Intelligent identification method, device and equipment for three super express items
By introducing Ghost convolutional layers and CBAM attention mechanisms into the YOLOv5 model, combined with multi-head self-attention mechanisms and PPO reinforcement learning, the problems of low identification efficiency and insufficient dynamic billing for oversized and overweight express parcels are solved, achieving more efficient and accurate identification and billing optimization.
Patent Information
- Application Number
- CN202510948307.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-10
- Publication Date
- 2025-11-11
AI Technical Summary
In existing technologies, the identification efficiency of oversized and overweight express parcels is low, the billing rules are not dynamic enough, and the data silo problem leads to fragmented identification logic, resulting in low sorting efficiency, high transportation costs, and large errors in manual billing.
We replaced the Ghost convolutional layer with the YOLOv5 model and added the CBAM attention mechanism. Combined with the multi-head self-attention mechanism, we constructed a package type identification model for oversized and overweight parcels. We identified the parcel type through multimodal data fusion and optimized the billing strategy using the PPO reinforcement learning algorithm.
It improved the accuracy and efficiency of identifying oversized and overweight express parcels, reduced the false detection rate, optimized the billing strategy, and achieved more precise transportation cost control.
Smart Images
Figure CN120929831A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer technology, and in particular to an intelligent identification method, device, and equipment for oversized and overweight express parcels. Background Technology
[0002] With the rapid development of the e-commerce industry, the number of express parcels has increased exponentially. Among them, "oversized and overweight parcels" cause problems such as low sorting efficiency, high transportation costs, and large errors in manual billing due to exceeding size and weight limits. Traditional identification of oversized and overweight parcels relies on manual measurement and static billing rules, which has the following drawbacks: Low identification efficiency: Manual measurement is time-consuming and prone to errors, especially during peak periods, which can easily cause sorting bottlenecks.
[0003] Existing billing rules are mostly based on fixed thresholds (such as single weight or volume standards) and do not take into account dynamic factors such as package shape, transportation route, and regional economic differences.
[0004] Data silo problem: Data from multiple sources such as distribution scanning, light curtain detection, and customer registration is not effectively integrated, resulting in fragmented judgment logic. Summary of the Invention
[0005] This invention provides an intelligent identification method, device, and equipment for oversized and overweight express parcels, which can quickly and accurately identify oversized and overweight express parcels.
[0006] On the one hand, the present invention provides an intelligent identification method for oversized and overweight express parcels, the method comprising: Acquire sample multimodal parcel data of sample express parcels; the sample multimodal parcel data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight. The standard convolutional layers in the YOLOv5 model are replaced with Ghost convolutional layers, and the CBAM attention mechanism is added to the Neck part to build the preset model; The sample multimodal wrapping data is input into the preset model, and the sample multimodal wrapping data is mapped to a high-dimensional space based on the embedding layer of the preset model to obtain the sample high-dimensional features; Based on the multi-head self-attention mechanism in the preset model, semantic features of the high-dimensional features of the sample are extracted from multiple levels and weighted and fused to obtain sample package type features; and the sample package type result of the sample express package is predicted based on the sample package type features. Based on the difference between the sample package type result and the sample package type label, the preset model is trained to obtain the express delivery oversized / overweight package identification model; Acquire the multimodal package data of the package to be tested, and input the multimodal package data into the rapid oversized / overweight package identification model to identify whether the package to be tested is an oversized / overweight express delivery package.
[0007] In one exemplary embodiment, obtaining sample multimodal package data of sample express parcels includes: The sample express parcel image is obtained using an image acquisition device; The sample package volume is obtained by capturing its length, width, and height in real time using a light curtain sensor. The weight of the sample express package is obtained based on the weighing equipment, and the sample weight data is obtained. Based on the sample package volume and the sample weight data, the sample package weight and volume data are obtained; Obtain the customer registration information and order data of the sample express parcels entered into the system to obtain the sample customer history records; The geographical location of the distribution center, real-time road conditions of the transportation route, and fuel cost fluctuations of the sample express parcels are obtained as sample environmental parameters. The sample package image, the sample package weight and volume data, the sample customer history records, and the sample environmental parameters are preprocessed to obtain the sample multimodal package data of the sample express package.
[0008] In one exemplary embodiment, inputting the sample multimodal wrapping data into the preset model includes: The sample multimodal wrapping data is augmented using at least one of HSV color space transformation, random rotation, and mosaic enhancement to obtain sample augmented data. The sample multimodal wrapping data and the sample augmentation data are input into the preset model.
[0009] In one exemplary embodiment, training the preset model based on the difference between the sample package type result and the sample package type label to obtain the express delivery oversized / overweight package identification model includes: A loss function is constructed using Focal Loss, and the target loss data is determined based on the loss function and the difference between the sample package type result and the sample package type label. Obtain a preset iteration period, and determine the training termination condition based on the target loss data and the preset iteration period; The model parameters of the preset model are adjusted according to the target loss data until the training termination condition is met, and the preset model at the end of training is determined as the express delivery oversized and overweight package identification model.
