A smart logistics sorting method, system and storage medium

Abstract

The present invention discloses an intelligent logistics sorting method, system and storage medium. The image information of the package is obtained through a transmission mechanism, the barcode information of the package is identified and the identification confidence is determined. If it is lower than a preset threshold, the pushing mechanism is controlled to sort the package into a collection box; if it is higher than or equal to the preset threshold, the destination is classified, the sorting mechanism is controlled to distribute the package to different conveyor belts according to the destination, the size and shape of the package at the same destination are detected, the storage grid is optimized, and the classification mechanism is controlled to sort the package at the same destination into the corresponding storage grid. The present invention identifies the barcode information of the package, sorts the package whose barcode information cannot be identified into the collection box, sorts the package whose barcode information can be identified into the corresponding conveyor belt according to the destination, and sorts the express package at the same destination into the corresponding storage grid according to the size, thereby achieving efficient package processing.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent logistics technology, and in particular relates to an intelligent logistics sorting method, system and storage medium. Background Art

[0002] With the development of society, online shopping has become the primary shopping method for consumers. When sending express parcels from their origin to their destination, they need to be sorted and stacked according to their destination to ensure rapid transit. Traditional manual sorting methods are inefficient and labor-intensive. Existing intelligent sorting first automatically reads the parcel's barcode using a reader / writer installed on a conveyor belt. A control system then receives the data from the reader / writer, processes it according to preset sorting rules, and generates sorting instructions. Finally, a sorting device, such as a pusher, chute, or pneumatic device, sorts parcels destined for different destinations into corresponding containers based on the sorting instructions issued by the control system.

[0003] However, when the barcode of the express package is not facing the reader, the existing intelligent sorting method cannot effectively identify the destination of the express package, and the control system cannot issue the correct sorting instruction for the express package. Therefore, the sorting device cannot sort the express package into the corresponding container, causing the express package to stay on the conveyor belt for a long time or enter the inappropriate container, resulting in low sorting efficiency and sorting accuracy.

[0004] Moreover, the existing intelligent sorting is to sort all express parcels of different sizes to the same destination into the same container according to the sorting instructions. The larger express parcels are stacked on top of the smaller express parcels, which is inconvenient to take out and easily causes the smaller express parcels to be squeezed and damaged. At the same time, the existing intelligent sorting only sets the sorting device on one side of the conveyor belt, which requires a longer conveyor belt for conveying and sorting express parcels, occupies a large space, and the express parcels are conveyed over a long distance on the conveyor belt, resulting in low sorting efficiency.

[0005] Overall, these shortcomings and defects limit the effectiveness of parcel sorting, and there is an urgent need to improve sorting efficiency and sorting accuracy. Summary of the Invention

[0006] In response to the above-mentioned defects in the prior art, the present invention provides an intelligent logistics sorting method, which includes the following steps.

[0007] Step 1: The image information of the package is obtained and preprocessed through the collection device of the transmission mechanism, and the barcode information of the package is identified based on the deep convolutional generative adversarial network.

[0008] Step 2: Determine the barcode recognition confidence based on the barcode information. If the recognition confidence is lower than a preset threshold, the barcode recognition of the package is determined to be abnormal, and the process proceeds to Step 3. If the recognition confidence is higher than or equal to the preset threshold, the barcode recognition of the package is determined to be normal, and the process proceeds to Step 4.

[0009] Step 3: Control the pushing mechanism to sort the parcels with abnormal barcode information into a collection box.

[0010] Step 4: Classify the parcels whose barcode information is normal by their destination, and control the sorting mechanism to distribute the parcels to different conveyor belts according to their destination.

[0011] Step 5: Detect the size and shape of the parcels with the same destination, optimize the allocation of storage cells, and control the sorting mechanism to sort the parcels with the same destination into the corresponding storage cells.

[0012] Step 6: Collect sorting data, which includes the barcode information recognition results, destination classification results, and package size and shape detection results of each package. Analyze the barcode information recognition rate, destination classification accuracy, and package storage efficiency, and optimize and adjust the destination classification and storage grid allocation based on the analysis results.

[0013] The step 1 of obtaining the image information of the package and performing preprocessing includes converting the obtained color image of the package into a grayscale image, removing noise through Gaussian filtering, and improving the image contrast through histogram equalization.

[0014] The Canny edge detection algorithm is used to determine the position of the barcode on the package, and the ROI region of interest (ROI) containing only the barcode is extracted and cropped.

[0015] The cropped image is adjusted to a predetermined size and pixel-normalized, which is expressed as:

[0016] .

[0017] The mean μ and standard deviation σ of image I are expressed as:

[0018] ;

[0019] .

[0020] Where I is the cropped image of the ROI region of interest containing only the barcode, Inorm is the normalized image, M is the number of image rows, N is the number of image columns, μ is the mean of the image, and σ is the standard deviation of the image.

[0021] Among them, in step 1, the recognition of the barcode information of the package based on the deep convolutional generative adversarial network includes constructing a generative network G, inputting a random noise vector z, and generating a feature map of the barcode through multiple convolutional layers, batch normalization, and activation function ReLU, which is expressed as:

[0022] .

[0023] Where z is a random noise vector, BatchNorm(z) is a batch normalization operation used to standardize the input z, ReLU is an activation function, and Conv2D is a two-dimensional convolution operation used to extract spatial relationships in input features.

[0024] Construct the discriminant network D, input the standardized image Inorm, and extract features through multiple convolutional layers and LeakyReLU activation function, which can be expressed as:

[0025] .

[0026] Among them, Inorm is the normalized image, Conv2D is a two-dimensional convolution operation used to extract spatial information from the input features, LeakyReLU is the activation function, and the Sigmoid function outputs a value between 0 and 1, indicating the probability that the image is a real barcode.

[0027] The adversarial loss function is constructed based on the loss of real images and the loss of generated images, which is expressed as:

[0028] .

[0029] Among them, Ex~pdata(x) represents the expected value of the real data, Denote the loss of real images, Expressed as the loss of generating images, x represents the real image, pdata(x) is the distribution of real data, Ez~p z (z) represents the expected value of random noise z, G(z) is the image generated by the network output, and p z (z) is the distribution of noise.

[0030] The discriminant network and the generative network are iteratively trained based on an alternating optimization strategy, including minimizing the loss function L GAN Update the discriminant network and minimize the loss of generated images Update the generative network.

[0031] The feature map output by the trained generative network is processed through the convolution layer to extract key features, and the extracted features are mapped to the feature vector F using the fully connected layer, which is expressed as:

[0032] F=[f1,f2,…,fi,…,fn].

[0033] Here, fi represents the characteristics of the barcode, and the characteristics of the barcode at least include the type, size, and clarity of the barcode.

[0034] The confidence level in step 2 includes defining a calculation formula for the confidence level C, which is expressed as:

[0035] .

[0036] Among them, C is the confidence of barcode recognition, ranging from 0 to 1, α is the parameter that controls the confidence growth rate, β is the baseline value used to adjust the offset of the confidence, T is the total length of barcode feature extraction, wi is the weight coefficient, and fi(t) is the value of barcode feature i at time t.

