveyor belt deviation detection device in the embodiment of the present invention will be described. Please refer to Figure 3 One embodiment of the conveyor belt deviation detection device in the embodiment of the present invention includes: An image acquisition module 301, configured to acquire conveyor belt images of the logistics transfer center and preprocess the acquired conveyor belt images of the logistics transfer center to obtain preprocessed conveyor belt images of the logistics transfer center; In this embodiment, the image acquisition module 301 includes: Using a high-resolution camera to take pictures of the conveyor belt from multiple angles to acquire conveyor belt images of the logistics transfer center at different times, different cargo carrying states, and different lighting conditions; Preprocess the acquired conveyor belt images of the logistics transfer center to obtain preprocessed conveyor belt images of the logistics transfer center; In this embodiment, the implementation process of the preprocessing includes: The grayscale operation is the primary link. The grayscale is used to simplify the image color information, reduce the computational complexity, convert the color image into a grayscale image, and discard the redundant computational amount brought by the color information, so that the subsequent processing can focus on the light and dark structure of the image, greatly reducing the computational complexity. For example, for a color RGB image of 1024×768 pixels, each pixel point has 3 color channels, and the data volume is reduced to one-third of the original after grayscale.

[0043] Apply denoising techniques to reduce noise interference in images and improve image clarity. The logistics site is filled with various noise sources, such as machine operation vibrations, electrical interference, etc. These noises will blur the details of the images. The common median filtering denoising method replaces the current pixel value with the median value within the pixel neighborhood, effectively removing salt-and-pepper noise, etc., and restoring the clear contours of the goods and the conveyor belt.

[0044] Enhance the image quality through image enhancement techniques such as contrast adjustment and histogram equalization to make the features more obvious. Contrast adjustment can stretch the grayscale value range of the image, making the originally dim and blurred edges of the goods sharp and clear; histogram equalization redistributes the grayscale histogram of the image pixels, making the overall brightness and contrast of the image more balanced and highlighting the key features.

[0045] Crop and scale the image to meet the requirements of the model input. If the model input requires a fixed size of 224×224 pixels, then the original images that are too large or too small need to be adjusted accordingly to ensure the uniformity of the input data format and avoid errors in model training caused by size differences.

[0046] The dataset construction module 302 is used to construct a dataset based on the preprocessed images of the conveyor belt in the logistics transfer center; the dataset is used to train the detection model; In this embodiment, the dataset construction module 302 includes: In the picture annotation link, use professional annotation tools to accurately identify and mark key information such as goods, conveyor belt edges, and tracks in the image; for example: use annotation tools such as LabelImg to carefully frame and mark the goods categories and positions, accurately outline the conveyor belt edges, and the edges and features of the tracks. For the goods, not only mark their types, but also record details such as their stacking forms and whether there are damages; for the conveyor belt edges, mark them to the millimeter level without missing any subtle arc changes, providing accurate learning targets for subsequent image recognition.

[0047] The model training module 303 is used to construct a detection model and train the constructed detection model using the dataset to obtain the trained detection model; The comprehensive detection module 304 is used to comprehensively detect the real-time obtained images of the conveyor belt in the logistics transfer center using the trained detection model to obtain the belt deviation result; In this embodiment, the detection model realizes deviation detection by fusing the SAM model and the LaneATT model; among them, the SAM model is used to realize the segmentation of the conveyor belt; the LaneATT model is used to realize the detection of the tracks on the left and right sides of the conveyor belt; deviation detection is performed based on the conveyor belt segmentation and the detection of the tracks on the left and right sides of the conveyor belt.

[0048] Among them, the SAM model includes: constructing the SAM model based on the attention mechanism in deep learning technology; The SAM model realizes the recognition and segmentation of the belt by focusing on the belt edge and texture on the conveyor belt; The SAM model is trained using a dataset so that the SAM model learns the shape and position of the goods and the edge features and texture features of the belt, and obtains the trained SAM model, so that the trained SAM model can recognize the belt and perform segmentation.