[0010] In one exemplary embodiment, the step of extracting semantic features of the high-dimensional features of the sample from multiple levels and performing weighted fusion based on the multi-head self-attention mechanism in the preset model to obtain sample package type features includes: The priorities of various data in the multimodal package data of the sample are determined; the sample package image, the sample package weight and volume data, and the sample environmental parameters have the same priority; the priority of the sample package weight and volume data is higher than the priority of the customer registration information of the sample express package, and the priority of the customer registration information of the sample express package is higher than the priority of the order data; Based on the delivery time requirements of the sample express parcels, determine the order priority of the sample express parcels; Based on the priorities of various data in the sample multimodal package data, the order priority of the sample express package, and the multi-head self-attention mechanism in the preset model, the semantic features of the high-dimensional features of the sample are extracted from multiple levels and the weight of each semantic feature is determined. For each sample's high-dimensional features, the product of the semantic features at each level and the weight of each semantic feature is calculated to obtain the sample-weighted features of each sample's high-dimensional features. The sample wrapping type feature is obtained by calculating the sum of the sample weighted features of the high-dimensional features of each sample.
[0011] In one exemplary embodiment, when the package to be tested is identified as an oversized or substandard express delivery item, the method further includes: The sample parcel attributes, sample transportation costs, and sample regional economic index of the area where the sample express parcels are located are obtained; the sample express parcels are labeled with sample billing strategy tags. The sample package attributes, sample transportation costs, and sample regional economic indices of the sample express delivery oversized items are input into the PPO reinforcement learning algorithm model to obtain the sample prediction billing strategy. Based on the difference between the sample predicted billing strategy and the sample billing strategy label, the PPO reinforcement learning algorithm model is trained to obtain the billing strategy prediction model; The package attributes, transportation costs, and regional economic index of the test package corresponding to the three major express delivery items are input into the billing strategy prediction model, and the target billing strategy of the test package is output.
[0012] In one exemplary embodiment, the packages to be tested are multiple, and the multiple packages to be tested are packages in the target area; after outputting the target billing policy for the packages to be tested, the method further includes: Based on the identification results of oversized / overweight items for each package to be tested, a data map of oversized / overweight item identification is constructed. The billing deviation distribution results are determined based on the target billing strategy corresponding to each test package and the location information of each test package in the target area; Construct a regional cost heatmap of the target area based on the billing deviation distribution results; The display panel shows the identification data of the three oversized components, the billing deviation distribution results, and the regional cost heat map.
[0013] On the other hand, an intelligent identification device for oversized and overweight express parcels is provided, the device comprising: The sample multimodal data acquisition module is used to acquire sample multimodal package data of sample express parcels; the sample multimodal package data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight. The model building module is used to replace the standard convolutional layers in the YOLOv5 model with Ghost convolutional layers and add the CBAM attention mechanism to the Neck part to build a preset model; The feature mapping module is used to input the sample multimodal wrapping data into the preset model, and map the sample multimodal wrapping data to a high-dimensional space based on the embedding layer of the preset model to obtain the sample high-dimensional features; The feature weighting module is used to extract semantic features of the high-dimensional features of the sample from multiple levels according to the multi-head self-attention mechanism in the preset model, and perform weighted fusion to obtain sample package type features; and predict the sample package type result of the sample express package based on the sample package type features. The model training module is used to train the preset model based on the difference between the sample package type result and the sample package type label to obtain the express oversized package identification model; The oversized / overweight package identification module is used to acquire the multimodal package data of the package to be tested, and input the multimodal package data into the fast oversized / overweight package identification model to identify whether the package to be tested is an oversized / overweight express delivery package.
[0014] On the other hand, an electronic device is provided, the device including a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to realize the intelligent identification method for oversized and overweight express parcels as described above.
[0015] On the other hand, a computer storage medium is provided, which stores at least one instruction or at least one program, which is loaded and executed by a processor to realize the intelligent identification method for oversized and overweight express parcels as described above.
[0016] On the other hand, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A 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 the intelligent identification method for oversized and overweight express parcels as described above.
[0017] The intelligent identification method, device, and equipment for oversized and overweight express parcels provided by this invention have the following technical effects: This invention acquires multimodal parcel data of sample express parcels. The multimodal parcel data includes parcel images, parcel weight and volume data, and customer history records. Each parcel is labeled with a parcel type tag, indicating whether it is a super-large or oversized parcel. The standard convolutional layers in the YOLOv5 model are replaced with Ghost convolutional layers, and a CBAM attention mechanism is added to the Neck part to construct a preset model. The multimodal parcel data is input into the preset model, and the embedding layer of the preset model maps the multimodal parcel data to a high-dimensional space to obtain high-dimensional features of the samples. Based on the multi-head self-attention mechanism in the preset model, semantic features of high-dimensional features of samples are extracted from multiple levels and weighted and fused to obtain sample package type features. The sample package type results of the sample express packages are predicted based on the sample package type features. The preset model is trained based on the difference between the sample package type results and the sample package type labels to obtain the express oversized / overweight package identification model. The multimodal package data of the package to be tested is obtained and input into the fast oversized / overweight package identification model to identify whether the package to be tested is an express oversized / overweight package, thereby improving the accuracy and efficiency of express oversized / overweight package identification. Attached Figure Description
[0018] To more clearly illustrate the technical solutions and advantages in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or 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.