[0037] The destination classification of the package in step 4 includes extracting and recognizing the information in the barcode through the OCR optical character recognition module, and pre-processing the recognized barcode information into a feature vector.

[0038] The generated feature vector is input into a pre-trained multi-layer perceptron (MLP) classification model.

[0039] The output of the classification model is processed using the Softmax activation function to calculate the probability distribution of each destination category, which is expressed as:

[0040] .

[0041] Where zi is the unnormalized score output by the classification model, and P(yi) is the probability of predicting class i.

[0042] Find the index of the maximum value in the probability distribution, determine the destination category with the highest probability, and use that category as the final classification result for the package.

[0043] Among them, detecting the size and shape of the same package in step 5 includes constructing a convolutional neural network (CNN) to identify the package shape and mark the category. The constructing of the convolutional neural network (CNN) includes constructing an input layer of the convolutional neural network (CNN) to receive the package image, extracting package image features through multiple convolutional layers, the package image features including shape information and edge information, mapping the extracted package image features to the package category through a fully connected layer, and using a training data set containing labeled package shapes and categories to train the CNN model until the model converges.

[0044] Based on a genetic algorithm, selection, crossover, and mutation operations are used to generate a target storage cell allocation scheme based on the shape and category of the package, as well as the package arrival status and the storage cell occupancy status. This includes randomly generating multiple storage cell allocation schemes, evaluating each scheme based on storage efficiency, access speed, and space utilization, selecting multiple initial schemes based on fitness using a tournament, crossovering the selected multiple initial schemes to generate a new storage cell allocation scheme, and repeating the evaluation, selection, and crossover operations until the target storage cell allocation scheme is obtained.

[0045] Generating intelligent sorting instructions and controlling the sorting mechanism to send the packages to the corresponding storage cells includes generating sorting instructions according to the target storage cell allocation plan, sending the sorting instructions to the sorting mechanism, and controlling the automated system to sort the packages to the target storage cells according to the target storage cell allocation plan.

[0046] The relevant information of each package includes at least one of the following: barcode information, destination classification result, package size and shape, and storage grid allocation.

[0047] Record each sorting situation, including at least one of the following: barcode recognition failure, destination classification error, package size mismatch.

[0048] Preprocess the collected data, including deduplication, cleaning, and filtering of invalid data.

[0049] The multi-layer perceptron (MLP) classification model is evaluated using the cross entropy loss function, expressed as:

[0050] .

[0051] Among them, y is the true label of the i-th sample, is the model's predicted probability for the i-th sample, and N is the number of samples.

[0052] The collected sorting data is added to the training set to form an incremental learning data set, and the model is continuously updated and retrained, which is expressed as:

[0053] .

[0054] Among them, θ new is the updated model parameter, θ old is the current model parameter, θ is the model parameter, η is the learning rate, is the gradient of the loss function with respect to the parameters.

[0055] The fitness function is established based on the collected sorting data, which is expressed as:

[0056] .

[0057] Among them, storage efficiency Eff= , S used is the used storage space, S total is the total capacity of the storage space, access speed , ∑Ts is the sum of the access time of the package, N is the total number of accessed packages, and α, β, γ are weight coefficients.

[0058] The weight coefficients α, β, and γ in the fitness function are adjusted through iterative learning to continuously optimize the storage grid allocation.

[0059] The present invention also provides an intelligent logistics sorting system, comprising: a transmission mechanism, a pushing mechanism, a sorting mechanism, a classification mechanism, a discrimination mechanism and an optimization mechanism.

[0060] The transmission mechanism is used to obtain image information of the package through the collection device of the transmission mechanism and perform preprocessing, and to identify the barcode information of the package based on the deep convolutional generative adversarial network.

[0061] The identification mechanism is used to determine the barcode recognition confidence based on the barcode information. If the recognition confidence is lower than a preset threshold, it is determined that the barcode information recognition of the package is abnormal. If the recognition confidence is higher than or equal to the preset threshold, it is determined that the barcode information recognition of the package is normal.

[0062] The pushing mechanism is used to sort the parcels with abnormal barcode information into the collection box.

[0063] The sorting mechanism is used to classify the destinations of packages with normal barcode information identification, and distribute the packages to different conveyor belts according to their destinations through the sorting mechanism.

[0064] The classification mechanism is used to detect the size and shape of packages with the same destination, optimize the allocation of storage cells, and sort the packages with the same destination into the corresponding storage cells through the classification mechanism.

[0065] The optimization mechanism is used to collect sorting data, which includes the barcode information recognition results, destination classification results, and package size and shape detection results of each package, analyze the barcode information recognition rate, destination classification accuracy and package storage efficiency, and optimize and adjust the destination classification and storage grid allocation based on the analysis results.

[0066] The present invention obtains image information of the package through the collection device of the transmission mechanism and pre-processes it, and identifies the barcode information of the package based on the deep convolution generation adversarial network; judges the barcode recognition confidence based on the barcode information, and if the recognition confidence is lower than the preset threshold, it is determined that the barcode information recognition of the package is abnormal, and the package with the abnormal barcode information recognition is sorted into a collection box through the pushing mechanism; if the recognition confidence is higher than or equal to the preset threshold, it is determined that the barcode information recognition of the package is normal, and the package with normal barcode information recognition is classified by destination, and the sorting mechanism is used to allocate the package to different conveyor belts according to the destination; detects the size and shape of the package with the same destination, optimizes the allocation of storage grids, and sorts the package with the same destination into the corresponding storage grid through the classification mechanism; collects sorting data, and optimizes and updates the barcode information recognition, destination classification and storage grid allocation according to the collected sorting data, so as to realize continuous optimization management of package sorting.

[0067] The barcode information of the package is identified through the transmission mechanism, and the packages with unidentifiable barcode information are sorted into the collection box through the pushing mechanism. The packages with identifiable barcode information are sorted into the corresponding conveyor belt according to the destination through the sorting mechanism. The express packages with the same destination are sorted into the corresponding storage grid according to their size through the classification mechanism, thereby realizing efficient package processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0069] Figure 1 It is a flow chart showing a smart logistics sorting method according to an embodiment of the present invention.

[0070] Figure 2 It is a structural diagram showing an intelligent logistics sorting system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0071] To make the objectives, technical solutions, and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It is apparent that the embodiments described are only some, not all, of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0072] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "an," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0073] It should be understood that although the terms "first," "second," "third," etc. may be used to describe "...," these "..." should not be limited to these terms. These terms are merely used to distinguish "...." For example, "first..." could also be referred to as "second...", and similarly, "second..." could also be referred to as "first..." without departing from the scope of the present invention.

[0074] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0075] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0076] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or device. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or device comprising the element.

[0077] Existing manual sorting is inefficient and labor-intensive. Intelligent sorting sorts all express parcels of different sizes with the same destination into the same container according to sorting instructions. Larger express parcels are stacked on top of smaller ones, which is inconvenient to take out and easily causes squeezing and damage to smaller parcels. At the same time, the existing intelligent sorting only sets the sorting device on one side of the conveyor belt, which requires a longer conveyor belt for conveying and sorting express parcels, occupies a large space, and the express parcels are transported a long distance on the conveyor belt, resulting in low sorting efficiency.

