A pulp cargo pile outward bulging identification and early warning method, system, terminal and medium based on deep learning
By constructing a deep learning-based pulp stack expansion recognition system, image processing technology is used to achieve real-time monitoring and early warning of stack expansion, solving the accuracy and real-time problems of manual observation and improving the safety and efficiency of port loading and unloading operations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- QINGDAO PORT INT CO LTD
- Filing Date
- 2026-03-20
- Publication Date
- 2026-06-23
AI Technical Summary
In existing technologies, the identification of pulp stack expansion relies on manual observation, which has low accuracy and poor real-time performance, leading to safety hazards and increased costs.
A deep learning-based method is used to construct a pulp stack expansion state recognition network. Image processing technology is used to realize real-time monitoring and expansion recognition. Candidate region boxes are generated by dilated convolution and region proposal network. The expansion recognition is combined with classification and regression mechanisms to trigger early warning.
It enables early detection and prevention of stack expansion, improves identification accuracy and response speed, reduces labor costs, and ensures operational safety and cargo stability.
Smart Images
Figure CN122265713A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer vision technology, specifically to a method, system, terminal, and medium for identifying and warning of outward expansion of pulp stacks based on deep learning. Background Technology
[0002] During loading and unloading at port terminals, the stacking of pulp stacks requires high stability to ensure operational safety and cargo integrity. However, due to various factors such as external vibrations, equipment malfunctions, and uneven stacking pressure during loading, unloading, transportation, and stacking, the stack may experience outward bulging, meaning a localized area of the stack surface protrudes outward. This significantly increases the risk of stack instability and may even cause cargo collapse or tipping, posing a significant safety hazard. Currently, the identification of outward bulging in pulp stacks mainly relies on manual observation, which typically involves two stages: operator inspection and emergency response. During inspection, operators visually observe changes in the shape of the stack surface or abnormal protrusions to determine if outward bulging has occurred. However, manual observation has significant limitations, being restricted by lighting conditions, observation angles, and operator experience, making it difficult to accurately judge subtle changes. Furthermore, current handling methods often lag behind the occurrence of outward bulging, typically only taking emergency remedial measures when the outward bulging is already obvious or even shows signs of collapse. This reactive approach not only increases operational costs but also increases the risk of safety accidents, threatening personnel safety and the economic value of goods. Summary of the Invention
[0003] To address the aforementioned issues, this invention provides a method, system, terminal, and medium for identifying and warning of bulging in pulp stacks based on deep learning. By introducing intelligent image processing technology, it enables real-time monitoring of the surface morphology of the stacks and automatic identification of bulging conditions, thereby achieving early detection and prevention of bulging risks, effectively ensuring operational safety and stable cargo stacking, reducing labor costs, and improving the overall efficiency of port loading and unloading operations.
[0004] In a first aspect, the technical solution of the present invention provides a method for identifying and warning of external expansion of pulp stacks based on deep learning, comprising the following steps: Construct and train a network for recognizing the external expansion state of pulp stacks; Acquire real-time images of pulp stacks; The real-time image of the pulp stack is preprocessed, and the feature map of the preprocessed real-time image of the pulp stack is extracted by dilated convolution. The feature map is then processed by a region proposal network to generate multiple candidate region boxes. Multiple candidate region boxes are input into the trained pulp stack expansion state recognition network for processing, and expansion is recognized for each candidate region box through classification and regression mechanisms. Early warning processing will be carried out based on the external expansion identification results.
[0005] In an optional implementation, the real-time image of the pulp stack is preprocessed, specifically including: The pixel values of the real-time images of the pulp stacks are mapped to a standard range through normalization processing; Multi-scale features of the image are extracted through a feature extraction layer using multi-layer convolution operations, batch normalization, and the ReLU activation function.
[0006] In an optional implementation, feature map extraction is performed on the preprocessed real-time image of the pulp stack using dilated convolution. Specifically, this involves performing the dilated convolution process using the following formula:
[0007] in, For the first Feature map of the layer For convolution kernel parameters, For convolution operations, For bias, This is the activation function.
[0008] In an optional implementation, a region proposal network is used to process the feature map to generate multiple candidate region boxes, specifically including: A region proposal network is used to perform a sliding window operation on the feature map to generate multiple candidate region boxes. The formula for generating the candidate region boxes is as follows:
[0009] in, Candidate region bounding boxes, The center point of the candidate region box, Define the width and height of the candidate region bounding box.