[0019] Figure 1 This is a schematic diagram of an intelligent identification system for oversized and overweight express parcels provided in the embodiments of this specification; Figure 2 This is a flowchart illustrating an intelligent identification method for oversized and overweight express parcels provided in the embodiments of this specification. Figure 3 This is a flowchart illustrating a method for obtaining sample multimodal parcel data of sample express parcels, as provided in the embodiments of this specification. Figure 4 This is a flowchart illustrating a method for training a preset model to obtain an oversized express parcel identification model based on the difference between the sample parcel type result and the sample parcel type label, as provided in an embodiment of this specification. Figure 5 This is a flowchart illustrating a method for determining the target billing strategy of a package under test based on a billing strategy prediction model, as provided in the embodiments of this specification. Figure 6 This is a schematic diagram of the structure of an intelligent identification device for oversized and overweight express parcels provided in the embodiments of this specification; Figure 7 This is a schematic diagram of the structure of a server provided in the embodiments of this specification. Detailed Implementation
[0020] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only 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.
[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0022] Please see Figure 1 , Figure 1 This is a schematic diagram of an intelligent identification system for oversized and overweight express parcels provided in the embodiments of this specification, as shown below. Figure 1 As shown, the intelligent identification system for oversized and overweight express parcels may include at least a server 01 and a client 02.
[0023] Specifically, in the embodiments of this specification, server 01 may include a standalone server, a distributed server, or a server cluster composed of multiple servers. It may also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Server 01 may include a network communication unit, a processor, and a memory, etc. Specifically, server 01 can be used to train a model for identifying overloaded express parcels.
[0024] Specifically, in the embodiments of this specification, the client 02 may include physical devices such as smartphones, desktop computers, tablets, laptops, digital assistants, smart wearable devices, smart speakers, in-vehicle terminals, and smart TVs. It may also include software running on the physical device, such as web pages provided to users by service providers, or applications provided to users by these service providers. Specifically, the client 02 can be used to query online whether a package to be tested is an oversized or overweight express delivery item based on the oversized / overweight item identification model.
[0025] In this embodiment, "oversized / overweight / overweight parcels" refer to items that exceed the prescribed dimensions, weight, and quantity. Specifically, "oversized / overweight / overweight parcels" include: Excessive length: The length of the item exceeds the maximum size limit stipulated by the courier company.
[0026] Overweight: The weight of the item exceeds the maximum weight limit stipulated by the courier company.
[0027] Excessive shipments: The number of items shipped at one time exceeds the maximum quantity limit stipulated by the courier company.
[0028] The following describes an intelligent identification method for oversized and overweight express parcels according to the present invention. Figure 2 This is a flowchart illustrating an intelligent identification method for oversized / overweight express parcels provided in an embodiment of this specification. This specification provides the operational steps described in the embodiments or flowchart, but based on conventional or non-creative labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or server product execution, the methods shown in the embodiments or drawings can be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment). Specifically, as shown in the embodiments or drawings... Figure 2 As shown, the method may include: S201: Obtain sample multimodal package data of sample express parcels; the sample multimodal package data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight. In the embodiments of this specification, the sample express parcel may include oversized / overweight parcels or non-oversized / overweight parcels. The sample parcel type label includes two types: an oversized / overweight label and a non-oversized / overweight label.
[0029] S203: Replace the standard convolutional layers in the YOLOv5 model with Ghost convolutional layers and add the CBAM attention mechanism to the Neck part to build the preset model.
[0030] In the embodiments described in this specification, the YOLOv5 model can be any one of the four models: Yolov5s, Yolov5m, Yolov5l, and Yolov5x.
[0031] Based on the YOLOv5n framework, the following optimizations are introduced: Ghost Convolution Replacement: The standard convolutional layers in CSPDarknet53 are replaced with GhostConv, reducing computation by 40% (FLOPs from 4.5G to 2.7G) and improving inference speed to 35FPS.
[0032] Enhanced attention mechanism: CBAM (Convolutional Block Attention Module) is added to the Neck part. By weighting channel and spatial attention, the detection accuracy of small targets (such as ultra-thin parts) is improved, reaching 98.7% mAP@0.5, and the inference speed is improved by 40%.
[0033] S205: Input the sample multimodal wrapping data into the preset model, and map the sample multimodal wrapping data to a high-dimensional space based on the embedding layer of the preset model to obtain the sample high-dimensional features.
[0034] In the embodiments of this specification, package image features (CNN output), weight and volume data, and customer history (such as frequent shipments of oversized items) can be mapped to a high-dimensional space through an embedding layer.
[0035] S207: Based on the multi-head self-attention mechanism in the preset model, extract the semantic features of the high-dimensional features of the sample from multiple levels and perform weighted fusion to obtain the sample package type features; and predict the sample package type result of the sample express package based on the sample package type features.
[0036] In the embodiments of this specification, a multi-head attention mechanism can be used for feature association. Semantic features of the high-dimensional features of the sample are extracted at multiple levels and weighted and fused to obtain sample package type features. Then, the sample package type result can be predicted based on the sample package type features. The sample package type result can include results indicating the sample package is an oversized / overweight / excessive package or a non-oversized / overweight / excessive package. By fusing image, business, and customer data through a multi-head attention mechanism, and combining online learning to dynamically adjust processing priorities, the false detection rate is reduced by 60%.
[0037] S209: Based on the difference between the sample package type result and the sample package type label, the preset model is trained to obtain the express delivery oversized / overloaded package identification model.
[0038] In the embodiments of this specification, the model parameters of the preset model can be continuously adjusted by comparing the differences between the sample package type results and the sample package type labels until the training is completed and the express delivery oversized package identification model is obtained.