[0078] like Figure 1 As shown, the present invention discloses a smart logistics sorting method, which includes:

[0079] Step 1: The image information of the package is obtained and preprocessed through the collection device of the transmission mechanism, and the barcode information of the package is identified based on the deep convolutional generative adversarial network.

[0080] The package is moved to a fixed location via a conveyor mechanism, where a high-resolution camera or laser scanner captures an image of the package. The captured image undergoes preprocessing, first removing background noise. The image is then converted to grayscale and binarized to highlight barcode features. Techniques such as contrast adjustment and edge enhancement are used to improve legibility. Finally, barcode recognition is performed on the preprocessed image. A trained model is used to extract features and decode them into usable information, providing the foundation for subsequent processing and sorting.

[0081] In one embodiment, obtaining image information of the package and performing preprocessing in step 1 includes converting the obtained color image of the package into a grayscale image, removing noise through Gaussian filtering, and improving image contrast through histogram equalization.

[0082] The Canny edge detection algorithm is used to determine the position of the barcode on the package, and the ROI region of interest (ROI) containing only the barcode is extracted and cropped.

[0083] The cropped image is adjusted to a predetermined size and pixel-normalized, which is expressed as:

[0084] .

[0085] The mean μ and standard deviation σ of image I are expressed as:

[0086] ;

[0087] .

[0088] Where I is the cropped image of the ROI region of interest containing only the barcode, Inorm is the normalized image, M is the number of image rows, N is the number of image columns, μ is the mean of the image, and σ is the standard deviation of the image.

[0089] A series of image processing techniques significantly improves the accuracy and robustness of barcode recognition. Converting color images to grayscale simplifies data processing and reduces computational complexity. Gaussian filtering effectively removes noise, ensuring image clarity, and histogram equalization enhances contrast, making the barcode stand out more clearly. The Canny edge detection algorithm precisely locates the barcode and extracts the region of interest (ROI). Normalization further eliminates the effects of image size and brightness variations. The combination of these technologies not only improves recognition accuracy but also provides reliable data support for subsequent automated processing, ultimately enhancing the efficiency of the overall parcel sorting system.

[0090] In one embodiment, the recognition of the barcode information of the package based on the deep convolutional generative adversarial network in step 1 includes constructing a generative network G, inputting a random noise vector z, and generating a feature map of the barcode through multiple convolutional layers, batch normalization, and activation function ReLU, which is expressed as:

[0091] .

[0092] Where z is a random noise vector, BatchNorm(z) is a batch normalization operation used to standardize the input z, ReLU is an activation function, and Conv2D is a two-dimensional convolution operation used to extract spatial relationships in input features.

[0093] Construct the discriminant network D, input the standardized image Inorm, and extract features through multiple convolutional layers and LeakyReLU activation function, which can be expressed as:

[0094] .

[0095] Among them, Inorm is the normalized image, Conv2D is a two-dimensional convolution operation used to extract spatial information from the input features, LeakyReLU is the activation function, and the Sigmoid function outputs a value between 0 and 1, indicating the probability that the image is a real barcode.

[0096] The adversarial loss function is constructed based on the loss of real images and the loss of generated images, which is expressed as:

[0097] .

[0098] Among them, Ex~pdata(x) represents the expected value of the real data, Denote the loss of real images, Expressed as the loss of generating images, x represents the real image, pdata(x) is the distribution of real data, Ez~p z (z) represents the expected value of random noise z, G(z) is the image generated by the network output, and pz (z) is the distribution of noise.

[0099] The discriminant network and the generative network are iteratively trained based on an alternating optimization strategy, including minimizing the loss function L GAN Update the discriminant network and minimize the loss of generated images Update the generative network.

[0100] The feature map output by the trained generative network is processed through the convolution layer to extract key features, and the extracted features are mapped to the feature vector F using the fully connected layer, which is expressed as:

[0101] F=[f1,f2,…,fi,…,fn].

[0102] Here, fi represents the characteristics of the barcode, and the characteristics of the barcode at least include the type, size, and clarity of the barcode.

[0103] The Generative Adversarial Network (GAN) architecture effectively improves barcode generation and recognition performance. The generative network G uses random noise vectors to generate barcode feature maps, generating diverse barcode samples and enhancing the diversity of training data, thereby improving the model's generalization ability. The discriminative network D, through feature extraction and probabilistic output, accurately assesses the authenticity of generated images, promoting optimization of the generative network. The design of the adversarial loss function ensures that the generation and recognition processes mutually reinforce each other, further improving the quality of the generated images. Furthermore, the combination of feature extraction and fully connected layers enables the generated feature vectors to clearly represent the key information of the barcode, providing strong support for subsequent barcode recognition. This process effectively improves the recognition accuracy and robustness of the overall system.

[0104] High-quality image acquisition and precise preprocessing significantly improve the accuracy and efficiency of barcode recognition. High-resolution image acquisition ensures image clarity, while preprocessing techniques such as denoising, grayscaling, and binarization effectively eliminate interference and enhance the barcode's features. Furthermore, the use of a deep convolutional generative adversarial network (DCGAN) for recognition automatically learns complex barcode patterns, enhancing recognition reliability. This optimized process not only reduces the risk of misidentification but also provides accurate data support for subsequent package sorting and management, thereby improving overall logistics efficiency.

[0105] Step 2: Determine the barcode recognition confidence based on the barcode information. If the recognition confidence is lower than a preset threshold, the barcode recognition of the package is determined to be abnormal, and the process proceeds to Step 3. If the recognition confidence is higher than or equal to the preset threshold, the barcode recognition of the package is determined to be normal, and the process proceeds to Step 4.

[0106] The confidence level of the barcode information reflects the reliability of the recognition result. If the recognition confidence level falls below the set threshold, the system deems it an anomaly, marks the package, and directs it to a collection bin for further processing, such as sorting it. If the recognition confidence level is higher than or equal to the preset threshold, the system deems the recognition normal, confirms the package information, and proceeds with destination sorting and subsequent processing. This process ensures accurate package processing, reduces the risk of misidentification, and improves overall operational efficiency.

[0107] In one embodiment, the confidence level in step 2 includes defining a calculation formula for the confidence level C, which is expressed as:

[0108] .

[0109] Among them, C is the confidence of barcode recognition, ranging from 0 to 1, α is the parameter that controls the confidence growth rate, β is the baseline value used to adjust the offset of the confidence, T is the total length of barcode feature extraction, wi is the weight coefficient, and fi(t) is the value of barcode feature i at time t.

[0110] By comprehensively considering the diversity and importance of barcode features, a more accurate recognition confidence assessment is provided. By introducing appropriate parameters, not only can the confidence growth rate and offset be flexibly adjusted, but the impact of different barcode features can also be accommodated, making the model more robust in complex situations. The use of weighting coefficients ensures that the contribution of important features to the confidence score is properly reflected, improving recognition reliability. This mechanism effectively reduces the risk of misidentification and ensures the accuracy of subsequent processing, thereby optimizing the overall efficiency and accuracy of the package processing system.

