Method, system and device for detecting and identifying damaged tray

Abstract

The invention discloses a damaged tray detection and identification method, system and device, belongs to the technical field of logistics automation, and solves the problems of function splitting, poor low-temperature adaptability, insufficient detection precision and low scheduling linkage efficiency in the prior art. According to the method, a tray stereo image, a three-dimensional point cloud and pressure distribution multi-dimensional data are collected through a low-temperature-resistant sensing assembly, after preprocessing and space-time alignment, tray identification and positioning are achieved by adopting an improved three-dimensional convolutional neural network, damage detection and grade judgment are completed through a multi-feature fusion strategy, and the damage detection and grade judgment accuracy is improved. And the feedback scheduling system realizes damage avoidance and task adjustment. The invention also discloses a corresponding detection system and device, which integrate low-temperature-resistant perception, data processing, identification detection and scheduling linkage modules and adapt to low-temperature scenes such as cold chains. According to the invention, tray identification, positioning and damage detection are integrated, the environment adaptability is high, the detection precision is high, linkage is efficient, safe and continuous logistics operation can be guaranteed, the deployment is flexible, and the popularization and application value is high.

Description

Technical Field

[0001] This invention belongs to the field of logistics automation technology, and in particular relates to a method, system and device for detecting and identifying damaged pallets. Background Technology

[0002] In automated logistics systems, pallets serve as core load-bearing and transfer components, and their structural integrity directly impacts cargo transfer safety, the operational stability of automated equipment, and overall operational efficiency. In low-temperature environments such as cold chain logistics, pallets are subjected to prolonged exposure to low temperatures, repeated impacts from handling, and dynamic loads, making them prone to structural defects such as cracking, deformation, and forkhole damage. If damaged pallets are not detected, identified, and avoided in a timely manner, they can not only lead to safety accidents such as cargo tipping and breakage but also cause automated equipment (such as AGVs) to jam, experience operational interruptions, or even malfunction, severely restricting the continuity and reliability of logistics operations.

[0003] Existing pallet detection and recognition technologies suffer from numerous technical shortcomings, making it difficult to meet the demands of automated operations in complex low-temperature environments: First, existing technologies are mostly single-type pallet identification, spatial positioning, or damage detection solutions, relying on manual inspection of damaged pallets, resulting in high labor costs, low inspection efficiency, and high false negative rates; Second, most damage detection solutions are based solely on single visual inspection technologies, and in complex environments such as low temperatures, fog, and low light, image acquisition quality is easily affected, leading to reduced damage detection accuracy and an inability to comprehensively identify various structural defects; Third, operational coordination is insufficient. Even if a damaged pallet is detected, it is difficult to quickly generate avoidance instructions and achieve efficient coordination with the scheduling system to dynamically adjust the task. Damaged pallets still interfere with the operational process, failing to guarantee continuous and stable operation in low-temperature scenarios.

[0004] The information disclosed in this background section is included only to enhance the understanding of the context of this disclosure, and therefore may contain information that does not constitute relevant technology currently known to those skilled in the art. Summary of the Invention

[0005] This application provides a method, system, and device for detecting and identifying damaged pallets, which solves the technical problems in the prior art such as fragmented pallet detection and identification functions, poor adaptability to complex low-temperature environments, insufficient detection accuracy, and inefficient linkage with automated scheduling systems, resulting in limited operational safety, continuity, and efficiency.

[0006] The technical solution adopted in this application is as follows: In a first aspect, this application provides a method for detecting and identifying damaged pallets, comprising the following steps: S1. Collect multi-dimensional raw data of the pallet area. The multi-dimensional raw data includes at least pallet stereoscopic image data, three-dimensional point cloud data, and pressure distribution data during the picking and placing of goods. S2. Preprocess the multi-dimensional raw data. The preprocessing includes image dehazing, point cloud noise reduction, and pressure data calibration. At the same time, the time synchronization and spatial coordinate alignment of the stereo image data and the three-dimensional point cloud data are completed. S3. Based on the preprocessed multi-dimensional data, an improved three-dimensional convolutional neural network algorithm is used to realize pallet type recognition and spatial posture positioning. S4. A multi-feature fusion detection strategy is adopted to detect the damage status of the pallet. The multi-feature fusion detection strategy includes edge discontinuity detection based on the image, geometric anomaly detection based on the point cloud, and uneven force detection based on the pressure distribution. A comprehensive damage score is calculated through a weighted voting mechanism, and the pallet damage level is determined according to a preset scoring threshold. S5. Feed back the pallet recognition result, spatial attitude positioning information and damage detection result to the scheduling system. If a damaged pallet is detected, generate an avoidance command and control the actuator to avoid the damaged pallet. The scheduling system will then reassign the job task. If the detected pallet is not damaged, continue to execute the job task assigned by the scheduling system.

[0007] Furthermore, the multi-dimensional raw data mentioned in step S1 is collected by a low-temperature sensing component, which includes a low-temperature binocular imaging device with anti-fogging and defrosting functions, a low-temperature and anti-interference laser detection device, and a low-temperature stable pressure sensing element installed at the end of the actuator fork arm.

[0008] Furthermore, the improved 3D convolutional neural network algorithm described in step S3 is based on a lightweight network architecture, introduces an attention mechanism, and achieves rapid identification of tray types and accurate positioning of spatial posture by optimizing network parameters and training strategies.

[0009] Furthermore, the multi-feature fusion detection strategy described in step S4 includes: The length and number of discontinuities at the edge of the tray in the preprocessed image are extracted as visual damage features. The flatness deviation and fork hole size deviation of the pallet were extracted from the preprocessed point cloud data as geometric damage features; The standard deviation of the pressure distribution data after calibration was extracted as a mechanical damage characteristic. The weight coefficients of each feature are determined by the analytic hierarchy process (AHP). Based on the weight coefficients and the values ​​of each feature, a comprehensive damage score is calculated. The pallets are then classified into three levels—no damage, minor damage, and severe damage—according to a preset scoring threshold.