[0039] S2011: Obtain the multimodal package data of the package to be tested, and input the multimodal package data into the rapid oversized / overweight package identification model to identify whether the package to be tested is an oversized / overweight express package.
[0040] In the embodiments of this specification, the multimodal package data to be tested is input into the rapid oversized / overweight package identification model, and the package type result is output, wherein the package type result indicates whether the package to be tested is an oversized / overweight express delivery package.
[0041] In the embodiments described in this specification, high-precision, low-noise package feature extraction can be achieved through multi-source fusion of data from light curtain sensing, weighing equipment and business system, and 3σ cleaning rules.
[0042] For example, such as Figure 3 As shown, the acquisition of sample multimodal package data of sample express parcels includes: S20101: Obtain a sample package image of the sample express package based on the image acquisition device; S20103: The length, width, and height of the sample express package are captured in real time based on the light curtain sensor to obtain the volume of the sample package; S20105: Obtain the weight of the sample express package based on the weighing device to obtain sample weight data; S20107: Based on the sample package volume and the sample weight data, obtain the sample package weight and volume data; S20109: Obtain the customer registration information and order data of the sample express parcels entered into the system, and obtain the sample customer history records; S201011: Obtain the geographical location of the distribution center of the sample express parcel, the real-time road conditions of the transportation route, and the fuel cost fluctuation as sample environmental parameters; S201013: Perform data preprocessing on the sample package image, the sample package weight and volume data, the sample customer history records, and the sample environmental parameters to obtain the sample multimodal package data of the sample express package.
[0043] In the embodiments described in this specification, the hardware sensing layer includes: a light curtain sensor that captures the length, width, and height of packages in real time (accuracy ±1cm), and a weighing device on the sorting line that acquires weight data (accuracy ±0.1kg). The system input layer includes: customer registration information (such as special packaging requirements) and OMS order data (shipping and receiving addresses, insurance information). Environmental parameters include: the geographical location of the distribution center, real-time road conditions along the transportation route, and fuel cost fluctuations. This allows for the acquisition of sample multimodal package data.
[0044] Data preprocessing may include: outlier removal: outliers in the light curtain scan are removed using the 3σ principle, and missing data is repaired using interpolation. Feature standardization: numerical data such as weight and volume are normalized to the [0,1] interval, and address text is encoded using Word2Vec. In this embodiment, the preprocessed data can achieve data standardization so that it can be input into the model for application.
[0045] In this embodiment of the specification, inputting the sample multimodal wrapping data into the preset model includes: The sample multimodal wrapping data is augmented using at least one of HSV color space transformation, random rotation, and mosaic enhancement to obtain sample augmented data. The sample multimodal wrapping data and the sample augmentation data are input into the preset model.
[0046] In the embodiments of this specification, before model training, data augmentation processing can be performed using HSV color space transformation, random rotation (±30°), and mosaic augmentation to improve the robustness of the model.
[0047] In this embodiment of the specification, the step of extracting semantic features of the high-dimensional features of the sample from multiple levels and performing weighted fusion based on the multi-head self-attention mechanism in the preset model to obtain sample package type features includes: The priorities of various data in the multimodal package data of the sample are determined; the sample package image, the sample package weight and volume data, and the sample environmental parameters have the same priority; the priority of the sample package weight and volume data is higher than the priority of the customer registration information of the sample express package, and the priority of the customer registration information of the sample express package is higher than the priority of the order data; Based on the delivery time requirements of the sample express parcels, determine the order priority of the sample express parcels; Based on the priorities of various data in the sample multimodal package data, the order priority of the sample express package, and the multi-head self-attention mechanism in the preset model, the semantic features of the high-dimensional features of the sample are extracted from multiple levels and the weight of each semantic feature is determined. For each sample's high-dimensional features, the product of the semantic features at each level and the weight of each semantic feature is calculated to obtain the sample-weighted features of each sample's high-dimensional features. The sample wrapping type feature is obtained by calculating the sum of the sample weighted features of the high-dimensional features of each sample.
[0048] In the embodiments described in this specification, a priority weight matrix can be defined, for example: light curtain data confidence level > customer registration information > OMS entered data. Urgent orders (such as fresh produce) have a 30% higher priority.
[0049] If the system's judgment result is inconsistent with the manual review, the model parameters are updated through online learning, with an iteration cycle of ≤1 hour.
[0050] In the embodiments described in this specification, such as Figure 4 As shown, the step of training the preset model based on the difference between the sample package type result and the sample package type label to obtain the express delivery oversized / overweight package identification model includes: S2091: Construct a loss function using Focal Loss, and determine the target loss data based on the loss function and the difference between the sample package type result and the sample package type label; S2093: Obtain a preset iteration period, and determine the training termination condition based on the target loss data and the preset iteration period; S2095: Adjust the model parameters of the preset model according to the target loss data until the training termination condition is met, and determine the preset model at the end of training as the express delivery oversized / overloaded item identification model.
[0051] In the embodiments of this specification, Focal Loss is used instead of cross-entropy loss to solve the problem of imbalance between positive and negative samples (the proportion of overweight samples is less than 5%). When the hyperparameters α=2 and γ=1, the recall rate is improved by 12%.