[0111] A confidence assessment mechanism intelligently determines the effectiveness of barcode recognition, thereby improving the system's overall accuracy and reliability. Calculating confidence allows the system to clearly distinguish between normal and abnormal recognition, reducing the risk of misjudgment. If the recognition result is unreliable, the package is promptly flagged and entered into the exception handling process, ensuring the accuracy of subsequent operations. For packages that are correctly identified, the system can quickly classify them by destination, improving processing efficiency. This mechanism not only optimizes the package sorting process but also enhances the system's adaptability to complex environments, thereby improving the efficiency of overall logistics management.

[0112] Step 3: Control the pushing mechanism to sort the parcels with abnormal barcode information into a collection box.

[0113] This automated sorting system automatically sorts parcels with barcode anomalies. This process identifies parcels marked as abnormal and then accurately moves them to dedicated collection bins, ensuring they are no longer mixed with normal parcels. This automated sorting not only increases processing speed but also reduces the risk of manual error, ensuring the overall accuracy and efficiency of the logistics system. Furthermore, centralized management of abnormal parcels facilitates subsequent manual inspection and processing, improving system transparency and traceability, and further optimizing parcel management processes.

[0114] The automated sorting mechanism efficiently handles parcels with barcode identification anomalies, effectively reducing the incidence of human error. Abnormal parcels are quickly and accurately sorted into dedicated collection bins, preventing them from being mixed up with normal parcels, thereby ensuring the accuracy and smoothness of the logistics chain. Furthermore, this process centralizes the management of abnormal parcels, facilitating subsequent manual inspection and processing, and improving the controllability and transparency of the system. Overall, this step not only optimizes the parcel management process and enhances system reliability, but also improves overall operational efficiency, providing strong support for subsequent logistics processing.

[0115] Step 4: Classify the parcels whose barcode information is normal by their destination, and control the sorting mechanism to distribute the parcels to different conveyor belts according to their destination.

[0116] By analyzing the identified barcode information, the final destination of each package is determined. The sorting mechanism automatically assigns packages to the corresponding conveyor belts based on their destination, ensuring that each package is quickly and accurately directed to its intended transportation route. This process utilizes efficient sorting algorithms and automated equipment, not only improving overall processing speed but also reducing human intervention and the risk of misclassification. Through this intelligent sorting and allocation, the system optimizes logistics processes, improves delivery efficiency and accuracy, and ensures that packages reach their destinations on time.

[0117] In one embodiment, classifying the destination of the package in step 4 includes extracting and recognizing information in the barcode using an OCR optical character recognition module, and pre-processing the recognized barcode information into a feature vector.

[0118] The generated feature vector is input into a pre-trained multi-layer perceptron (MLP) classification model.

[0119] The output of the classification model is processed using the Softmax activation function to calculate the probability distribution of each destination category, which is expressed as:

[0120] .

[0121] Where zi is the unnormalized score output by the classification model, and P(yi) is the probability of predicting class i.

[0122] Find the index of the maximum value in the probability distribution, determine the destination category with the highest probability, and use that category as the final classification result for the package.

[0123] The combination of OCR technology and a multi-layer perceptron (MLP) model enables efficient and accurate barcode information processing and package classification. The OCR module quickly extracts barcode information and converts it into a feature vector suitable for analysis, while the pre-trained MLP model enhances classification accuracy by learning from historical data. Using the Softmax activation function to process the output clearly calculates the probability distribution of each destination category, enabling the system to reliably identify the destination with the highest probability. This automated processing not only increases classification speed and reduces the risk of human intervention errors, but also optimizes the overall logistics process, ensuring that packages are delivered to their destinations promptly and accurately, thereby improving customer satisfaction and operational efficiency.

[0124] The intelligent destination classification mechanism significantly improves the efficiency and accuracy of package processing. Leveraging information extracted by OCR and the powerful classification capabilities of the Multi-Layer Perceptron (MLP) model, the system can quickly and accurately assign packages to the appropriate conveyor belts, reducing the need for manual operation and reducing errors. Calculating probability distributions using the Softmax activation function ensures the scientific and reliable nature of classification decisions. This process not only speeds up logistics operations and optimizes resource utilization, but also improves customer satisfaction, ensures timely delivery of packages, and further enhances the competitiveness and efficiency of the overall logistics system.

[0125] Step 5: Detect the size and shape of the parcels with the same destination, optimize the allocation of storage cells, and control the sorting mechanism to sort the parcels with the same destination into the corresponding storage cells.

[0126] Sensors or cameras capture the physical characteristics of packages and analyze their size and shape. This data helps the system determine the optimal storage location for each package, ensuring that packages destined for the same destination are efficiently grouped together. A sorting mechanism then automatically sorts the packages into the appropriate storage compartments based on the analysis. This intelligent sorting and storage method not only improves storage space utilization but also facilitates subsequent retrieval and reduces the risk of package damage, thereby optimizing the entire logistics management process and enhancing operational accuracy and efficiency.

[0127] In one embodiment, detecting the size and shape of the same package in step 5 includes constructing a convolutional neural network (CNN) to identify the package shape and label the category. Constructing the convolutional neural network (CNN) includes constructing an input layer of the convolutional neural network (CNN) to receive a package image, extracting package image features through multiple convolutional layers, wherein the package image features include shape information and edge information, mapping the extracted package image features to the package category through a fully connected layer, and training the CNN model using a training dataset containing labeled package shapes and categories until the model converges.

[0128] Based on a genetic algorithm, selection, crossover, and mutation operations are used to generate a target storage cell allocation scheme based on the shape and category of the package, as well as the package arrival status and the storage cell occupancy status. This includes randomly generating multiple storage cell allocation schemes, evaluating each scheme based on storage efficiency, access speed, and space utilization, selecting multiple initial schemes based on fitness using a tournament, crossovering the selected multiple initial schemes to generate a new storage cell allocation scheme, and repeating the evaluation, selection, and crossover operations until the target storage cell allocation scheme is obtained.

[0129] Generate intelligent sorting instructions and control the sorting mechanism to send the packages to the corresponding storage cells, including generating sorting instructions according to the target storage cell allocation plan, sending the sorting instructions to the sorting mechanism, and controlling the automated system to sort the packages to the target storage cells according to the target storage cell allocation plan.

[0130] By combining a convolutional neural network (CNN) and a genetic algorithm, package shape recognition and intelligent storage compartment allocation are optimized. The CNN efficiently and accurately identifies package shape and category, ensuring accurate classification. The genetic algorithm, by simulating the process of natural selection, generates multiple storage compartment allocation schemes to optimize space utilization and storage efficiency. By generating intelligent sorting instructions, the system automatically routes packages to designated storage compartments, reducing manual intervention and improving processing speed and accuracy. This process not only enhances the intelligence of logistics management, but also effectively reduces the risk of package misplacement, optimizes overall operational efficiency, and provides strong support for timely and accurate delivery.

[0131] Intelligent detection and optimized storage compartment allocation significantly improves the efficiency and accuracy of package management. By analyzing package size and shape, the system optimizes storage space, ensuring that packages destined for the same destination are stored together for easy access. This intelligent sorting mechanism not only reduces the risk of package damage but also enhances accessibility and reduces the need for manual intervention. Furthermore, optimized storage compartment allocation streamlines the logistics process, improving overall operational efficiency and responsiveness, thereby providing customers with more timely and reliable service.