[0010] Secondly, this application provides a damaged pallet detection and identification system, which includes at least a low-temperature sensing module, a data processing module, an identification and damage detection module, a scheduling and linkage module, and a power management module; communication connections are established between the modules to realize data interaction and command transmission, and to adapt to low-temperature operating environments; The low-temperature sensing module is used to collect multi-dimensional raw data of the pallet area. It includes a low-temperature binocular imaging device, a laser detection device, and a pressure sensing element. The low-temperature binocular imaging device is used to collect three-dimensional image data of the pallet, the laser detection device is used to collect three-dimensional point cloud data of the pallet, and the pressure sensing element is installed at the end of the fork arm of the mobile operation platform to collect pressure distribution data during the picking and placing of goods. The data processing module is connected to the low-temperature sensing module and is used to perform the data preprocessing and data synchronization alignment operations in step S2 of the first aspect above. The identification and damage detection module is connected to the data processing module and includes a pallet identification unit and a damage detection unit; wherein, the pallet identification unit is used to perform the pallet type identification and spatial attitude positioning operation in step S3 of the first aspect above, and the damage detection unit is used to perform the damage status detection and level determination operation in step S4 of the first aspect above. The scheduling and linkage module is connected to the identification and damage detection module and the scheduling system respectively, and is used to perform the result feedback, avoidance instruction generation and task redistribution linkage operation in step S5 of the first aspect above. The power management module is used to provide stable power to each module. It is equipped with a low-temperature adaptive energy storage unit and a power heating component to ensure the stability of power supply in low-temperature environments.

[0011] The damaged pallet detection and identification system also includes an edge computing module, which is integrated locally on the mobile operation platform and used to deploy data processing algorithms, pallet recognition algorithms, and damage detection algorithms. The edge computing module establishes a communication connection with a remote server to realize statistical analysis of damaged pallet data and iterative updates of algorithm models.

[0012] Furthermore, the low-temperature resistant binocular imaging device has anti-fogging and defrosting functions, and adopts low-temperature adapted optical components and image sensing elements; the laser detection device has low-temperature resistance and anti-interference characteristics; the pressure sensing element adopts a low-temperature stable sensing structure; and data transmission between modules is achieved through a low-temperature resistant communication interface.

[0013] Furthermore, the power management module also has overload protection functions to ensure power supply safety; wherein, the low-temperature adaptive energy storage unit has stable discharge capability in the target low-temperature operating environment.

[0014] Furthermore, the scheduling linkage module communicates with the scheduling system through a wireless low-temperature resistant communication module. When a severely damaged pallet is detected, it generates an emergency avoidance command and triggers an alarm signal.

[0015] Thirdly, this application provides a damaged pallet detection and identification device, including a supporting body and a damaged pallet detection and identification system as described in any of the second aspects above integrated on the supporting body; the supporting body is a mobile work platform adapted for low-temperature operations, the pressure sensing element in the low-temperature sensing module is integrated at the end of the fork arm of the mobile work platform, and the edge computing module, data processing module and power management module are integrated in the control cabin of the mobile work platform to realize integrated operation of pallet identification, damage detection and operation linkage.

[0016] Fourthly, this application provides a damaged pallet detection and identification device, including a memory and a processor, wherein the memory is used to store computer programs or instructions; when the computer programs or instructions are executed by the processor, the method described in any one of the first aspects is implemented.

[0017] Fifthly, this application provides a computer-readable storage medium storing a computer program or instructions that, when executed by a processor, implement the method described in any one of the first aspects above.

[0018] Sixthly, this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the method described in any one of the first aspects above.

[0019] Compared with the prior art, this application has the following significant advantages:

[0020] 1. Integrated Functionality Significantly Improves Operational Efficiency. This invention integrates pallet type recognition, spatial posture positioning, and damage detection functions, overcoming the limitations of fragmented functions in existing technologies. It eliminates the need for manual inspection of damaged pallets, significantly reducing labor costs. Simultaneously, it enables the simultaneous execution of detection and pallet transfer operations, shortening the time required for single-pallet detection and identification without affecting the automated operation rhythm, resulting in a significant improvement in operational efficiency.

[0021] 2. Strong environmental adaptability and high detection accuracy. Through the synergistic effect of low-temperature resistant sensing components and multi-feature fusion detection strategies, it effectively resists environmental interference such as low temperature, fog, and low light, avoiding the limitations of single detection technologies; the accuracy of pallet recognition, spatial positioning accuracy, and damage detection accuracy are significantly improved, and the rate of missed detection of severe damage is significantly reduced. It can accurately identify various structural defects such as cracks, deformation, and fork hole damage, reducing the risk of missed detection and false detection.

[0022] 3. Highly efficient linkage response ensures safe and continuous operation: The scheduling linkage module coordinates rapidly with the scheduling system and the execution mechanism with low communication latency. It generates differentiated avoidance instructions and task redistribution schemes for different damage levels, and triggers alarm signals at the same time to avoid safety hazards such as cargo tipping and equipment jamming from the source, thus ensuring the continuity of operation.

[0023] 4. Excellent low-temperature stability and wide applicability. The core components (low-temperature sensing components, low-temperature adaptable energy storage units, and power supply heating components) are all designed for low-temperature resistance. With pre-processing operations such as image defogging and pressure data temperature compensation, they can work stably in low-temperature environments [-30℃, 0℃], as well as in fog, frost, and low-light environments, avoiding performance degradation or failure of core components. They are suitable for various low-temperature operation scenarios such as cold chain logistics and low-temperature warehousing.

[0024] 5. Flexible and reusable deployment with high promotional value. The described damaged pallet detection and identification method can be quickly deployed and reused through electronic devices, storage media, and computer program products; the system and device can be integrated into various mobile operation platforms, the edge computing module supports iterative updates of algorithm models, and it is adaptable to various automated operation scenarios such as cold chain logistics and low-temperature warehousing. It has good versatility and scalability, and can be widely used in various logistics automation operation scenarios, possessing good promotional and application value.

[0025] The beneficial effects of aspects two through five above can be referenced to aspect one or any possible implementation thereof, and will not be elaborated upon here. Based on the implementations provided in the above aspects, this application can also be further combined to provide more implementations.

[0026] Other advantages, objectives and features of this application will be partly apparent from the description below, and partly understood by those skilled in the art through study and practice of this application. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating the damaged pallet detection and identification method provided in this application embodiment; Figure 2 This is a schematic diagram of the improved three-dimensional convolutional neural network algorithm structure described in step S3 of the embodiments of this application; Figure 3This is a flowchart illustrating the multi-feature fusion detection strategy described in step S4 of the embodiments of this application; Figure 4 This is a schematic diagram of the structure of the damaged pallet detection and identification system provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the damaged pallet detection and identification device provided in the embodiments of this application. Detailed Implementation

[0029] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0030] The term "and / or" as used in this application includes any and all combinations of one or more of the associated listed items. In this application, "at least one" means one or more, and "more than one" means two or more. The terms "first," "second," and other ordinal terms used in this application may be used to describe various constituent elements, but these constituent elements are not limited by these terms. The purpose of using these terms is solely to distinguish one constituent element from others and should not be construed as indicating or implying relative importance. For example, without departing from the scope of this application, a first constituent element may be named a second constituent element, and similarly, a second constituent element may be named a first constituent element.