[0049] Model training begins. A large number of image data containing conveyor belts are input to learn the unique texture features of the belt, such as the fine lines unique to rubber materials, the traces after wear, and the special light and shadow changes at the connection with the metal rollers. This model can automatically focus on key information such as the edge and texture of the belt by learning the features and structures in the image, thus achieving efficient segmentation of the belt. During actual deployment, the model parameters need to be configured according to the computing resources of the logistics transfer center first. If there is a high-performance GPU cluster on-site, the depth and width of the model can be appropriately increased to improve the segmentation accuracy; otherwise, the model structure needs to be streamlined to ensure real-time performance. After the model loads the image, the attention mechanism quickly takes effect, locking the belt area like a searchlight and automatically filtering out interference factors such as goods and surrounding equipment. After introducing the SAM model, the original method of relying on manual inspections to judge the running state and wear condition of the belt has been innovated. The model can process dozens of high-definition images per second and output the belt segmentation results in real time. Once abnormal conditions such as belt tearing and deviation are detected, an alarm is immediately triggered, raising the level of automated monitoring to a new height and greatly optimizing the accuracy and response speed of the conveyor belt monitoring system.

[0050] The LaneATT model includes: using the Augmentor module to perform data augmentation processing on the conveyor belt pictures of the logistics transfer center in the dataset to obtain the dataset after data augmentation processing; Constructing the LaneATT model based on a convolutional neural network; Training the constructed LaneATT model using the dataset after data augmentation processing to obtain the trained LaneATT model; Using the trained LaneATT model to detect the left and right tracks of the conveyor belt; The use of the Augmentor module to perform data augmentation on the conveyor belt images of the logistics transfer center in the dataset includes: randomly rotating the conveyor belt images of the logistics transfer center in the dataset to obtain the rotated conveyor belt images of the logistics transfer center; in this embodiment, the random rotation operation simulates the slight deviation of the camera shooting angle, randomly rotates the image within the range of -30° to 30°, so that the model can adapt to the track forms under different installation angles; mirror-flipping the rotated conveyor belt images of the logistics transfer center to obtain the mirror-flipped conveyor belt images of the logistics transfer center; in this embodiment, the mirror-flipping copies and flips the image from the left-right and up-down directions to expand data diversity, especially suitable for track scenarios with symmetric structures; scaling the mirror-flipped conveyor belt images of the logistics transfer center to obtain the scaled conveyor belt images of the logistics transfer center; in this embodiment, the scaling operation randomly changes the image size in the ratio of 0.8 to 1.2, and then combines with cropping to a fixed size, so that the model can identify the track features at different distances and different resolutions.

[0051] Augmentor constructs an augmentation pipeline through concise Python code. After being processed by Augmentor, the limited original dataset is expanded several times or even dozens of times, and the model encounters more diverse track scenarios during training, and the robustness is significantly improved. After being put into actual use, whether it is direct sunlight, shadow occlusion or slight wear on the track surface, the LaneAT model can accurately locate the track to ensure the smooth and orderly transportation of goods.

[0052] The detection model includes: using the validation set to verify the SAM model and the LaneATT model respectively; Based on the verification results of the SAM model and the LaneATT model on the validation set, corresponding weights are assigned respectively; Based on the SAM model, the LaneATT model and the corresponding weights, a weighted fusion is performed to obtain the detection model.

[0053] In this embodiment, a weight of 0.6 is assigned to the belt segmentation result of the SAM model, and a weight of 0.4 is assigned to the track detection result of the LaneATT model. The fused output contains both the accurate belt area and the stable track positioning.

[0054] The detection model further includes: performing belt deviation detection based on the belt segmentation result and the track positioning information; The belt deviation detection based on the belt segmentation result and the track positioning information includes: when the belt segmentation result is outside the track range, it is considered that the current belt is deviated; When it is detected that the current belt is deviated, an alarm warning is given and displayed through the visual monitoring interface.

[0055] aboveFigure 3 The conveyor belt deviation detection device in the embodiment of the present invention will be described in detail from the perspective of modular functional entities. Next, the electronic device in the embodiment of the present invention will be described in detail from the perspective of hardware processing.