[0132] Step 6: Collect sorting data, which includes the barcode information recognition results, destination classification results, and package size and shape detection results of each package. Analyze the barcode information recognition rate, destination classification accuracy, and package storage efficiency, and optimize and adjust the destination classification and storage grid allocation based on the analysis results.

[0133] Collecting and analyzing data generated during the sorting process enables continuous optimization of management. The system regularly records the sorting status of each package, including barcode recognition accuracy, destination classification results, and storage bin usage. Analysis of this data can reveal potential issues in the recognition and classification process, helping to optimize barcode recognition algorithms, destination classification models, and storage bin allocation strategies. By incorporating this information into the system, the system can dynamically adjust its operations to continuously improve efficiency and accuracy, enabling the package sorting process to adapt to environmental changes, ensuring flexibility and efficiency in logistics management, and ultimately enhancing overall service levels and customer satisfaction.

[0134] In one embodiment, the relevant information of each package includes at least one of the following: barcode information, destination classification result, package size and shape, and storage compartment allocation.

[0135] Record each sorting situation, including at least one of the following: barcode recognition failure, destination classification error, package size mismatch.

[0136] Preprocess the collected data, including deduplication, cleaning, and filtering of invalid data.

[0137] The multi-layer perceptron (MLP) classification model is evaluated using the cross entropy loss function, expressed as:

[0138] .

[0139] Among them, y is the true label of the i-th sample, is the model's predicted probability for the i-th sample, and N is the number of samples.

[0140] The collected sorting data is added to the training set to form an incremental learning data set, and the model is continuously updated and retrained, which is expressed as:

[0141] .

[0142] Among them, θ new is the updated model parameter, θ old is the current model parameter, θ is the model parameter, η is the learning rate, is the gradient of the loss function with respect to the parameters.

[0143] The fitness function is established based on the collected sorting data, which is expressed as:

[0144] .

[0145] Among them, storage efficiency Eff= , S used is the used storage space, S total is the total capacity of the storage space, access speed , ∑Ts is the sum of the access time of the package, N is the total number of accessed packages, and α, β, γ are weight coefficients.

[0146] The weight coefficients α, β, and γ in the fitness function are adjusted through iterative learning to continuously optimize the storage grid allocation.

[0147] Through systematic data collection and analysis, precise optimization of the parcel sorting process is achieved. The collected sorting data not only covers the key characteristics of the parcels and the sorting status, but also ensures data accuracy through preprocessing. This enables the system to identify problems such as barcode recognition failures and destination classification errors, allowing for timely strategy adjustments. Based on the results of data mining analysis, the system continuously optimizes barcode recognition, destination classification, and storage bin allocation, improving overall sorting efficiency and accuracy. Through this continuous improvement process, logistics management has become more flexible and efficient, reducing error rates, enhancing customer satisfaction, and improving overall operational efficiency.

[0148] Continuous data collection and analysis form a dynamic optimization mechanism that continuously improves the parcel sorting process. The system promptly identifies and corrects issues such as barcode recognition failures and destination classification errors, thereby reducing errors and improving processing efficiency. The preprocessing stage ensures data accuracy, making the analysis results more reliable. Based on data mining analysis, the system can specifically adjust barcode recognition, destination classification, and storage bin allocation strategies to improve overall sorting efficiency and accuracy. This cyclical optimization not only enhances the adaptability of the logistics system but also significantly improves customer satisfaction, ultimately driving overall operational improvements and profitability growth for the company.

[0149] The present invention uses a transmission mechanism to identify the barcode information of the package, uses a pushing mechanism to sort the packages whose barcode information cannot be identified into a collection box, uses a sorting mechanism to sort the packages whose barcode information can be identified onto the corresponding conveyor belt according to the destination, and uses a classification mechanism to sort the express packages with the same destination into the corresponding storage grid according to the size, thereby realizing efficient package processing.

[0150] Corresponding to the intelligent logistics sorting method provided by the present invention, the present invention also provides an intelligent logistics sorting system. Figure 2As shown, the system includes: a transmission mechanism, a pushing mechanism, a sorting mechanism, a classification mechanism, a discrimination mechanism and an optimization mechanism.

[0151] The transmission mechanism is used to obtain image information of the package through the collection device of the transmission mechanism and perform preprocessing, and to identify the barcode information of the package based on the deep convolutional generative adversarial network.

[0152] The identification mechanism is used to determine the barcode recognition confidence based on the barcode information. If the recognition confidence is lower than a preset threshold, it is determined that the barcode information recognition of the package is abnormal. If the recognition confidence is higher than or equal to the preset threshold, it is determined that the barcode information recognition of the package is normal.

[0153] The pushing mechanism is used to sort the parcels with abnormal barcode information into the collection box.

[0154] The sorting mechanism is used to classify the destinations of packages with normal barcode information identification, and distribute the packages to different conveyor belts according to their destinations through the sorting mechanism.

[0155] The classification mechanism is used to detect the size and shape of packages with the same destination, optimize the allocation of storage cells, and sort the packages with the same destination into the corresponding storage cells through the classification mechanism.

[0156] The optimization mechanism is used to collect sorting data, which includes the barcode information recognition results, destination classification results, and package size and shape detection results of each package, analyze the barcode information recognition rate, destination classification accuracy and package storage efficiency, and optimize and adjust the destination classification and storage grid allocation based on the analysis results.

[0157] The present invention also provides an intelligent logistics sorting line, which consists of a transmission mechanism, a pushing mechanism, a sorting mechanism, a collecting mechanism and a classification mechanism.

[0158] The transmission mechanism consists of a first baffle, a second baffle, a first conveyor belt, a main conveyor wheel, a second rotating motor, a second fixed base plate, a third telescopic rod, a belt, a slave conveyor wheel, a second U-shaped bracket and a code reader.

[0159] A first conveyor belt is placed between the first baffle and the second baffle. The first conveyor belt has multiple equally spaced power rollers built in. One end of the power roller is rotatably connected to the first baffle through a bearing, and the other end of the power roller is rotatably connected to the second baffle through a bearing.

[0160] Preferably, the top surfaces of the first baffle and the second baffle are flush with the first conveyor belt or slightly lower than the height of the first conveyor belt.

[0161] Preferably, the first baffle, the second baffle and the first conveyor belt together constitute a main conveying line.

[0162] The main conveying line is supported by two groups of third telescopic rods, one end of one group of the third telescopic rods is connected to the bottom surface of the first baffle, one end of the other group of the third telescopic rods is connected to the bottom surface of the second baffle, and the other end of the third telescopic rod is provided with a second fixed bottom plate.

[0163] Preferably, a group of the third telescopic rods includes a plurality of third telescopic rods, and the plurality of third telescopic rods in the same group are arranged at equal distances along the main conveying line.

[0164] The secondary transmission wheel is placed on the side of the first baffle, and the wheel axle of the secondary transmission wheel is connected to one of the power rollers of the first conveyor belt.

[0165] The second rotating motor is placed on the ground and close to the slave transmission wheel.

[0166] The wheel shaft of the main transmission wheel is connected to the motor shaft of the second rotating motor, and the main transmission wheel and the slave transmission wheel are connected through a belt.