[0031] Before introducing the embodiments of this application, the technical adaptation (operational scenario adaptation) and deployment (including hardware deployment and software deployment) involved in this application will be introduced first.

[0032] Operational scenario adaptation: Based on the actual needs of the target operational scenario (e.g., cold storage, low-temperature storage center), confirm the type and specifications of pallets in the operational area (e.g., the size and structure of plastic pallets and wooden pallets), transfer path parameters (e.g., driving route, pick-up and drop-off points), and communication protocols of the remote scheduling system (e.g., TCP / IP, MODBUS); complete the path planning and positioning calibration of the carrier (low-temperature adapted AGV) (through LiDAR positioning technology) to ensure that the driving accuracy of the carrier meets the requirements of pick-up and drop-off and detection and identification, and ensure the collaborative adaptation of the operation process with the detection and identification operation of this application.

[0033] Hardware deployment: In accordance with the core requirements of the technical solution of this application, complete the installation and deployment of each functional module and hardware device, including the deployment of low temperature sensing components, core modules, communication and power supply.

[0034] (1) Deployment of low temperature sensing components: The low temperature resistant binocular imaging device is installed at the front of the main vehicle through a low temperature adapter bracket. The bracket has shock absorption and anti-fogging functions, and the imaging field of view covers the pallet recognition area (within 0.5-3m of the front of the main vehicle), ensuring that the collected three-dimensional images of the pallet are clear and complete. The laser detection device is deployed in coordination with the binocular imaging device, and the installation height is consistent with that of the binocular imaging device to ensure the spatial compatibility of the three-dimensional point cloud data and image data. The pressure sensing element is integrated into the bearing surface at the end of the fork arm of the main vehicle through an embedded installation method. It adopts a sealed protective structure to prevent low temperature and water vapor from affecting the sensing accuracy, and is used to collect pressure distribution data in real time during the loading and unloading process.

[0035] (2) Core module deployment: The data processing module, edge computing module, and power management module are integrated into the control cabin of the carrier body. The control cabin is equipped with a cold-proof and heat-insulating layer and a small cooling fan is installed inside to balance the low-temperature protection and equipment heat dissipation requirements. The scheduling and linkage module is integrated with the control system of the carrier body and realizes command interaction through a standardized interface. The alarm device (audio-visual alarm) is installed on the top of the carrier body to facilitate on-site personnel to quickly detect abnormalities.

[0036] (3) Communication and power supply deployment: Each module is connected by a low-temperature resistant cable (with an outer layer wrapped in a cold-proof and anti-interference sheath), and data transmission is achieved using a low-temperature resistant communication interface (RS485, Ethernet); the power management module uses 24V DC power supply to uniformly distribute power to each module, and the low-temperature adaptive energy storage unit and power heating component are integrated inside the power management module to ensure stable power supply in a low-temperature environment of [-30℃, 0°].

[0037] Software and Algorithm Deployment: The computer program and algorithm configuration file corresponding to the damaged pallet detection and identification method described in this application are imported into an electronic device (or edge computing module) via a computer-readable storage medium to complete the software installation and adaptation. Specific deployment details are as follows:

[0038] (1) Algorithm deployment: The core algorithms, such as data acquisition and preprocessing algorithm, improved three-dimensional convolutional neural network algorithm (including lightweight network architecture, attention mechanism, optimized network parameters and training strategy implementation code), multi-feature fusion detection algorithm (including weighted voting mechanism, analytic hierarchy process weight calculation implementation code), and scheduling linkage control algorithm, are deployed to the edge computing module and data processing module to achieve local fast computing and response; the edge computing module establishes a communication connection with the remote server to synchronize damaged pallet data and iteratively update the algorithm model.

[0039] (2) Model training and calibration: The training dataset covers pallet sample data of different types (plastic, wood) and different damage states (no damage, slight damage, severe damage). The sample data was collected in a low temperature environment of [-30℃, 0°] and includes stereo image data, three-dimensional point cloud data, and pressure distribution data. The training process follows the preset optimized network parameters and training strategy. Through multiple iterations of training and verification, the optimal algorithm model is saved. After deployment, according to the actual situation of the target operation scenario, parameters such as feature extraction threshold and comprehensive damage score threshold are calibrated to ensure the accuracy and stability of the algorithm operation.

[0040] refer to Figure 1 , Figure 1 This is a flowchart illustrating the damaged pallet detection and identification method provided in an embodiment of this application. Figure 1 As shown, the damaged pallet detection and identification method includes: S1. Collect multi-dimensional raw data of the pallet area. The multi-dimensional raw data includes at least pallet stereoscopic image data, three-dimensional point cloud data, and pressure distribution data during the picking and placing of goods. S2. Preprocess the multi-dimensional raw data. The preprocessing includes image dehazing, point cloud noise reduction, and pressure data calibration. At the same time, the time synchronization and spatial coordinate alignment of the stereo image data and the three-dimensional point cloud data are completed. S3. Based on the preprocessed multi-dimensional data, an improved three-dimensional convolutional neural network algorithm is used to realize pallet type recognition and spatial posture positioning. S4. A multi-feature fusion detection strategy is adopted to detect the damage status of the pallet. The multi-feature fusion detection strategy includes edge discontinuity detection based on the image, geometric anomaly detection based on the point cloud, and uneven force detection based on the pressure distribution. A comprehensive damage score is calculated through a weighted voting mechanism, and the pallet damage level is determined according to a preset scoring threshold. S5. Feed back the pallet recognition result, spatial attitude positioning information and damage detection result to the scheduling system. If a damaged pallet is detected, generate an avoidance command and control the actuator to avoid the damaged pallet. The scheduling system will then reassign the job task. If the detected pallet is not damaged, continue to execute the job task assigned by the scheduling system.