[0056] Figure 4 FIG. 5 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. The electronic device 400 may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPUs) 410 (for example, one or more processors) and a memory 420, and one or more storage media 430 for storing application programs 433 or data 432 (for example, one or more mass storage devices). Among them, the memory 420 and the storage media 430 may be transient storage or persistent storage. The program stored in the storage media 430 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the electronic device 400. Further, the processor 410 may be configured to communicate with the storage media 430 and execute a series of instruction operations in the storage media 430 on the electronic device 400.

[0057] The electronic device 400 may further include one or more power supplies 440, one or more wired or wireless network interfaces 450, one or more input / output interfaces 450, and / or one or more operating systems 431, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art can understand that Figure 4 the shown structure of the electronic device does not constitute a limitation on the electronic device, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0058] The present invention also provides a computer-readable storage medium. The computer-readable storage medium may be a non-volatile computer-readable storage medium, or may also be a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the conveyor belt deviation detection method for the logistics transfer center.

[0059] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system or device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0060] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0061] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. A method for detecting the deviation of a conveyor belt in a logistics transfer center, characterized in that, Including: Obtain pictures of the conveyor belt in the logistics transfer center, and preprocess the obtained pictures of the conveyor belt in the logistics transfer center to obtain the preprocessed pictures of the conveyor belt in the logistics transfer center; Construct a dataset based on the preprocessed pictures of the conveyor belt in the logistics transfer center; the dataset is used to train the detection model; Construct a detection model, and use the dataset to train the constructed detection model to obtain the trained detection model; Use the trained detection model to comprehensively detect the pictures of the conveyor belt in the logistics transfer center obtained in real time, and obtain the result of belt deviation; The detection model realizes deviation detection by fusing the SAM model and the LaneATT model; among them, the SAM model is used to realize the segmentation of the conveyor belt; the LaneATT model is used to realize the detection of the left and right tracks of the conveyor belt; deviation detection is carried out based on the conveyor belt segmentation and the detection of the left and right tracks of the conveyor belt.

2. The conveyor belt deviation detection method for a logistics transfer center according to claim 1, characterized in that, The obtaining pictures of the conveyor belt in the logistics transfer center, and preprocessing the obtained pictures of the conveyor belt in the logistics transfer center to obtain the preprocessed pictures of the conveyor belt in the logistics transfer center, includes: Take pictures of the conveyor belt from multiple angles through a camera to obtain pictures of the conveyor belt in the logistics transfer center under different time periods, different cargo carrying states, and different lighting conditions; Perform grayscale processing on the obtained pictures of the conveyor belt in the logistics transfer center to obtain grayscale pictures of the conveyor belt in the logistics transfer center; Perform denoising processing on the obtained grayscale pictures of the conveyor belt in the logistics transfer center to obtain the denoised grayscale pictures of the conveyor belt in the logistics transfer center; Perform image enhancement processing on the obtained denoised grayscale pictures of the conveyor belt in the logistics transfer center to obtain the enhanced grayscale pictures of the conveyor belt in the logistics transfer center; Perform image cropping and scaling processing on the obtained enhanced grayscale pictures of the conveyor belt in the logistics transfer center to obtain the processed grayscale pictures of the conveyor belt in the logistics transfer center.

3. The conveyor belt deviation detection method for the logistics transfer center according to claim 1, wherein The constructing a dataset based on the preprocessed pictures of the conveyor belt in the logistics transfer center, includes: Use the LabelImg annotation tool to annotate the goods, the belt in the conveyor belt, and the tracks in the preprocessed pictures of the conveyor belt in the logistics transfer center; Among them, the annotating the goods, the belt in the conveyor belt, and the tracks in the preprocessed pictures of the conveyor belt in the logistics transfer center, includes: annotating the edge and wear condition of the belt in the conveyor belt; annotating the shape and position of the goods; annotating the edge and features of the tracks; Construct a dataset based on the annotated preprocessed pictures of the conveyor belt in the logistics transfer center; Construct a training set and a validation set based on the dataset; The training set is used to train the detection model; the validation set is used to verify the detection model and determine the fusion weights in the detection model.