[0052] During training, a preset iteration period can be set, and then the training termination condition can be determined based on the target loss data and the preset iteration period. For example, the training termination condition can be that the target loss data is less than a preset threshold and the preset iteration period is less than a preset duration.
[0053] For example, if the system's judgment result is inconsistent with the manual review, the model parameters are updated through online learning. For example, the preset iteration cycle is ≤1 hour.
[0054] In the embodiments of this specification, when the package to be tested is identified as an oversized or substandard express delivery item, such as... Figure 5 As shown, the method further includes: S501: Obtain the sample package attributes, sample transportation cost, and sample regional economic index of the area where the sample express delivery overloaded item is located; the sample express delivery overloaded item is labeled with a sample billing strategy tag. S503: Input the sample package attributes, sample transportation costs, and sample regional economic index of the sample express three-dimensional oversized items into the PPO reinforcement learning algorithm model to obtain the sample prediction billing strategy. S505: Based on the difference between the sample predicted billing strategy and the sample billing strategy label, train the PPO reinforcement learning algorithm model to obtain the billing strategy prediction model; S507: Input the package attributes, transportation costs, and regional economic index of the test package corresponding to the three oversized express parcels into the billing strategy prediction model, and output the target billing strategy for the test package.
[0055] In the embodiments of this specification, the PPO (Proximal Policy Optimization Algorithms), proposed by OpenAI in 2017, is a reinforcement learning algorithm based on policy optimization. By employing the PPO reinforcement learning algorithm to construct a dynamic charging model, and combining regional economic indices and transportation cost parameters, the charging deviation rate is optimized to below 3%.
[0056] In the embodiments of this specification, the sample package attributes of the sample express delivery oversized / overweight items may include data such as package size and weight; transportation costs may include, but are not limited to, route distance and fuel prices; and regional economic indices may include, but are not limited to, destination GDP and consumption levels. Based on the difference between the sample predicted billing strategy and the sample billing strategy label, loss data is determined. Then, the model parameters of the PPO reinforcement learning algorithm model are adjusted according to the loss data until the training termination condition is met. The training termination condition can be determined based on the loss data, and the PPO reinforcement learning algorithm model at the end of training is determined as the billing strategy prediction model.
[0057] A reward function was set during model training, with positive rewards for improved billing accuracy and reduced customer complaint rates, and negative penalties for freight deviations exceeding 10% and system computation latency exceeding 200ms. The algorithm was implemented using the Proximal Policy Optimization (PPO) algorithm to train the agent and explore the optimal pricing strategy. The Q-value update formula is as follows: The discount factor γ = 0.95 and the learning rate α = 0.001.
[0058] In this embodiment of the specification, the packages to be tested are multiple, and the multiple packages to be tested are packages in the target area; after outputting the target billing policy for the packages to be tested, the method further includes: Based on the identification results of oversized / overweight items for each package to be tested, a data map of oversized / overweight item identification is constructed. The billing deviation distribution results are determined based on the target billing strategy corresponding to each test package and the location information of each test package in the target area; Construct a regional cost heatmap of the target area based on the billing deviation distribution results; The display panel shows the identification data of the three oversized components, the billing deviation distribution results, and the regional cost heat map.
[0059] In the embodiments of this specification, the over-limit item identification results corresponding to each test package are used to construct and display an over-limit item identification data map; the billing deviation distribution results can also be determined and displayed to the user based on the target billing strategy corresponding to each test package and the location information of each test package in the target area; the regional cost heat map of the target area can also be constructed and displayed based on the billing deviation distribution results, so that the user can intuitively understand the cost heat map of the target area.
[0060] In the embodiments described in this specification, blockchain notarization and Tableau visualization dashboards are used to achieve end-to-end data traceability and visualized monitoring of sorting efficiency and costs.
[0061] 1. Management dashboard: Build a real-time monitoring dashboard based on Tableau or Power BI to display the accuracy of oversized parts identification, billing deviation distribution, regional cost heat map, etc.
[0062] 2. Blockchain Evidence Storage: The judgment results and billing records are recorded on the blockchain (Hyperledger Fabric) to ensure that the data is immutable and to support auditing and dispute tracing.
[0063] As can be seen from the technical solutions provided in the embodiments of this specification above, the embodiments of this specification acquire sample multimodal package data of sample express parcels; the sample multimodal package data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight; the standard convolutional layers in the YOLOv5 model are replaced with Ghost convolutional layers, and a CBAM attention mechanism is added to the Neck part to construct a preset model; the sample multimodal package data is input into the preset model, and the sample multimodal package data is mapped to the embedding layer based on the preset model. High-dimensional features of samples are obtained in high-dimensional space; based on the multi-head self-attention mechanism in the preset model, semantic features of the high-dimensional features of samples are extracted from multiple levels and weighted and fused to obtain sample package type features; the sample package type results of sample express packages are predicted based on the sample package type features; based on the difference between the sample package type results and the sample package type labels, the preset model is trained to obtain the express oversized / overweight package identification model; the multimodal package data of the package to be tested is obtained and input into the fast oversized / overweight package identification model to identify whether the package to be tested is an express oversized / overweight package, thereby improving the accuracy and efficiency of express oversized / overweight package identification.