[0041] Specifically, in step S1, the AGV (carrying body) travels according to the instructions of the remote scheduling system. After confirming that it has entered the pallet recognition area through its own LiDAR positioning module, it automatically triggers the activation of the low-temperature sensing component. The low-temperature binocular imaging device immediately activates the anti-fog and defrosting function to eliminate the interference of low-temperature fog and frost on the imaging, and continuously collects 3D image data of the pallet (image resolution adapted to recognition requirements, and acquisition frequency matched with the AGV's travel speed); the laser detection device is activated simultaneously to collect 3D point cloud data of the pallet, and at the same time, initially removes environmental noise such as ground and surrounding debris, and outputs effective point cloud data; when the AGV fork arm moves to the appropriate height above the pallet (10-20cm above the pallet surface) to prepare for picking up or placing goods, the pressure sensing element is activated simultaneously to collect pressure distribution data in real time during the picking and placing process and record the pressure change curve. All three types of data are transmitted in real time to the AGV's local data processing module through the low-temperature communication interface.

[0042] Specifically, in step S2, after receiving the multi-dimensional raw data, the data processing module immediately performs preprocessing operations. For the stereo image data, a dark channel prior algorithm is used for dehazing, combined with a histogram equalization algorithm to enhance the contrast of image details, and finally, a Gaussian filtering algorithm is used to eliminate image noise, outputting a clear tray image. For the three-dimensional point cloud data, a combination of direct-pass filtering and statistical filtering algorithms is used for noise reduction, removing abnormal isolated points and environmental noise points, and retaining the effective point cloud corresponding to the tray outline and structural features. For the pressure distribution data, a temperature compensation algorithm is used to eliminate the influence of low temperature environment on sensing accuracy, combined with a linear calibration algorithm to correct the system error of the sensing element, outputting accurate pressure data. After preprocessing, based on a timestamp synchronization algorithm, the acquisition time of the three types of data is precisely aligned; through a preset extrinsic parameter matrix (obtained in advance through calibration experiments), the spatial coordinates of the stereo image data and the three-dimensional point cloud data are aligned, mapping the two-dimensional image coordinates and the three-dimensional spatial coordinates to the same coordinate system, outputting standardized multi-dimensional data, which is then transmitted to the recognition and damage detection module.

[0043] Specifically, in step S3, the pallet recognition unit of the identification and damage detection module receives the preprocessed multi-dimensional data and calls the improved three-dimensional convolutional neural network algorithm to perform the recognition and positioning operation. The three-dimensional convolutional neural network algorithm is a deep learning algorithm based on three-dimensional convolution operations, which extracts and learns the three-dimensional features after fusing the pallet stereoscopic image data and three-dimensional point cloud data, so as to achieve accurate recognition and positioning of pallet type and spatial posture.

[0044] In some embodiments, reference Figure 2 , Figure 2The diagram below shows the structure of the improved three-dimensional convolutional neural network algorithm described in step S3. The improved three-dimensional convolutional neural network algorithm described in step S3 includes a data input layer 201, a feature extraction layer 203, a feature fusion layer 205, an output layer 207, and a regression layer 209. For example, the algorithm implementation steps are as follows: First, the data input layer 201 receives preprocessed standardized multi-source data and converts the two-dimensional stereo image data and three-dimensional point cloud data into a three-dimensional input tensor of the same dimension; Second, the feature extraction layer 203 performs multiple rounds of convolution operations through a 3×3×3 three-dimensional convolution kernel, and after each round of convolution, it is combined with batch normalization and ReLU activation function to suppress gradient vanishing, enhance the nonlinear expression capability of features, and gradually extract the shallow texture features, mid-level structural features and deep semantic features of the tray; Third, the feature fusion layer 205 splices and fuses the multi-scale features extracted by different convolution layers to improve the integrity of features; Fourth, the output layer 207 outputs the tray type recognition result through a fully connected layer and a Softmax classifier, and outputs the spatial position coordinates (x, y, z) and attitude angles (heading angle, pitch angle, roll angle) of the tray through the regression layer 209.

[0045] It should be noted that the lightweight network architecture described in step S3 is a network architecture that improves computational efficiency by simplifying the network structure and reducing the number of parameters while ensuring the accuracy of the algorithm's recognition. For example, the lightweight network architecture technology implementation steps are as follows: First, depthwise separable convolution is used to replace traditional 3D convolution, splitting the convolution operation into depthwise convolution (convolution for each input channel individually) and pointwise convolution (1×1×1 convolution fusing channel features), significantly reducing the number of parameters and computational load; Second, a bottleneck structure is introduced, further compressing the number of network parameters through dimensionality scaling operations. Residual connections are used inside the bottleneck structure to ensure effective feature transfer; Third, redundant fully connected layers at the end of the network are removed and replaced with global average pooling, simplifying the network structure while improving generalization ability; Fourth, the number of network channels is optimized by removing feature channels with low contribution through channel pruning techniques, further improving the network's operating speed.

[0046] It should be noted that the attention mechanism described in step S3 is a feature enhancement mechanism that can automatically focus on key feature regions of the tray and suppress interference from irrelevant background features. For example, the implementation steps of this attention mechanism are as follows: First, global average pooling and global max pooling are performed on the feature map output by the feature extraction layer 203 to obtain a dual-channel feature descriptor; channel weight coefficients are generated through two fully connected layers and a Sigmoid activation function, and the weight coefficients are multiplied with the original feature map by channel weighting to enhance the expression of key feature channels; Second, the channel-weighted feature map is subjected to max pooling and average pooling in the channel dimension, and the two pooling results are concatenated and then multiplied with a 1×1×1 convolution kernel and a Sigmoid activation function to generate a spatial weight map; Third, the spatial weight map and the feature map are multiplied by spatial weighting to focus on the tray region features and suppress background interference; Fourth, the feature map enhanced by the dual attention mechanism is input into the subsequent feature fusion layer 205.