[0010] In an optional implementation, the expansion of each candidate region box is identified through a classification and regression mechanism, specifically including: The classification layer processes each candidate region box using a normalized exponential function and outputs an outward probability distribution. The coordinates of each candidate region box are obtained by using a regression layer and linear transformation. The output layer generates an outward expansion recognition result based on the classification confidence and candidate region bounding box regression results. This includes outputting an outward expansion, the corresponding confidence, and the coordinates of the outward expansion region when the classification confidence exceeds the threshold, and outputting a non-outward expansion region when the classification confidence does not exceed the threshold.
[0011] In one optional implementation, an early warning process is performed based on the external expansion identification result, specifically including: If the identification result is "outward expansion", then an outward expansion warning signal will be triggered; If the identification result is not external expansion, the external expansion warning signal will not be triggered.
[0012] In one optional implementation, training the pulp stack expansion state recognition network specifically includes: Construct an annotated dataset containing multiple candidate region boxes, and annotate the location of the outer expansion region within each candidate region box; Network optimization is performed using loss functions, including cross-entropy loss function for classification loss and smoothing loss function for regression loss. The SGD optimizer is used for optimization, and the model converges by setting an initial learning rate and combining a learning rate decay strategy. During training, the loss value is calculated through forward propagation, and the network weights are updated through backpropagation.
[0013] Secondly, the technical solution of the present invention provides a deep learning-based system for identifying and warning of pulp stack bulging, comprising: A network construction and training module is used to build and train a pulp stack expansion state recognition network; The real-time image acquisition module is used to acquire real-time images of the pulp stacks; The candidate region box extraction module is used to preprocess the real-time image of the pulp stack, extract the feature map of the preprocessed real-time image of the pulp stack through dilated convolution, and use the region proposal network to process the feature map to generate multiple candidate region boxes. The outward expansion recognition module is used to input multiple candidate region boxes into the trained pulp stack outward expansion state recognition network for processing, and to perform outward expansion recognition on each candidate region box through classification and regression mechanisms. The early warning processing module is used to perform early warning processing based on the external expansion identification results.
[0014] Thirdly, the technical solution of the present invention provides a terminal, comprising: The memory is used to store the deep learning-based pulp stack expansion identification and early warning program; A processor is configured to implement the steps of the deep learning-based pulp stack expansion identification and early warning method as described in any of the preceding claims when executing the deep learning-based pulp stack expansion identification and early warning program.
[0015] Fourthly, the present invention provides a computer-readable storage medium storing a deep learning-based pulp stack expansion identification and early warning program. When the deep learning-based pulp stack expansion identification and early warning program is executed by a processor, it implements the steps of the deep learning-based pulp stack expansion identification and early warning method as described in any of the above claims.
[0016] This invention provides a method, system, terminal, and medium for identifying and warning of bulging in pulp stacks based on deep learning. Compared with existing technologies, it has the following advantages: It constructs and trains a network for identifying the bulging state of pulp stacks, and uses this network to process real-time images of the pulp stacks to identify bulging areas and issue warnings. Through multi-scale feature extraction and precise classification and regression mechanisms using a deep learning model, this invention can identify minute morphological changes on the surface of the stack, especially subtle bulging areas that are difficult for humans to detect. Compared to manual observation, which is prone to missing bulging under insufficient light or limited field of view, this method maintains a detection accuracy of over 90%, effectively reducing missed detections and misjudgments. Furthermore, it requires no manual intervention, adapting to the dynamic monitoring needs of high-density cargo stacks and saving human resources. In addition, the early warning function allows the risk of bulging to be detected before collapse, thereby avoiding safety accidents caused by cargo instability. It proactively prevents safety hazards, improves the safety level of port operations, avoids additional economic losses, and ensures the quality of cargo transportation and storage. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of a deep learning-based method for identifying and warning of pulp stack expansion, provided by an embodiment of the present invention.
[0019] Figure 2 This is a schematic diagram of a specific implementation of the deep learning-based method for identifying and warning of pulp stack expansion.
[0020] Figure 3 This is a schematic diagram of the process for processing images of pulp stacks according to the present invention.
[0021] Figure 4 This is a schematic block diagram of a deep learning-based pulp stack expansion identification and early warning system provided in an embodiment of the present invention.