[0064] This specification also provides an intelligent identification device for oversized and overweight express parcels, such as... Figure 6 As shown, the device includes: The sample multimodal data acquisition module 610 is used to acquire sample multimodal package data of sample express parcels; the sample multimodal package data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight. Model building module 620 is used to replace the standard convolutional layers in the YOLOv5 model with Ghost convolutional layers and add the CBAM attention mechanism to the Neck part to build a preset model; The feature mapping module 630 is used to input the sample multimodal wrapping data into the preset model, and map the sample multimodal wrapping data to a high-dimensional space based on the embedding layer of the preset model to obtain the sample high-dimensional features; The feature weighting module 640 is used to extract semantic features of the high-dimensional features of the sample from multiple levels according to the multi-head self-attention mechanism in the preset model and perform weighted fusion to obtain sample package type features; and predict the sample package type result of the sample express package based on the sample package type features. The model training module 650 is used to train the preset model based on the difference between the sample package type result and the sample package type label to obtain the express oversized package identification model. The oversized / overweight package identification module 660 is used to acquire the multimodal package data of the package to be tested, and input the multimodal package data into the fast oversized / overweight package identification model to identify whether the package to be tested is an oversized / overweight express delivery package.
[0065] In one exemplary embodiment, the sample multimodal data acquisition module includes: The sample image acquisition unit is used to acquire sample package images of the sample express package based on the image acquisition device; The sample volume acquisition unit is used to capture the length, width, and height of the sample express package in real time based on the light curtain sensor to obtain the sample package volume; The sample weight acquisition unit is used to acquire the weight of the sample express package based on the weighing device, and obtain sample weight data; The sample data acquisition unit is used to obtain the sample package weight and volume data based on the sample package volume and the sample weight data. The sample record acquisition unit is used to acquire the customer registration information and order data of the sample express parcels entered in the system, and obtain the sample customer history records; The sample environment parameter acquisition unit is used to acquire the geographical location of the distribution center of the sample express parcel, the real-time road conditions of the transportation route, and the fluctuation of fuel costs as sample environment parameters. The sample multimodal data acquisition unit is used to preprocess the sample package image, the sample package weight and volume data, the sample customer history records, and the sample environmental parameters to obtain the sample multimodal package data of the sample express package.
[0066] In one exemplary embodiment, the feature mapping module includes: The sample enhancement unit is used to perform data enhancement processing on the sample multimodal wrapping data using at least one of HSV color space transformation, random rotation, and mosaic enhancement to obtain sample enhanced data; The data input unit is used to input the sample multimodal wrapping data and the sample augmentation data into the preset model.
[0067] In one exemplary embodiment, the model training module includes: The target loss determination unit is used to construct a loss function using Focal Loss and determine the target loss data based on the loss function and the difference between the sample package type result and the sample package type label. The training condition determination unit is used to obtain a preset iteration period and determine the training termination condition based on the target loss data and the preset iteration period. The model determination unit is used to adjust the model parameters of the preset model according to the target loss data until the training termination condition is met, and to determine the preset model at the end of training as the express delivery oversized and overweight package identification model.
[0068] In one exemplary embodiment, the feature weighting module includes: The first priority determination unit is used to determine the priority of various data in the sample multimodal package data; the sample package image, the sample package weight and volume data, and the sample environmental parameters correspond to the same priority; the priority of the sample package weight and volume data is higher than the priority of the customer registration information of the sample express package, and the priority of the customer registration information of the sample express package is higher than the priority of the order data. The second priority determination unit is used to determine the order priority of the sample express parcel based on the delivery time requirements of the sample express parcel. The weight determination unit is used to extract the semantic features of the high-dimensional features of the sample from multiple levels and determine the weight of each semantic feature based on the priority of various data in the sample multimodal package data, the order priority of the sample express package, and the multi-head self-attention mechanism in the preset model. The sample weighting unit is used to calculate the product of the semantic features at each level and the weight of each semantic feature for each sample's high-dimensional features, so as to obtain the sample weighted features of each sample's high-dimensional features. The sample feature determination unit is used to calculate the sum of the sample weighted features of the high-dimensional features of each sample to obtain the sample package type feature.
[0069] In one exemplary embodiment, when the package to be tested is identified as an oversized or substandard express delivery item, the device further includes: The sample attribute acquisition module is used to acquire the sample package attributes, sample transportation costs, and sample regional economic index of the area where the sample express delivery overloaded items are located; the sample express delivery overloaded items are labeled with sample billing strategy tags. The sample attribute input module is used to input the sample package attributes, sample transportation costs, and sample regional economic index of the sample express three-dimensional oversized items into the PPO reinforcement learning algorithm model to obtain the sample prediction billing strategy. The prediction model training module is used to train the PPO reinforcement learning algorithm model based on the difference between the sample predicted billing strategy and the sample billing strategy label, so as to obtain the billing strategy prediction model. The target strategy determination module is used to input the package attributes, transportation costs, and regional economic index of the test package corresponding to the express delivery oversized / overweight parcel into the billing strategy prediction model, and output the target billing strategy of the test package.
[0070] In one exemplary embodiment, the apparatus further includes: The data graph construction module is used to construct a data graph for identifying oversized / overweight items based on the identification results of each package to be tested. The distribution result determination module is used to determine the billing deviation distribution result based on the target billing strategy corresponding to each package under test and the location information of each package under test in the target area; The heatmap construction module is used to construct a regional cost heatmap of the target area based on the billing deviation distribution results. The graph display module is used to display the three oversized parts identification data graph, the billing deviation distribution result, and the regional cost heat map in the display panel.