[0047] It should be noted that the optimization of network parameters and training strategy described in step S3 is a technical solution to improve network training efficiency and recognition accuracy by reasonably setting the initial values ​​of network parameters and selecting a suitable training optimizer and loss function. For example, the optimized network parameters and training strategy include: First, parameter initialization: the parameters of convolutional and fully connected layers are initialized using the He normal initialization method to ensure gradient stability in the early stages of network training; the batch normalization layer parameters are initialized to a mean of 0 and a variance of 1, and the bias term is initialized to 0; Second, optimizer selection: the Adam optimizer is used, with an initial learning rate of 0.001. The learning rate decay strategy uses cosine annealing decay, with the learning rate gradually decaying according to a cosine curve every 10 epochs to avoid overfitting in the later stages of training; Third, loss function design: a joint loss function is used, including classification loss and regression loss. The classification loss uses the cross-entropy loss function to optimize the tray type recognition result; the regression loss uses the smooth L1 loss function to optimize the regression accuracy of the tray's spatial position and pose angle. The weights of the joint loss function are 0.4 for classification loss and 0.6 for regression loss; Fourth, training process optimization: mini-batch stochastic gradient descent is used. The training process includes SGD, with a batch size of 32. An early stopping strategy is introduced during training: training stops when the validation set loss shows no decrease for five consecutive epochs, saving the optimal model. The fifth step involves regularization: L2 regularization (weight decay coefficient of 0.0001) is introduced in the fully connected layers, along with random dropout (dropout probability of 0.5) to suppress network overfitting.

[0048] Based on the above technology, the improved 3D convolutional neural network algorithm compares and matches the features of the dataset with the optimized network parameters and training model to achieve rapid identification of pallet type. At the same time, it outputs the spatial coordinates, posture angle and other positioning information of the pallet. The identification and positioning results are fed back to the scheduling and linkage module in real time and transmitted to the damage detection unit simultaneously to provide a position reference for subsequent damage detection.

[0049] In step S4, after obtaining the spatial coordinates and attitude angle information of the pallet, the damage detection area is first accurately locked based on this positioning reference. Local data of the pallet's location is extracted from the preprocessed multi-dimensional data using coordinate clipping technology, eliminating interference from the background environment and surrounding debris to ensure the targeted and accurate extraction of damage features. For the locked pallet area data, damage features are extracted dimensionally according to a multi-feature fusion detection strategy.

[0050] In some embodiments, reference Figure 3 , Figure 3 This is a flowchart illustrating the multi-feature fusion detection strategy described in step S4. Figure 3 As shown, the strategy includes: S301, Extract the discontinuous length and number of discontinuities of the tray edge in the preprocessed image as visual damage features; S303, extract the flatness deviation and fork hole size deviation of the pallet from the preprocessed point cloud data as geometric damage features; S305, extract the standard deviation of the pressure distribution data after calibration as a mechanical damage characteristic; S307, the weight coefficients of each feature are determined by the analytic hierarchy process, and a comprehensive damage score is calculated based on the weight coefficients and the values ​​of each feature. The tray is divided into three levels: no damage, minor damage, and severe damage according to the preset scoring threshold.

[0051] Specifically, the visual damage feature extraction uses the Canny edge detection algorithm to extract the edge contours of the preprocessed tray stereo image local data. Morphological closing operations are used to eliminate edge burr interference, resulting in a smooth and complete tray edge contour. The extracted edge contours are fitted with straight lines based on the Hough line detection algorithm to identify discontinuous line segments in the contours. The length (unit: mm) of each discontinuous line segment and the total number of discontinuous line segments are counted. The longest discontinuous length and the total number of discontinuous line segments are used as core visual damage feature parameters, denoted as Fv1 and Fv2, respectively. The longest discontinuous length is used to characterize the severity of cracks and edge damage on the tray surface, and the total number of discontinuous line segments is used to characterize the dispersion of damage locations. The two work together to reflect the visual damage state of the tray.

[0052] It is understood that the geometric damage feature extraction involves performing planar fitting on the preprocessed local 3D point cloud data of the pallet. The RANSAC random sampling consensus algorithm is used to fit the reference plane on the upper surface of the pallet, and the vertical distance deviation from each point in the point cloud data to the reference plane is calculated. The maximum distance deviation is taken as the flatness deviation (unit: mm), denoted as Fg1, which is used to characterize the degree of deformation of the upper surface of the pallet. The pallet point cloud data is segmented by a region growing algorithm to accurately extract the point cloud subset of the region where the fork opening is located. The actual size (length and width) of the fork opening is calculated based on the minimum bounding rectangle algorithm. It is compared with the standard fork opening size to obtain the size deviation rate (unit: %), denoted as Fg2, which is used to characterize the deformation and damage degree of the fork opening. The flatness deviation Fg1 and the fork opening size deviation rate Fg2 together constitute the geometric damage feature parameters, which comprehensively reflect the geometric anomalies at the pallet structural level.

[0053] For example, the mechanical damage feature extraction is based on calibrated pressure distribution data. The real-time pressure values ​​(unit: N) of the pressure sensor are statistically analyzed during the loading and unloading process, and the standard deviation of the pressure distribution is calculated, denoted as Fm1, to characterize the uniformity of the pressure distribution. The larger the standard deviation, the more uneven the force on the pallet, which indirectly reflects the possible hidden damage to the internal structure of the pallet (such as the breakage of internal supports), resulting in an imbalance of force when bearing load. To improve the reliability of the mechanical features, the maximum and minimum pressure values ​​of the pressure distribution are recorded simultaneously, and the pressure difference is calculated. If the pressure difference exceeds a preset threshold (e.g., 30% of the standard bearing pressure), it is marked as an abnormal mechanical feature, further enhancing the ability to identify hidden damage.

[0054] It should be noted that, due to the differences in the dimensions and value ranges of the various damage characteristic parameters, the extracted characteristic parameters need to be standardized first. The min-max standardization method is used to map each characteristic parameter to the [0,1] interval. The standardization formula is: F' = (F - F min ) / (F max - F min ), where F is the original eigenvalue, F min F is the minimum sample value of this feature. max F' is the maximum sample value of the feature, and F' is the standardized feature value. After standardization, we obtain the standardized values ​​of visual damage features Fv1' and Fv2', the standardized values ​​of geometric damage features Fg1' and Fg2', and the standardized value of mechanical damage features Fm1'.

[0055] For example, the weight coefficients of each feature are determined using the Analytic Hierarchy Process (AHP), and a hierarchical structure model is constructed. The target layer is the comprehensive judgment of pallet damage, the criterion layer is the visual damage features, geometric damage features, and mechanical damage features, and the solution layer is the specific feature parameters. A judgment matrix is ​​constructed, and the rationality of the construction is ensured by a consistency test (consistency ratio CR < 0.1). Finally, the weight coefficients of each criterion layer are determined (e.g., visual damage feature weight 0.35, geometric damage feature weight 0.4, mechanical damage feature weight 0.25). The weights of the feature parameters of each solution layer under the corresponding criterion layer are equally distributed (i.e., Fv1' and Fv2' both have a weight of 0.5 under visual damage features; Fg1' and Fg2' both have a weight of 0.5 under geometric damage features).