[0022] Figure 5 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation
[0023] To enable those skilled in the art to better understand the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
[0025] To address the problems of low accuracy, poor real-time performance, insufficient efficiency, and safety hazards caused by reliance on manual observation in existing technologies, this invention provides a deep learning-based method for identifying and warning of external expansion in pulp stacks. This method uses a pulp stack external expansion status recognition network to process real-time images of the pulp stacks, enabling the identification and warning of expansion areas. It overcomes the technical shortcomings of manual inspection, such as difficulty in detecting subtle external expansions, inability to provide timely warnings, and delayed emergency response. Simultaneously, it significantly improves the accuracy, response speed, and automation level of stack status monitoring, thereby enabling early detection and prevention of stack external expansion risks, effectively ensuring operational safety and stable cargo stacking, reducing labor costs, and improving the overall efficiency of port loading and unloading operations.
[0026] Figure 1 This is a schematic flowchart of a deep learning-based method for identifying and warning of pulp stack bulging, provided in an embodiment of the present invention. Figure 1 The executing entity can be a deep learning-based pulp stack expansion identification and early warning system. The deep learning-based pulp stack expansion identification and early warning method provided in this embodiment of the invention is executed by a computer device; correspondingly, the deep learning-based pulp stack expansion identification and early warning system runs on the computer device. Depending on different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.
[0027] like Figure 1 As shown, the method includes the following steps.
[0028] S1. Construct and train a network for recognizing the outward expansion state of pulp stacks.
[0029] S2, acquire real-time images of the pulp stack.
[0030] S3 preprocesses the real-time image of the pulp stack and extracts feature maps from the preprocessed real-time image of the pulp stack through dilated convolution. A region proposal network is then used to process the feature maps to generate multiple candidate region boxes.
[0031] S4 inputs multiple candidate region boxes into the trained pulp stack expansion state recognition network for processing, and performs expansion recognition on each candidate region box through classification and regression mechanisms.
[0032] S5 performs early warning processing based on the external expansion identification results.
[0033] This embodiment first trains the pulp stack expansion state recognition network to more accurately identify expansion, specifically including the following steps.
[0034] S001, construct an annotated dataset containing multiple candidate region boxes, and annotate the location of the outer expansion region within each candidate region box.
[0035] S002, network optimization is performed through loss functions, including cross-entropy loss function for classification loss and smoothing loss function for regression loss.
[0036] S003 uses the SGD optimizer for optimization, and converges the model by setting an initial learning rate and combining it with a learning rate decay strategy.
[0037] S004, during training, the loss value is calculated through forward propagation, and the network weights are updated through backpropagation.
[0038] In this embodiment, during the training phase, a labeled dataset containing multiple candidate bounding boxes is first prepared, clearly indicating the locations of the bulging regions. Next, a loss function guides network optimization. The classification loss uses the cross-entropy loss function to measure the model's performance in classifying bulging and non-bulging regions, while the regression loss uses a smoothing loss function to optimize the matching accuracy between candidate boxes and actual bulging regions. During optimization, the SGD optimizer is used, gradually improving the model's convergence efficiency by setting an initial learning rate (exemplarily 0.001) and combining it with a learning rate decay strategy. The training process includes forward propagation to calculate the loss value, backpropagation to update network weights, and saving model checkpoints after every few training steps to ensure the stability and recoverability of the training process. Through multiple rounds of training iterations, the network gradually learns the ability to accurately identify the bulging features of pulp stacks, laying a solid foundation for subsequent testing.
[0039] In the testing phase, the image of the pulp stack to be tested is processed by a feature extraction layer to extract multi-scale image features, and a region proposal network generates candidate boxes. Subsequently, a classification layer predicts the probability of bulging and non-bulging in the candidate regions, and a regression layer further optimizes the position of the candidate boxes to obtain more accurate bulging regions. Finally, combining the confidence score of the detection results with a preset threshold (exemplarily set to 0.7), when the confidence score exceeds the threshold, the system outputs the coordinates of the bulging region and its confidence score, and triggers a corresponding early warning signal, thereby realizing intelligent monitoring of the bulging state of the pulp stack and early prevention of safety hazards.
[0040] In this embodiment, after training the pulp stack expansion state recognition network, the trained pulp stack expansion state recognition network is used to process multiple candidate region boxes of the real-time acquired pulp stack image to identify the expansion region and issue an early warning. Figure 2 This is a schematic diagram of a specific implementation of the deep learning-based method for identifying and warning of pulp stack bulging. First, a high-definition camera is used to capture images. Then, image preprocessing is achieved through normalization and noise reduction. After that, feature extraction is performed, and the features are detected to identify stack bulging. When bulging is present, an early warning is triggered. The detection results can be stored for subsequent optimization analysis and training.