[0167] Two ends of the second U-shaped bracket are respectively placed on the top surfaces of the first baffle and the second baffle, and the second U-shaped bracket is close to the first main rotating roller.

[0168] A code reader is disposed on the inner top surface of the second U-shaped bracket.

[0169] The pushing mechanism consists of a collection box, an ultrasonic ranging module, a transmission baffle, a transmission plate, a second support rod, a cylinder, a second telescopic rod, an L-shaped push plate, a connecting block, a support plate, a fourth telescopic rod and a third fixed base plate.

[0170] One end of the fourth telescopic rod is respectively placed on the four corners of the bottom surface of the support plate, and the other end of the fourth telescopic rod is respectively placed on the top surface of the third fixed bottom plate. The support plate is in contact with the first baffle.

[0171] Preferably, a locking wheel is provided on the bottom surface of the third fixed base plate.

[0172] The cylinder is placed on the top surface of the support plate.

[0173] The L-shaped push plate contacts the top surface of the support plate and is close to the first baffle. A connecting block is provided on the side surface of the L-shaped push plate.

[0174] One end of the second telescopic rod is slidably placed in the cylinder, and the other end of the second telescopic rod is placed on the connecting block.

[0175] The collection box is placed on the ground and close to the second baffle. The collection box is a structure with an open top.

[0176] An ultrasonic distance measuring module is arranged on the inner side of the collection box, and the ultrasonic distance measuring module is close to the top surface of the collection box.

[0177] The transmission plate is located between the collection box and the second baffle, one end of the transmission plate is in contact with the second baffle, and the other end of the transmission plate is tilted downward and extends to above the collection box.

[0178] The transmission plate is close to the reader, and the L-shaped push plate corresponds to one end of the transmission plate.

[0179] Transmission baffles are vertically arranged on both sides of the top surface of the transmission plate, and the length of the transmission baffles is equal to the length of the transmission plate.

[0180] Second support rods are correspondingly arranged on both sides of the bottom surface of the transmission plate.

[0181] Preferably, there are multiple second support rods, and the multiple second support rods are equidistantly arranged along the length direction of the bottom surface of the transmission plate, and the lengths of the multiple second support rods gradually decrease from one end to the other end of the transmission plate.

[0182] The sorting mechanism consists of a first U-shaped bracket, a paddle motor, a connecting rod, a first telescopic rod, a support arm, a paddle, a camera and a rotating shaft.

[0183] Two ends of the first U-shaped bracket are respectively placed on the top surfaces of the first baffle and the second baffle, and are located above the first conveyor belt.

[0184] Preferably, there are multiple first U-shaped brackets, and the multiple first U-shaped brackets are arranged at equal intervals along the length direction of the first conveyor belt.

[0185] Cameras are respectively disposed on the top surfaces of the first U-shaped brackets.

[0186] A paddle motor is correspondingly disposed in the middle of the top surface of the first U-shaped bracket.

[0187] One end of the connecting rod is connected to the motor shaft of the paddle motor, and the other end of the connecting rod is placed on the side of one end of the first telescopic rod, and the other end of the first telescopic rod is placed in the middle of one side of the paddle.

[0188] The support arm is located directly below the connecting rod, and one end of the support arm is placed on the side of the connecting rod and close to the first telescopic rod.

[0189] One end of the rotating shaft is placed on the middle of the inner top surface of the first U-shaped bracket through a bearing, and the other end of the supporting arm is placed on the side of the rotating shaft through a through hole sleeve opened in the middle.

[0190] Preferably, a limit block is provided at the other end of the rotating shaft.

[0191] The collecting mechanism is composed of a second conveyor belt baffle, a connecting plate, a vertical plate, a first supporting rod and a second conveyor belt.

[0192] The two sets of second conveyor belts are respectively placed on both sides of the main conveyor line in a one-to-one correspondence.

[0193] Preferably, a group includes multiple second conveyor belts, and the multiple second conveyor belts in the same group are arranged equidistantly along the first conveyor belt. The multiple second conveyor belts in the same group correspond to the multiple dial plates one by one.

[0194] The second conveyor belt is equipped with conveyor belt rollers, and second conveyor belt baffles are respectively arranged on both sides of the second conveyor belt. The conveyor belt rollers and the second conveyor belt baffles are rotatably connected through bearings, and the second conveyor belt baffles are placed on the first support rod.

[0195] The height of the second conveyor belt baffle gradually decreases from one end to the other end.

[0196] Preferably, there are multiple first support rods, and the multiple support rods are equidistantly arranged along the side length direction of the second conveyor belt baffle, and the lengths of the multiple first support rods gradually decrease from one end to the other end of the second conveyor belt baffle.

[0197] Preferably, the height of the second conveyor belt gradually decreases from one end close to the main conveyor line to the other end.

[0198] The third rotating motor is placed on the side of one of the second conveyor belt baffles, and the motor shaft of the third rotating motor passes through a through hole opened on the side of one of the second conveyor belt baffles and is connected to one of the conveyor belt rollers of the second conveyor belt.

[0199] The express parcels falling on the second conveyor belt are conveyed by sliding due to gravity, or the express parcels are conveyed by rotating the second conveyor belt due to the third rotating motor.

[0200] The two ends of the connecting plate are respectively placed on the other end top surfaces of the two second conveyor belt baffles in the same group.

[0201] One end of the vertical plate is placed in the middle of the bottom surface of the connecting plate, and the other end of the vertical plate is close to the second conveyor belt.

[0202] The classification mechanism consists of a first fixed base plate, a third support rod, an arc-shaped fixing member, a first rotating motor, a rotating shaft, a first limiting plate, a classification collection basket, a partition plate, a limiting plate fixing leg and a second limiting plate.

[0203] The classification collection basket is placed on the ground, and one end of the classification collection basket is placed below the second conveyor belt baffle.

[0204] The classification collection basket is a structure with an open top surface.

[0205] The partition plate is placed in the classification collection basket, and the two sides of the partition plate are detachably connected to the two inner side surfaces of the classification collection basket respectively, the bottom surface of the partition plate is in contact with the bottom surface of the classification collection basket, and the top surface of the partition plate is flush with the top surface of the classification collection basket.

[0206] Preferably, there are multiple partition plates, and the multiple partition plates are arranged at non-equidistant intervals along the width direction of the classification collection basket. The multiple partition plates divide the classification collection basket into multiple storage compartments with different space sizes, and the widths of the multiple storage compartments gradually increase from one end to the other end of the classification collection basket.

[0207] The third support rod is close to the other end of the classification collection basket, and the length of the third support rod is greater than the height of the classification collection basket.

[0208] One end of the third support rod is placed on the first fixed bottom plate, the other end of the third support rod is placed on the bottom surface of the arc-shaped fixing member, and the first rotating motor is placed on the top surface of the arc-shaped fixing member.

[0209] Preferably, there are multiple groups of the third support rods, and the multiple groups of the third support rods are arranged without spacing along the length direction of the other end of the classification collection basket. There are two third support rods in each group, and the two third support rods in the same group correspond to the first rotating motors with opposite rotation directions.

[0210] The fixing foot of the limiting plate is close to one end of the classification collection basket and is located below the other end of the second conveyor belt baffle.