[0056] The comprehensive damage score is calculated based on a weighted voting mechanism. The specific calculation process is as follows: First, the weighted scores of each criterion layer are calculated: visual damage feature score Sv = 0.5×Fv1' + 0.5×Fv2', geometric damage feature score Sg = 0.5×Fg1' + 0.5×Fg2', and mechanical damage feature score Sm = Fm1'. Then, the comprehensive damage score S = 0.35×Sv +0.4×Sg + 0.25×Sm is calculated. The comprehensive score S ranges from [0,1] and is converted into a percentage score (i.e., S×100) for easy grade determination.

[0057] Damage levels are categorized based on preset scoring thresholds. A comprehensive score of ≤20 indicates no damage; the pallet surface has no obvious damage, its geometry is normal, and stress is evenly distributed, allowing for normal transport. A comprehensive score of ≤60 (<20) indicates minor damage; the pallet may have slight cracks, minor deformation, or uneven stress distribution, which does not affect short-term load-bearing capacity but requires regular inspection. A comprehensive score >60 indicates severe damage; the pallet may have severe cracks, obvious deformation, or severe uneven stress distribution, making it unsuitable for transport and requiring immediate remediation. An abnormal feature priority judgment logic is also introduced. If any core feature parameter exceeds the severe damage threshold (e.g., longest discontinuity >50mm, flatness deviation >15mm, pressure distribution standard deviation >50% of the standard value), the comprehensive score is not calculated, and the pallet is directly judged as severely damaged, ensuring no severe damage is missed and improving the reliability and timeliness of the judgment.

[0058] In step S5, the scheduling linkage module feeds back the above results (pallet type, spatial coordinates, attitude angle, damage level and various characteristic parameters, etc.) to the remote scheduling system through the wireless low temperature resistance communication module, and at the same time issues linkage instructions to the AGV actuators (fork arm, driving system) to execute differentiated response strategies. Specifically: ① For undamaged pallets, a normal operation instruction is issued. The AGV adjusts its fork arm posture based on positioning information, precisely aligns with the pallet fork holes, completes the picking and placing operation, and then transfers the goods according to the preset path, sending a signal indicating completion of the operation; ② For slightly damaged pallets, a minor avoidance instruction is immediately generated, controlling the fork arm to stop moving and rise to a safe height. The AGV moves away from the damaged pallet at low speed, and simultaneously sends feedback to the dispatch system, which then reassigns the transfer tasks to nearby undamaged pallets; ③ For severely damaged pallets, an emergency avoidance instruction is immediately generated, forcibly stopping the AGV, controlling the fork arm to quickly rise to a safe height and driving the AGV to retreat to a safe distance. At the same time, an audible and visual alarm device (installed on top of the AGV) is triggered to remind on-site personnel to handle the situation promptly. The severely damaged information is synchronized to the remote dispatch system and the on-site monitoring terminal. The dispatch system marks the pallet as a high-risk damaged item and prohibits subsequent operation scheduling; after the on-site personnel have completed the handling and sent a recovery signal, the alarm is deactivated, and the AGV returns to normal operation.

[0059] Based on the same technical concept, embodiments of this application also provide a damaged pallet detection and identification system. (Reference) Figure 4 , Figure 4 This is a schematic diagram of a damaged pallet detection and identification system provided in an embodiment of this application. The system includes at least a low-temperature sensing module 401, a data processing module 403, an identification and damage detection module 405, a scheduling and linkage module 407, a power management module 409, and an edge computing module 411. Communication connections are established between the modules to achieve data interaction and command transmission, adapting to low-temperature operating environments. Specifically: The low-temperature sensing module 401 is used to collect multi-dimensional raw data of the pallet area. It includes a low-temperature binocular imaging device, a laser detection device, and a pressure sensing element. The low-temperature binocular imaging device is used to collect three-dimensional image data of the pallet, the laser detection device is used to collect three-dimensional point cloud data of the pallet, and the pressure sensing element is installed at the end of the fork arm of the mobile operation platform to collect pressure distribution data during the picking and placing of goods. The data processing module 403 is connected to the low-temperature sensing module and is used to perform data preprocessing and data synchronization and alignment operations in step S2. The identification and damage detection module 405 is connected to the data processing module and includes a pallet identification unit and a damage detection unit; wherein, the pallet identification unit is used to perform the pallet type identification and spatial posture positioning operation in step S3, and the damage detection unit is used to perform the damage status detection and level determination operation in step S4. The scheduling linkage module 407 is connected to the identification and damage detection module and the scheduling system respectively, and is used to perform the result feedback, avoidance instruction generation and task redistribution linkage operation in step S5 of claim 1. The power management module 409 is used to provide stable power supply to each module. It is equipped with a low-temperature adaptive energy storage unit and a power heating component to ensure the stability of power supply in low-temperature environments. The edge computing module 411 is integrated locally on the mobile operating platform and is used to deploy data processing algorithms, pallet recognition algorithms, and damage detection algorithms. The edge computing module establishes a communication connection with a remote server to realize statistical analysis of damaged pallet data and iterative updates of algorithm models.

[0060] In some embodiments, the low-temperature sensing module 401 collects multi-dimensional raw data from the tray area and has both self-checking and parameter adaptive adjustment functions. After startup, the module sends a real-time acquisition status signal (normal / abnormal) to the data processing module. If the data processing module reports substandard data quality, the module automatically adjusts the acquisition parameters (such as imaging exposure time and laser detection frequency). If a device malfunction is detected (such as the binocular imaging device failing to activate the anti-fog and defrosting function), the module immediately switches to a backup device to ensure uninterrupted data acquisition. The collected data is transmitted to the data processing module in real-time, providing raw data support for subsequent processing.

[0061] For example, the data processing module 403 performs data preprocessing and spatiotemporal synchronization alignment operations. After receiving the raw data transmitted by the low-temperature sensing module 401, it completes image dehazing, point cloud noise reduction, and pressure data calibration according to a preset algorithm, while simultaneously achieving time synchronization and spatial coordinate alignment of the three types of data; it provides real-time feedback on the preprocessing progress and data quality assessment results (qualified / unqualified). If the data is unqualified, it requests the low-temperature sensing module 401 to re-acquire or adjust the parameters; the preprocessed standardized data is synchronously transmitted to the identification and damage detection module, and temporarily stored in the local storage unit of the edge computing module for subsequent traceability and algorithm iteration.