[0041] In the above specific implementation, in terms of hardware configuration, a high-definition industrial camera is used to acquire real-time images of the pulp stack, combined with a high-performance edge computing device for deep learning model inference, and an audible and visual alarm device is used to realize the early warning function. The software process starts with image acquisition and preprocessing, normalizing, denoising, and resizing the images. Then, the preprocessed images are used to accurately locate the bulging areas through multi-scale feature extraction and a region proposal network, and the coordinates and confidence scores of the bulging areas are output. If the detection result exceeds the set threshold of 0.7, the system triggers an audible and visual alarm and displays the abnormal information intuitively on the monitoring interface, while storing the results in the database for subsequent analysis. Through all-weather real-time monitoring and accurate early warning, early identification and risk control of pulp stack bulging problems are achieved, significantly improving the safety and efficiency of port operations, and the robustness and adaptability of the model are continuously improved through long-term data collection and retraining.
[0042] Figure 3 This is a flowchart illustrating the process of processing images of pulp stacks, specifically including the following steps.
[0043] S301 performs preprocessing on real-time images of pulp stacks.
[0044] In this embodiment, the image preprocessing includes normalizing the pixel values of the real-time image of the pulp stack to a standard range, and then using a feature extraction layer to extract multi-scale features of the image through multi-layer convolution operations, batch normalization, and ReLU activation function.
[0045] It should be noted that this embodiment uses real-time images of the pallet in RGB format as input and preprocesses the input RGB format images.
[0046] S302 extracts feature maps from the pre-processed real-time image of the pulp stack using dilated convolution.
[0047] This embodiment uses dilated convolution to capture subtle morphological changes on the surface of the stack. The expression for the dilated convolution process is as follows:
[0048] in, For the first Feature map of the layer For convolution kernel parameters, For convolution operations, For bias, This is the activation function.
[0049] S303 uses a region proposal network to process the feature map and generate multiple candidate region boxes.
[0050] This embodiment uses a region proposal network to perform a sliding window operation on the feature map, generating multiple candidate region boxes. The formula for generating the candidate region boxes is as follows:
[0051] in, Candidate region bounding boxes, The center point of the candidate region box, Define the width and height of the candidate region bounding box.
[0052] S304 identifies the outward expansion of each candidate region box through classification and regression mechanisms.
[0053] S304.1 uses a classification layer to process each candidate region box using a normalized exponential function and outputs an outward probability distribution.
[0054] S304.2, use a regression layer to obtain the coordinate positions of each candidate region box through linear transformation.
[0055] S304.3 The output layer generates an outward expansion recognition result based on the classification confidence and candidate region bounding box regression results. This includes outputting an outward expansion, the corresponding confidence, and the coordinates of the outward expansion region when the classification confidence exceeds the threshold, and outputting a non-outward expansion region when the classification confidence does not exceed the threshold.
[0056] This embodiment uses classification and regression layers to classify and adjust the frame of candidate regions for outward expansion. The classification layer outputs the probability distribution of outward expansion using a normalized exponential function, while the regression layer predicts the precise position of the frame using a linear transformation. Finally, the output layer generates the final detection result (outward expansion or non-outward expansion) based on the classification confidence and frame regression results. Marking outward expansion areas of the pallet triggers a corresponding early warning signal. This architecture achieves efficient detection and intelligent monitoring of pallet outward expansion through multi-layered feature extraction and accurate classification and regression.
[0057] This embodiment performs early warning processing based on the external expansion identification result, including triggering an external expansion early warning signal if the identification result is external expansion, and not triggering an external expansion early warning signal if the identification result is not external expansion.
[0058] The foregoing has described in detail an embodiment of a deep learning-based method for identifying and warning of pulp stack bulging. Based on the deep learning-based method for identifying and warning of pulp stack bulging described in the above embodiment, this invention also provides a deep learning-based system for identifying and warning of pulp stack bulging, corresponding to the method.
[0059] Figure 4 This is a schematic block diagram of a deep learning-based pulp stack expansion identification and early warning system provided in an embodiment of the present invention. In this embodiment, the deep learning-based pulp stack expansion identification and early warning system 400 can be divided into multiple functional modules according to its functions, such as... Figure 4 As shown. The functional modules may include: a recognition network construction and training module 410, a real-time image acquisition module 420, a candidate region box extraction module 430, an extrusion recognition module 440, and an early warning processing module 450. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and which are stored in memory.