[0211] Preferably, there are multiple groups of the limiting plate fixing legs, and the multiple groups of the limiting plate fixing legs are arranged without spacing along the length direction of one end of the classification collection basket. The multiple groups of the limiting plate fixing legs and the multiple groups of the third support rods correspond one to one. There are two limiting plate fixing legs in each group, and the two sides of the second limiting plate correspond to the side faces of the two limiting plate fixing legs in the same group.

[0212] One end of the rotating shaft is connected to the motor shaft of the first rotating motor, and the other end of the rotating shaft is placed on the corresponding second limiting plate through a bearing. The height of the rotating shaft gradually increases from one end to the other end.

[0213] The plurality of groups of rotating shafts correspond to the plurality of vertical plates in a one-to-one manner, and each vertical plate is located between two adjacent groups of rotating shafts.

[0214] The distance between the two rotating shafts in the same group gradually decreases from one end close to the second limiting plate to the other end.

[0215] The first limiting plate is respectively sleeved on the side surfaces of the two rotating shafts in the same group through through holes opened at both ends, and bearings are respectively arranged between the first limiting plate and the two rotating shafts.

[0216] The intelligent logistics sorting line also includes an intelligent logistics sorting system, which includes a signal converter, a data processor and a controller.

[0217] The signal converter is placed on a side of the first baffle, the data processor is placed on a side of the first baffle, and the controller is placed on a side of the first baffle.

[0218] The camera, the reader, and the ultrasonic distance measurement module are connected to the signal converter via respective data lines.

[0219] The first rotary motor, the second rotary motor, the third rotary motor, the first telescopic rod, the second telescopic rod, the third telescopic rod, the fourth telescopic rod, the paddle motor, and the cylinder are respectively connected to the controller via data transmission lines.

[0220] The signal converter is connected to the data processor via a data transmission line, and the data processor is connected to the controller via a data transmission line.

[0221] The signal converter can convert the electrical signals of the image data collected by the camera, the barcode data of the express package collected by the reader, and the distance data collected by the ultrasonic ranging module into digital signals.

[0222] The controller can be implemented by one or at least two application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing instructions of the data processing device.

[0223] The data processor and the signal converter exchange information.

[0224] When the smart logistics sorting system is executed, it mainly implements the following steps:

[0225] Before sorting the express parcels, the smart logistics sorting system presets the destination of the express parcels corresponding to each classification collection basket in advance. For example, classification collection basket No. 1 corresponds to Jiangsu Province, classification collection basket No. 2 corresponds to Shaanxi Province, classification collection basket No. 3 corresponds to Zhejiang Province..., the smart logistics sorting system divides each classification collection basket into multiple storage compartments of different space sizes through multiple partitions according to the size of the express parcels in advance. For example: each classification collection basket is divided into storage compartment No. 1, storage compartment No. 2, storage compartment No. 3 according to the space from small to large..., and the distance between the two rotating shafts in the same group corresponding to each classification collection basket is preset in advance, so that the distance between each storage compartment and the two rotating shafts corresponds to each other, for example: a distance less than the preset value 1 corresponds to storage compartment No. 1, a distance less than the preset value 2 corresponds to storage compartment No. 2, a distance less than the preset value 3 corresponds to storage compartment No. 3...

[0226] During the process of sorting express parcels, the express parcels are placed on the first conveyor belt at intervals, the controller starts the second rotary motor, the second rotary motor drives the main conveying wheel to rotate, the main conveying wheel drives the slave conveying wheel to rotate, and the slave conveying wheel drives the first conveyor belt to rotate, so that the express parcels are conveyed on the first conveyor belt, and the reader collects the barcode data on the express parcels.

[0227] When the barcode reader cannot collect the barcode data of the express package and the control system cannot identify the destination in the barcode data, the express package continues to be conveyed forward by the first conveyor belt and approaches the cylinder. The controller controls the cylinder to start, and the cylinder drives the telescopic rod to extend. The telescopic rod drives the L-shaped push plate to quickly push the express package onto the transmission plate, and slides along the transmission plate into the collection box for subsequent re-sorting. The ultrasonic ranging module in the collection box monitors the internal situation of the collection box in real time. When the collection box is full, the ultrasonic ranging module sends a signal to the terminal to remind the staff to transfer the express packages in the collection box in time.

[0228] When the code reader is able to collect the barcode data of the express package, the collected barcode data is sent to the signal converter, the signal converter converts the electrical signal of the received barcode data into a digital signal and sends it to the data processor, the data processor processes the digital signal of the received barcode data, identifies the destination of the express package in the barcode data, finds the classification collection basket corresponding to the destination, and determines whether the paddle motor swings left or right according to whether the classification collection basket is on the left or right side of the first conveyor belt, the controller controls the camera corresponding to the classification collection basket to start, the camera collects images of the express package on the first conveyor belt, and sends the collected image data signal to the signal converter, the signal converter converts the electrical signal of the received image data signal into a digital signal and sends it to the data processor, the data processor processes the digital signal of the received image data, and when the express package appears in the image data, the controller starts the paddle motor, the first telescopic rod, the third rotating motor and The first rotating motor and the paddle motor drive the connecting rod and the first telescopic rod to swing a preset distance in the opposite direction of the determined swing direction, and the first telescopic rod extends downward so that the paddle reaches the side of the express package, and the paddle motor drives the paddle to swing according to the determined swing direction, pushing the express package to the corresponding second conveyor belt, and the third drive motor drives the second conveyor belt to rotate, and the multiple groups of first rotating motors corresponding to the second conveyor belt drive the corresponding multiple groups of rotating shafts to rotate. The transmission of the second conveyor belt and the separation of the vertical plate transmit the express package to one of the groups of rotating shafts, and the two rotating shafts of the same group with opposite rotation directions drive the express package forward along the two rotating shafts. The distance between the two rotating shafts gradually increases, causing the express package to fall from the gap between the two rotating shafts into the storage compartment corresponding to the classified collection basket, thereby realizing the classified storage of the express package, which can facilitate the subsequent loading work, so that when loading, the large express package is placed at the bottom first, and the small express package is stacked on the large express package, which can make full use of the space inside the vehicle and protect the express package.

[0229] Furthermore, a flexible gasket is placed on the inner bottom surface of the collection box. The flexible gasket is made of rubber, and the edge of the flexible gasket is connected to the edge of the inner bottom surface of the collection box.

[0230] Furthermore, a rubber strip is provided on the side of the L-shaped push plate, and the length of the rubber strip is equal to the length of the side of the L-shaped push plate.

[0231] Preferably, there are a plurality of rubber strips, and the plurality of rubber strips are arranged equidistantly along the length direction of the side surface of the L-shaped push plate.

[0232] The present invention uses a pushing mechanism to sort express parcels whose barcode information cannot be identified into a collection box, which can prevent the express parcels from staying on the first conveyor belt for a long time or entering inappropriate storage compartments, thereby improving sorting efficiency and sorting accuracy.