[0062] For example, the identification and damage detection module 405 realizes pallet type identification, spatial posture positioning, and damage state detection and level determination, and includes a pallet identification unit and a damage detection unit. The pallet identification unit calls the improved three-dimensional convolutional neural network algorithm deployed by the edge computing module 411, receives standardized data transmitted by the data processing module 403, and outputs identification and positioning results. If incomplete data features are found (such as missing point cloud data), a reprocessing request is immediately sent to the data processing module 403. The damage detection unit locks the detection area based on the positioning results, calls the multi-feature fusion detection algorithm, extracts visual, geometric, and mechanical damage features, completes the damage level determination, and transmits the detection results to the scheduling and linkage module in real time. If there is an anomaly in the detection results (such as a sudden change in feature parameters), re-detection is automatically triggered to avoid misjudgment.

[0063] For example, the scheduling linkage module 407 realizes linkage of detection result feedback, avoidance command generation, and task reallocation. It establishes real-time communication with the remote scheduling system through a wireless low-temperature communication module, transmitting detection results and operation status information; after receiving the detection results from the identification and damage detection module 405, it quickly generates differentiated commands (normal operation command, minor avoidance command, and emergency avoidance command) and sends them to the AGV actuator; when a severely damaged pallet is detected, it immediately triggers an audible and visual alarm and simultaneously sends a high-risk warning signal to the scheduling system; it provides real-time feedback on command execution status, and if command execution fails (e.g., avoidance action is not completed), it triggers the identification and damage detection module 405 to re-detect to confirm the damage status, avoiding false linkage.

[0064] For example, the power management module 409 provides stable power to each module, adapting to low-temperature operating environments. The module is equipped with a low-temperature adapted lithium battery pack (low-temperature adapted energy storage unit) and a power heating component. When the ambient temperature is below a preset threshold (e.g., -10℃), the heating component is automatically activated to ensure stable discharge of the lithium battery pack. It also monitors the power supply voltage, current, and power consumption of each module, including the low-temperature sensing module 401, data processing module 403, and identification and damage detection module 405, dynamically adjusting power supply priorities (prioritizing power supply to core modules). Furthermore, it has overload, overvoltage, and short-circuit protection functions. When abnormal power consumption of a module is detected, it immediately issues a warning signal and adjusts the power supply strategy to prevent module damage due to power supply failure.

[0065] For example, the edge computing module 411 is used to deploy core algorithms and achieve local rapid computation and algorithm iteration updates. The module is integrated into the AGV control cabin, locally deploying core algorithms such as data preprocessing algorithms, improved 3D convolutional neural network algorithms, and multi-feature fusion detection algorithms. It executes some data processing and identification detection computation tasks in parallel, reducing data transmission latency and improving processing efficiency. It establishes a communication connection with a remote server, periodically synchronizing damaged pallet data (such as damage type, damage location, detection time, etc.) for remote statistical analysis. It receives algorithm model iteration update packages from the remote server, completes update verification, and synchronously optimizes the algorithm model of the identification and damage detection module 405 to ensure the long-term stability and accuracy of the system. When the local algorithm malfunctions, it automatically switches to the remote algorithm running mode to ensure uninterrupted core functions.

[0066] Based on the same technical concept, this application embodiment also provides a damaged pallet detection and identification device, including a carrier body and the aforementioned damaged pallet detection and identification system integrated on the carrier body; the carrier body is a mobile work platform (e.g., AGV) adapted to low-temperature operations, the pressure sensing element in the low-temperature sensing module 401 is integrated at the end of the fork arm of the mobile work platform, and the edge computing module 411, data processing module 403 and power management module 409 are integrated in the control cabin of the mobile work platform to realize integrated operation of pallet identification, damage detection and operation linkage.

[0067] Based on the same technical concept, this application also provides a damaged pallet detection and identification device, see reference. Figure 5 , Figure 5 This is a schematic diagram of the structure of a damaged pallet detection and identification device provided in an embodiment of this application. Figure 5 As shown, the damaged pallet detection and identification device includes a memory 501 and a processor 502. The memory 501 is used to store computer instructions; when the processor 502 executes the computer instructions, it implements the method steps in any of the method embodiments.

[0068] The memory 501 includes at least one type of computer-readable storage medium, including flash memory, hard disk, multimedia card, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium may be an internal storage unit of an electronic device, such as the hard disk or memory of the electronic device. In other embodiments, the computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, secure digital card (SD card), flash memory card, etc., equipped on the electronic device. Of course, the computer-readable storage medium may include both internal storage units and external storage devices of the electronic device. In this embodiment, the computer-readable storage medium is typically used to store the operating system and various application software installed on the electronic device, such as the program code of the damaged tray detection and identification method in the embodiment. In addition, the computer-readable storage medium may also be used to temporarily store various types of data that have been output or will be output.

[0069] In some embodiments, processor 502 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other chip. Processor 502 is typically used to control the overall operation of the processing device, such as performing control and processing related to data interaction or communication with other entities. In this embodiment, processor 502 is used to run program code stored in memory 501 or process data.

[0070] Based on the same technical concept, this application also provides a computer-readable storage medium, which includes a computer program or instructions stored in the storage medium. When the computer program or instructions are executed by a processing device, they implement the method steps in any method embodiment. Further details can be found in the method embodiments, which will not be repeated here. In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium can be an internal storage unit of an electronic device, such as the hard disk or memory of the electronic device. In other embodiments, the computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, secure digital card (SD card), flash memory card, etc., equipped on the electronic device. Of course, the computer-readable storage medium can also include both internal storage units and external storage devices of the electronic device. In this embodiment, the computer-readable storage medium is typically used to store the operating system and various application software installed on the electronic device, such as the program code of the damaged tray detection and identification method in the embodiment. Furthermore, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or will be output.

[0071] Based on the same technical concept, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform a damaged pallet detection and identification method provided in the above-described method embodiments.

[0072] It should be noted that the order of description of the embodiments in this application is not intended to limit the priority of the embodiments.

[0073] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this application and in its specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0074] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many forms under the guidance of this application without departing from the spirit and scope of protection of the claims. All equivalent transformations made under the inventive concept of this application using the content of this application's specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of this application.