[0060] The recognition network construction training module 410 is used to construct and train the pulp stack expansion state recognition network.
[0061] The real-time image acquisition module 420 is used to acquire real-time images of pulp stacks.
[0062] The candidate region box extraction module 430 is used to preprocess the real-time image of the pulp stack, extract feature maps from the preprocessed real-time image of the pulp stack through dilated convolution, and generate multiple candidate region boxes by processing the feature maps using a region proposal network.
[0063] The outward expansion recognition module 440 is used to input multiple candidate region boxes into the trained pulp stack outward expansion state recognition network for processing, and to perform outward expansion recognition on each candidate region box through classification and regression mechanisms.
[0064] The early warning processing module 450 is used to perform early warning processing based on the external expansion identification results.
[0065] The deep learning-based pulp stack expansion identification and early warning system of this embodiment is used to implement the aforementioned deep learning-based pulp stack expansion identification and early warning method. Therefore, the specific implementation of this system can be found in the embodiment section of the deep learning-based pulp stack expansion identification and early warning method above. Thus, its specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.
[0066] Furthermore, since the deep learning-based pulp stack expansion identification and early warning system of this embodiment is used to implement the aforementioned deep learning-based pulp stack expansion identification and early warning method, its function corresponds to the function of the above method, and will not be repeated here.
[0067] Figure 5 A schematic diagram of a terminal 500 provided in an embodiment of the present invention includes: a processor 510, a memory 520, and a communication unit 530. The processor 510 is used to implement the following steps when executing the deep learning-based pulp stack expansion identification and early warning program stored in the memory 520: Construct and train a network for recognizing the external expansion state of pulp stacks; Acquire real-time images of pulp stacks; The real-time image of the pulp stack is preprocessed, and the feature map of the preprocessed real-time image of the pulp stack is extracted by dilated convolution. The feature map is then processed by a region proposal network to generate multiple candidate region boxes. Multiple candidate region boxes are input into the trained pulp stack expansion state recognition network for processing, and expansion is recognized for each candidate region box through classification and regression mechanisms. Early warning processing will be carried out based on the external expansion identification results.
[0068] The terminal 500 includes a processor 510, a memory 520, and a communication unit 530. These components communicate via one or more buses. Those skilled in the art will understand that the server structure shown in the figure does not constitute a limitation of the present invention. It can be a bus topology or a star topology, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0069] The memory 520 can be used to store the execution instructions of the processor 510. The memory 520 can be implemented by any type of volatile or non-volatile memory terminal or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory 520 are executed by the processor 510, the terminal 500 is able to perform some or all of the steps in the above method embodiments.
[0070] The processor 510 serves as the control center of the storage terminal, connecting various parts of the electronic terminal via various interfaces and lines. It executes software programs and / or modules stored in the memory 520, and calls data stored in the memory to perform various functions of the electronic terminal and / or process data. The processor can be composed of integrated circuits (ICs), such as a single packaged IC or multiple packaged ICs with the same or different functions connected together. For example, the processor 510 may consist only of a central processing unit (CPU). In this embodiment of the invention, the CPU may have a single processing core or include multiple processing cores.
[0071] The communication unit 530 is used to establish a communication channel, enabling the storage terminal to communicate with other terminals. It can receive user data sent by other terminals or send user data to other terminals.
[0072] The present invention also provides a computer storage medium, which may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.
[0073] The computer storage medium stores a deep learning-based pulp stack expansion identification and early warning program. When the deep learning-based pulp stack expansion identification and early warning program is executed by the processor, it performs the following steps: Construct and train a network for recognizing the external expansion state of pulp stacks; Acquire real-time images of pulp stacks; The real-time image of the pulp stack is preprocessed, and the feature map of the preprocessed real-time image of the pulp stack is extracted by dilated convolution. The feature map is then processed by a region proposal network to generate multiple candidate region boxes. Multiple candidate region boxes are input into the trained pulp stack expansion state recognition network for processing, and expansion is recognized for each candidate region box through classification and regression mechanisms. Early warning processing will be carried out based on the external expansion identification results. Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or other media capable of storing program code. It includes several instructions to cause a computer terminal (which may be a personal computer, server, or a second terminal, network terminal, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.
[0074] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0075] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0076] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0077] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.