[0071] The apparatus and method embodiments described herein are based on the same inventive concept.
[0072] This specification provides an electronic device including a processor and a memory. The memory stores at least one instruction or at least one program, which is loaded and executed by the processor to implement the intelligent identification method for oversized express parcels as provided in the above method embodiments.
[0073] Embodiments of the present invention also provide a computer storage medium, which can be disposed in a terminal to store at least one instruction or at least one program related to implementing an intelligent identification method for oversized and overweight express parcels in the method embodiments. The at least one instruction or at least one program is loaded and executed by the processor to implement the intelligent identification method for oversized and overweight express parcels provided in the above method embodiments.
[0074] Embodiments of the present invention also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A 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 the intelligent identification method for oversized express parcels provided in the above-described method embodiments.
[0075] Optionally, in the embodiments of this specification, the storage medium may be located at at least one of the multiple network servers in a computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0076] The memory described in the embodiments of this specification can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for the functions, etc.; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory may also include a memory controller to provide the processor with access to the memory.
[0077] The intelligent identification method for oversized and overweight express parcels provided in this specification can be executed on a mobile terminal, computer terminal, server, or similar computing device. Taking a server as an example... Figure 7 This is a hardware structure block diagram of the server for an intelligent identification method for oversized and overweight express parcels provided in the embodiments of this specification. Figure 7As shown, the server 700 can vary significantly due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 710 (CPUs 710 may include, but are not limited to, microprocessors (MCUs) or programmable logic devices (FPGAs), a memory 730 for storing data, and one or more storage media 720 (e.g., one or more mass storage devices) for storing application programs 723 or data 722. The memory 730 and storage media 720 may be temporary or persistent storage. The program stored in the storage media 720 may include one or more modules, each module may include a series of instruction operations on the server. Furthermore, the CPU 710 may be configured to communicate with the storage media 720 and execute the series of instruction operations stored in the storage media 720 on the server 700. Server 700 may also include one or more power supplies 760, one or more wired or wireless network interfaces 750, one or more input / output interfaces 740, and / or one or more operating systems 721, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.
[0078] The input / output interface 740 can be used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of server 700. In one example, the input / output interface 740 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the input / output interface 740 may be a radio frequency (RF) module used for wireless communication with the Internet.
[0079] Those skilled in the art will understand that Figure 7 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, server 700 may also include... Figure 7 The more or fewer components shown, or having the same Figure 7 The different configurations shown.
[0080] As can be seen from the embodiments of the intelligent identification method, device, electronic device, or storage medium for oversized and overweight express parcels provided by the present invention, the present invention acquires sample multimodal parcel data of sample express parcels; the sample multimodal parcel data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcel is an oversized or overweight parcel; the standard convolutional layers in the YOLOv5 model are replaced with Ghost convolutional layers, and a CBAM attention mechanism is added to the Neck part to construct a preset model; the sample multimodal parcel data is input into the preset model, and the sample multimodal parcel data is then processed by the embedding layer based on the preset model. Modal package data is mapped to a high-dimensional space to obtain high-dimensional features of the samples. Based on the multi-head self-attention mechanism in the preset model, semantic features of the high-dimensional features of the samples are extracted from multiple levels and weighted and fused to obtain sample package type features. The sample package type results of the sample express packages are predicted based on the sample package type features. Based on the difference between the sample package type results and the sample package type labels, the preset model is trained to obtain the express oversized / overweight package identification model. The multimodal package data of the package to be tested is obtained and input into the fast oversized / overweight package identification model to identify whether the package to be tested is an express oversized / overweight package, thereby improving the accuracy and efficiency of express oversized / overweight package identification.
[0081] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0082] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0083] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer storage medium, such as a read-only memory, a disk, or an optical disk.
[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for intelligent identification of oversized and overweight express parcels, characterized in that, The method includes: Acquire sample multimodal parcel data of sample express parcels; the sample multimodal parcel data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight. The standard convolutional layers in the YOLOv5 model are replaced with Ghost convolutional layers, and the CBAM attention mechanism is added to the Neck part to build the preset model; The sample multimodal wrapping data is input into the preset model, and the sample multimodal wrapping data is mapped to a high-dimensional space based on the embedding layer of the preset model to obtain the sample high-dimensional features; Based on the multi-head self-attention mechanism in the preset model, semantic features of the high-dimensional features of the sample are extracted from multiple levels and weighted and fused to obtain sample package type features; and the sample package type result of the sample express package is predicted based on the sample package type features. Based on the difference between the sample package type result and the sample package type label, the preset model is trained to obtain the express delivery oversized / overweight package identification model; Acquire the multimodal package data of the package to be tested, and input the multimodal package data into the rapid oversized / overweight package identification model to identify whether the package to be tested is an oversized / overweight express delivery package.
2. The method according to claim 1, characterized in that, The acquisition of sample multimodal package data for sample express parcels includes: The sample express parcel image is obtained using an image acquisition device; The sample package volume is obtained by capturing its length, width, and height in real time using a light curtain sensor. The weight of the sample express package is obtained based on the weighing equipment, and the sample weight data is obtained. Based on the sample package volume and the sample weight data, the sample package weight and volume data are obtained; Obtain the customer registration information and order data of the sample express parcels entered into the system to obtain the sample customer history records; The geographical location of the distribution center, real-time road conditions of the transportation route, and fuel cost fluctuations of the sample express parcels are obtained as sample environmental parameters. The sample package image, the sample package weight and volume data, the sample customer history records, and the sample environmental parameters are preprocessed to obtain the sample multimodal package data of the sample express package.