Claims

1. A method for detecting and identifying damaged pallets, characterized in that, Includes the following steps: S1. Collect multi-dimensional raw data of the pallet area. The multi-dimensional raw data includes at least pallet stereoscopic image data, three-dimensional point cloud data, and pressure distribution data during the picking and placing of goods. S2. Preprocess the multi-dimensional raw data. The preprocessing includes image dehazing, point cloud noise reduction, and pressure data calibration. At the same time, the time synchronization and spatial coordinate alignment of the stereo image data and the three-dimensional point cloud data are completed. S3. Based on the preprocessed multi-dimensional data, an improved three-dimensional convolutional neural network algorithm is used to realize pallet type recognition and spatial posture positioning. S4. A multi-feature fusion detection strategy is adopted to detect the damage status of the pallet. The multi-feature fusion detection strategy includes edge discontinuity detection based on the image, geometric anomaly detection based on the point cloud, and uneven force detection based on the pressure distribution. A comprehensive damage score is calculated through a weighted voting mechanism, and the pallet damage level is determined according to a preset scoring threshold. S5. Feed back the pallet recognition result, spatial attitude positioning information and damage detection result to the scheduling system. If a damaged pallet is detected, generate an avoidance command and control the actuator to avoid the damaged pallet. The scheduling system will then reassign the job task. If the detected pallet is not damaged, continue to execute the job task assigned by the scheduling system.

2. The method for detecting and identifying damaged pallets according to claim 1, characterized in that, The multi-dimensional raw data mentioned in step S1 is collected by a low-temperature sensing component, which includes a low-temperature binocular imaging device with anti-fogging and defrosting functions, a low-temperature and anti-interference laser detection device, and a low-temperature stable pressure sensing element installed at the end of the actuator fork arm.

3. The method for detecting and identifying damaged pallets according to claim 1, characterized in that, The improved 3D convolutional neural network algorithm described in step S3 is based on a lightweight network architecture and introduces an attention mechanism. By optimizing network parameters and training strategies, it achieves rapid identification of tray types and accurate positioning of spatial posture.

4. The method for detecting and identifying damaged pallets according to claim 1, characterized in that, The multi-feature fusion detection strategy described in step S4 includes: The length and number of discontinuities at the edge of the tray in the preprocessed image are extracted as visual damage features. The flatness deviation and fork hole size deviation of the pallet were extracted from the preprocessed point cloud data as geometric damage features; The standard deviation of the pressure distribution data after calibration was extracted as a mechanical damage characteristic. The weight coefficients of each feature are determined by the analytic hierarchy process (AHP). Based on the weight coefficients and the values ​​of each feature, a comprehensive damage score is calculated. The pallets are then classified into three levels—no damage, minor damage, and severe damage—according to a preset scoring threshold.

5. A damaged pallet detection and identification system, characterized in that, It includes at least a low-temperature sensing module, a data processing module, an identification and damage detection module, a scheduling and linkage module, and a power management module; communication connections are established between the modules to realize data interaction and command transmission, adapting to low-temperature operating environments; The low-temperature sensing module is used to collect multi-dimensional raw data of the pallet area. It includes a low-temperature binocular imaging device, a laser detection device, and a pressure sensing element. The low-temperature binocular imaging device is used to collect three-dimensional image data of the pallet, the laser detection device is used to collect three-dimensional point cloud data of the pallet, and the pressure sensing element is installed at the end of the fork arm of the mobile operation platform to collect pressure distribution data during the picking and placing of goods. The data processing module is connected to the low-temperature sensing module and is used to perform the data preprocessing and data synchronization alignment operations in step S2 of claim 1. The identification and damage detection module is connected to the data processing module and includes a pallet identification unit and a damage detection unit; wherein, the pallet identification unit is used to perform the pallet type identification and spatial posture positioning operation in step S3 of claim 1, and the damage detection unit is used to perform the damage status detection and level determination operation in step S4 of claim 1. The scheduling and linkage module is connected to the identification and damage detection module and the scheduling system respectively, and is used to perform the result feedback, avoidance instruction generation and task reallocation linkage operation in step S5 of claim 1. The power management module is used to provide stable power to each module. It is equipped with a low-temperature adaptive energy storage unit and a power heating component to ensure the stability of power supply in low-temperature environments.

6. The damaged pallet detection and identification system according to claim 5, characterized in that, It also includes an edge computing module, which is integrated locally on the mobile work platform and is used to deploy data processing algorithms, pallet recognition algorithms, and damage detection algorithms. The edge computing module establishes a communication connection with a remote server to realize statistical analysis of damaged pallet data and iterative updates of algorithm models.

7. The damaged pallet detection and identification system according to claim 5, characterized in that, The low-temperature resistant binocular imaging device has anti-fogging and defrosting functions, and adopts low-temperature adapted optical components and image sensing elements; the laser detection device has low-temperature resistance and anti-interference characteristics; the pressure sensing element adopts a low-temperature stable sensing structure; and data transmission between modules is achieved through a low-temperature resistant communication interface.

8. The damaged pallet detection and identification system according to claim 5, characterized in that, The power management module also has overload protection to ensure power supply safety; the low-temperature adaptive energy storage unit has stable discharge capability in the target low-temperature operating environment.

9. The damaged pallet detection and identification system according to claim 5, characterized in that, The scheduling linkage module communicates with the scheduling system through a wireless low-temperature resistant communication module. When a severely damaged pallet is detected, it generates an emergency avoidance command and triggers an alarm signal.

10. A damaged pallet detection and identification device, characterized in that, The system includes a support body and a damaged pallet detection and identification system according to any one of claims 5-9 integrated on the support body; the support body is a mobile work platform adapted for low-temperature operations, the pressure sensing element in the low-temperature sensing module is integrated at the end of the fork arm of the mobile work platform, and the edge computing module, data processing module and power management module are integrated in the control cabin of the mobile work platform to realize integrated operation of pallet identification, damage detection and operation linkage.

11. A damaged pallet detection and identification device, characterized in that, It includes a memory and a processor, the memory being used to store computer programs or instructions; when the computer programs or instructions are executed by the processor, the method of any one of claims 1-4 is implemented.

12. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions, which, when executed by a processor, implement the method of any one of claims 1-4.

13. A computer program product or computer program, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, the method of any one of claims 1-4 is implemented.

Images