Claims
1. A method for identifying and warning of bulging in pulp stacks based on deep learning, characterized in that, Includes the following steps: Construct and train a network for recognizing the external expansion state of pulp stacks; Acquire real-time images of pulp stacks; The real-time image of the pulp stack is preprocessed, and the feature map of the preprocessed real-time image of the pulp stack is extracted by dilated convolution. The feature map is then processed by a region proposal network to generate multiple candidate region boxes. Multiple candidate region boxes are input into the trained pulp stack expansion state recognition network for processing, and expansion is recognized for each candidate region box through classification and regression mechanisms. Early warning processing will be carried out based on the external expansion identification results.
2. The method for identifying and warning of pulp stack bulging based on deep learning according to claim 1, characterized in that, Preprocessing of real-time images of pulp stacks includes: The pixel values of the real-time images of the pulp stacks are mapped to a standard range through normalization processing; Multi-scale features of the image are extracted through a feature extraction layer using multi-layer convolution operations, batch normalization, and the ReLU activation function.
3. The method for identifying and warning of pulp stack bulging based on deep learning according to claim 2, characterized in that, Feature map extraction is performed on the preprocessed real-time image of the pulp stack using dilated convolution. Specifically, the dilated convolution process is executed using the following formula. in, For the first Feature map of the layer For convolution kernel parameters, For convolution operations, For bias, This is the activation function.
4. The method for identifying and warning of pulp stack bulging based on deep learning according to claim 3, characterized in that, The feature map is processed using a region proposal network to generate multiple candidate region boxes, specifically including: A region proposal network is used to perform a sliding window operation on the feature map to generate multiple candidate region boxes. The formula for generating the candidate region boxes is as follows: in, Candidate region bounding boxes, The center point of the candidate region box, Define the width and height of the candidate region bounding box.
5. The method for identifying and warning of pulp stack bulging based on deep learning according to claim 4, characterized in that, The expansion of each candidate region box is identified through classification and regression mechanisms, specifically including: The classification layer processes each candidate region box using a normalized exponential function and outputs an outward probability distribution. The coordinates of each candidate region box are obtained by using a regression layer and linear transformation. The output layer generates an outward expansion recognition result based on the classification confidence and candidate region bounding box regression results. This includes outputting an outward expansion, the corresponding confidence, and the coordinates of the outward expansion region when the classification confidence exceeds the threshold, and outputting a non-outward expansion region when the classification confidence does not exceed the threshold.
6. The method for identifying and warning of pulp stack bulging based on deep learning according to claim 5, characterized in that, Based on the external expansion identification results, early warning processing is carried out, specifically including: If the identification result is "outward expansion", then an outward expansion warning signal will be triggered; If the identification result is not external expansion, the external expansion warning signal will not be triggered.
7. The method for identifying and warning of pulp stack bulging based on deep learning according to claim 5, characterized in that, Training the pulp stack expansion state recognition network specifically includes: Construct an annotated dataset containing multiple candidate region boxes, and annotate the location of the outer expansion region within each candidate region box; Network optimization is performed using loss functions, including cross-entropy loss function for classification loss and smoothing loss function for regression loss. The SGD optimizer is used for optimization, and the model converges by setting an initial learning rate and combining a learning rate decay strategy. During training, the loss value is calculated through forward propagation, and the network weights are updated through backpropagation.
8. A deep learning-based system for identifying and warning of bulging in pulp stacks, characterized in that, include: A network construction and training module is used to build and train a pulp stack expansion state recognition network; The real-time image acquisition module is used to acquire real-time images of the pulp stacks; The candidate region box extraction module is used to preprocess the real-time image of the pulp stack, extract the feature map of the preprocessed real-time image of the pulp stack through dilated convolution, and use the region proposal network to process the feature map to generate multiple candidate region boxes. The outward expansion recognition module is used to input multiple candidate region boxes into the trained pulp stack outward expansion state recognition network for processing, and to perform outward expansion recognition on each candidate region box through classification and regression mechanisms. The early warning processing module is used to perform early warning processing based on the external expansion identification results.
9. A terminal, characterized in that, include: The memory is used to store the deep learning-based pulp stack expansion identification and early warning program; The processor is configured to implement the steps of the deep learning-based pulp stack expansion identification and early warning method as described in any one of claims 1-7 when executing the deep learning-based pulp stack expansion identification and early warning program.
10. A computer-readable storage medium, characterized in that, The readable storage medium stores a deep learning-based pulp stack expansion identification and early warning program, which, when executed by a processor, implements the steps of the deep learning-based pulp stack expansion identification and early warning method as described in any one of claims 1-7.