3. The method according to claim 2, characterized in that, The step of inputting the sample multimodal wrapping data into the preset model includes: The sample multimodal wrapping data is augmented using at least one of HSV color space transformation, random rotation, and mosaic enhancement to obtain sample augmented data. The sample multimodal wrapping data and the sample augmentation data are input into the preset model.
4. The method according to claim 3, characterized in that, The step of training the preset model based on the difference between the sample package type result and the sample package type label to obtain the express delivery oversized / overweight package identification model includes: A loss function is constructed using Focal Loss, and the target loss data is determined based on the loss function and the difference between the sample package type result and the sample package type label. Obtain a preset iteration period, and determine the training termination condition based on the target loss data and the preset iteration period; The model parameters of the preset model are adjusted according to the target loss data until the training termination condition is met, and the preset model at the end of training is determined as the express delivery oversized and overweight package identification model.
5. The method according to any one of claims 1-4, characterized in that, The semantic features of the high-dimensional features of the sample are extracted from multiple levels and weighted and fused according to the multi-head self-attention mechanism in the preset model to obtain the sample package type features, including: The priorities of various data in the multimodal package data of the sample are determined; the sample package image, the sample package weight and volume data, and the sample environmental parameters have the same priority; the priority of the sample package weight and volume data is higher than the priority of the customer registration information of the sample express package, and the priority of the customer registration information of the sample express package is higher than the priority of the order data; Based on the delivery time requirements of the sample express parcels, determine the order priority of the sample express parcels; Based on the priorities of various data in the sample multimodal package data, the order priority of the sample express package, and the multi-head self-attention mechanism in the preset model, the semantic features of the high-dimensional features of the sample are extracted from multiple levels and the weight of each semantic feature is determined. For each sample's high-dimensional features, the product of the semantic features at each level and the weight of each semantic feature is calculated to obtain the sample-weighted features of each sample's high-dimensional features. The sample wrapping type feature is obtained by calculating the sum of the sample weighted features of the high-dimensional features of each sample.
6. The method according to claim 1, characterized in that, When the package to be tested is identified as an oversized or substandard express delivery item, the method further includes: The sample parcel attributes, sample transportation costs, and sample regional economic index of the area where the sample express parcels are located are obtained; the sample express parcels are labeled with sample billing strategy tags. The sample package attributes, sample transportation costs, and sample regional economic indices of the sample express delivery oversized items are input into the PPO reinforcement learning algorithm model to obtain the sample prediction billing strategy. Based on the difference between the sample predicted billing strategy and the sample billing strategy label, the PPO reinforcement learning algorithm model is trained to obtain the billing strategy prediction model; The package attributes, transportation costs, and regional economic index of the test package corresponding to the three major express delivery items are input into the billing strategy prediction model, and the target billing strategy of the test package is output.
7. The method according to claim 6, characterized in that, The number of packages to be tested is multiple, and these multiple packages are packages within the target area; after outputting the target billing policy for the packages to be tested, the method further includes: Based on the identification results of oversized / overweight items for each package to be tested, a data map of oversized / overweight item identification is constructed. The billing deviation distribution results are determined based on the target billing strategy corresponding to each test package and the location information of each test package in the target area; Construct a regional cost heatmap of the target area based on the billing deviation distribution results; The display panel shows the identification data of the three oversized components, the billing deviation distribution results, and the regional cost heat map.
8. An intelligent identification device for oversized and overweight express parcels, characterized in that, The device includes: The sample multimodal data acquisition module is used to acquire sample multimodal package data of sample express parcels; the sample multimodal package data includes sample parcel images, sample parcel weight and volume data, and sample customer history records, and the sample express parcels are labeled with sample parcel type tags; the sample parcel type tags indicate whether the sample express parcels are oversized or overweight. The model building module is used to replace the standard convolutional layers in the YOLOv5 model with Ghost convolutional layers and add the CBAM attention mechanism to the Neck part to build a preset model; The feature mapping module is used to input the sample multimodal wrapping data into the preset model, and map the sample multimodal wrapping data to a high-dimensional space based on the embedding layer of the preset model to obtain the sample high-dimensional features; The feature weighting module is used to extract semantic features of the high-dimensional features of the sample from multiple levels according to the multi-head self-attention mechanism in the preset model, and perform weighted fusion to obtain sample package type features; and predict the sample package type result of the sample express package based on the sample package type features. The model training module is used to train the preset model based on the difference between the sample package type result and the sample package type label to obtain the express oversized package identification model; The oversized / overweight package identification module is used to acquire the multimodal package data of the package to be tested, and input the multimodal package data into the fast oversized / overweight package identification model to identify whether the package to be tested is an oversized / overweight express delivery package.
9. An electronic device, characterized in that, The device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the intelligent identification method for oversized and overweight express parcels as described in any one of claims 1-7.
10. A computer storage medium, characterized in that, The computer storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the intelligent identification method for oversized and overweight express parcels as described in any one of claims 1-7.