4. The method for detecting belt deviation of a logistics transfer center according to claim 1, wherein, The SAM model is used to realize the segmentation of the conveyor belt, including: Construct the SAM model based on the attention mechanism in deep learning technology; The SAM model realizes the recognition and segmentation of the belt by focusing on the edge and texture of the belt on the conveyor belt; Train the SAM model using a dataset so that the SAM model learns the shape and position of the goods and the edge and texture features of the belt, and obtain the trained SAM model, so that the trained SAM model can identify the belt and perform segmentation.

5. The method for detecting the deviation of the conveyor belt in the logistics transfer center according to claim 1, characterized in that, The LaneATT model is used to detect the tracks on both the left and right sides of the conveyor belt, including: Use the Augmentor module to perform data augmentation on the conveyor belt pictures in the logistics transfer center in the dataset to obtain the dataset after data augmentation processing; Build the LaneATT model based on the convolutional neural network; Use the dataset after data augmentation processing to train the constructed LaneATT model to obtain the trained LaneATT model; Use the trained LaneATT model to detect the tracks on both the left and right sides of the conveyor belt; The use of the Augmentor module to perform data augmentation on the conveyor belt pictures in the logistics transfer center in the dataset includes: randomly rotating the conveyor belt pictures in the logistics transfer center in the dataset to obtain the rotated conveyor belt pictures in the logistics transfer center; performing mirror flipping on the rotated conveyor belt pictures in the logistics transfer center to obtain the mirror-flipped conveyor belt pictures in the logistics transfer center; performing scaling processing on the mirror-flipped conveyor belt pictures in the logistics transfer center to obtain the scaled conveyor belt pictures in the logistics transfer center.

6. The method for detecting conveyor belt deviation in a logistics transfer center according to claim 1, wherein The detection model includes: Use the validation set to verify the SAM model and the LaneATT model respectively; Assign corresponding weights to the verification results of the SAM model and the LaneATT model based on the validation set; Perform weighted fusion based on the SAM model, the LaneATT model, and the corresponding weights to obtain the detection model.

7. The method for detecting the deviation of the conveyor belt in the logistics transfer center according to claim 1, characterized in that, The detection model also includes: performing belt deviation detection based on the belt segmentation result and the track positioning information; The performing belt deviation detection based on the belt segmentation result and the track positioning information includes: when the belt segmentation result is outside the track range, it is considered that the current belt is deviated; When it is detected that the current belt is deviated, an alarm warning is issued and displayed through the visual monitoring interface.

8. A conveyor belt deviation detection device for a logistics transfer center, characterized in that, Includes: A picture acquisition module for acquiring conveyor belt pictures in the logistics transfer center and preprocessing the acquired conveyor belt pictures in the logistics transfer center to obtain the preprocessed conveyor belt pictures in the logistics transfer center; A dataset construction module for constructing a dataset based on the preprocessed conveyor belt pictures in the logistics transfer center; the dataset is used to train the detection model; A model training module for constructing a detection model, using the dataset to train the constructed detection model to obtain the trained detection model; A comprehensive detection module for comprehensively detecting the conveyor belt pictures in the logistics transfer center obtained in real time using the trained detection model to obtain the belt deviation result; The detection model realizes deviation detection by fusing the SAM model and the LaneATT model; among them, the SAM model is used to realize conveyor belt segmentation; the LaneATT model is used to realize the detection of the tracks on both the left and right sides of the conveyor belt; deviation detection is performed based on conveyor belt segmentation and the detection of the tracks on both the left and right sides of the conveyor belt.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the conveyor belt deviation detection method for a logistics transfer center described in any one of claims 1 to 7.

10. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, When the computer program is executed by a processor, it implements the steps of the conveyor belt deviation detection method for a logistics transfer center described in any one of claims 1 to 7.

Images