[0233] The sorting mechanism will sort the express parcels with identifiable barcode information onto the corresponding second conveyor belt according to their destinations, and the classification mechanism will sort the express parcels with the same destination into the corresponding storage compartments according to their sizes. This will enable the express parcels to be stored in a classified manner, facilitating subsequent loading work. When loading, large express parcels will be placed at the bottom first, and small express parcels will be stacked on top of the large express parcels. This will make full use of the space inside the vehicle, protect the express parcels, and reduce the squeezing damage of large express parcels to small express parcels.

[0234] By respectively arranging a plurality of collecting mechanisms and sorting mechanisms on the sides of the first baffle and the second baffle, the length of the first conveyor belt and the conveying time of the express parcels can be reduced, thereby improving the sorting efficiency.

[0235] It should be noted that the computer-readable medium described above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable storage media may include, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wire, optical cable, RF (radio frequency), or any suitable combination thereof.

[0236] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0237] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0238] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0239] The units involved in the embodiments described in this disclosure may be implemented in software or hardware, wherein the name of a unit does not necessarily limit the unit itself.

[0240] The above introduces the preferred embodiments of the present invention, which is intended to make the spirit of the present invention clearer and easier to understand, and is not intended to limit the present invention. Any modifications, replacements, and improvements made within the spirit and principles of the present invention should be included in the scope of protection outlined by the claims attached to the present invention.

Claims

1. A smart logistics sorting method, comprising: Step 1: The image information of the package is acquired and preprocessed through the collection device of the transmission mechanism, and the barcode information of the package is recognized based on the deep convolutional generative adversarial network; Step 2: Determine the barcode recognition confidence level based on the barcode information. If the recognition confidence level is lower than a preset threshold, the barcode recognition of the package is determined to be abnormal, and the process proceeds to Step 3. If the recognition confidence level is higher than or equal to the preset threshold, the barcode recognition of the package is determined to be normal, and the process proceeds to Step 4. Step 3: Control the pushing mechanism to sort the parcels with abnormal barcode information into a collection box; Step 4: Classify the parcels whose barcode information is normal by their destination, and control the sorting mechanism to distribute the parcels to different conveyor belts according to their destination; Step 5: Detect the size and shape of parcels with the same destination, optimize the allocation of storage cells, and control the sorting mechanism to sort the parcels with the same destination into the corresponding storage cells; Step 6: Collect sorting data, including barcode recognition results, destination classification results, and package size and shape detection results for each package. Analyze the barcode recognition rate, destination classification accuracy, and package storage efficiency, and optimize and adjust the destination classification and storage slot allocation based on the analysis results. In step 1, the barcode information of the package identified by the deep convolutional generative adversarial network includes: Construct a generative network G, input a random noise vector z, and generate a feature map of the barcode through multiple convolutional layers, batch normalization, and activation function ReLU, which is expressed as: ； Where z is a random noise vector, BatchNorm(z) is a batch normalization operation used to standardize the input z, ReLU is an activation function, and Conv2D is a two-dimensional convolution operation used to extract spatial relationships in the input features. Construct the discriminant network D, input the standardized image Inorm, and extract features through multiple convolutional layers and LeakyReLU activation function, which can be expressed as: ； Where Inorm is the normalized image, Conv2D is a two-dimensional convolution operation used to extract spatial information from the input features, LeakyReLU is the activation function, and the Sigmoid function outputs a value between 0 and 1, indicating the probability that the image is a real barcode. The adversarial loss function is constructed based on the loss of real images and the loss of generated images, which is expressed as: ； Among them, Ex~pdata(x) represents the expected value of the real data, Denote the loss of real images, Expressed as the loss of generating images, x represents the real image, pdata(x) is the distribution of real data, Ez~p z (z) represents the expected value of random noise z, G(z) is the image generated by the network output, and p z (z) is the distribution of noise; The discriminant network and the generative network are iteratively trained based on an alternating optimization strategy, including minimizing the loss function L GAN Update the discriminant network and minimize the loss of generated images Update the generative network; The feature map output by the trained generative network is processed through the convolution layer to extract key features, and the extracted features are mapped to the feature vector F using the fully connected layer, which is expressed as: F=[f1,f2,…,fi,…,fn]; Wherein, fi represents the characteristics of the barcode, and the characteristics of the barcode include at least the type, size, and clarity of the barcode; The destination classification of the package in step 4 includes: The information in the barcode is extracted and recognized through the OCR optical character recognition module, and the recognized barcode information is pre-processed and converted into a feature vector; Input the generated feature vector into the pre-trained multi-layer perceptron MLP classification model; The output of the classification model is processed using the Softmax activation function to calculate the probability distribution of each destination category, which is expressed as: ； Where zi is the unnormalized score output by the classification model, and P(yi) is the probability of predicting category i; Find the index of the maximum value in the probability distribution, determine the destination category with the highest probability, and use that category as the final classification result for the package; The sorting data in step 6 also includes at least one of the following: package-related information, sorting status; Relevant information about each package, including at least one of the following: barcode information, destination classification result, package size and shape, and storage compartment allocation; Record each sorting situation, including at least one of the following: barcode recognition failure, destination classification error, package size mismatch; Preprocess the collected data, including deduplication, cleaning, and filtering of invalid data; The cross entropy loss function is used to evaluate the multi-layer perceptron MLP classification model, which is expressed as: ； Among them, y is the true label of the i-th sample, is the model's predicted probability for the i-th sample, and N is the number of samples; The collected sorting data is added to the training set to form an incremental learning data set, and the model is continuously updated and retrained, which is expressed as: ； Among them, θ new is the updated model parameter, θ old is the current model parameter, θ is the model parameter, η is the learning rate, is the gradient of the loss function with respect to the parameters; The fitness function is established based on the collected sorting data, which is expressed as: ； Among them, storage efficiency Eff= , S used is the used storage space, S total is the total capacity of the storage space, access speed , ∑Ts is the sum of the access time of the package, N is the total number of accessed packages, α, β, γ are weight coefficients; The weight coefficients α, β, and γ in the fitness function are adjusted through iterative learning to optimize the storage grid allocation.

2. The intelligent logistics sorting method according to claim 1, characterized in that: The image information of the package is obtained and preprocessed in step 1, including: The obtained parcel color image is converted into a grayscale image, noise is removed by Gaussian filtering, and image contrast is improved by histogram equalization; The Canny edge detection algorithm is used to determine the location of the barcode on the package, and the ROI region of interest (ROI) containing only the barcode is extracted and cropped. The cropped image is adjusted to a predetermined size and pixel normalized, which is expressed as: ； The mean μ and standard deviation σ of image I are expressed as: ， ； Where I is the cropped image of the ROI region of interest containing only the barcode, Inorm is the normalized image, M is the number of image rows, N is the number of image columns, μ is the mean of the image, and σ is the standard deviation of the image.

3. The intelligent logistics sorting method according to claim 1, characterized in that: Detecting the size and shape of the package in step 5 includes: Constructing a convolutional neural network (CNN) to identify package shapes and label categories, wherein constructing the convolutional neural network (CNN) includes constructing an input layer of the convolutional neural network (CNN) to receive a package image, extracting package image features through multiple convolutional layers, wherein the package image features include shape information and edge information, mapping the extracted package image features to package categories through a fully connected layer, and training the CNN model using a training dataset containing labeled package shapes and categories until the model converges; Based on a genetic algorithm, a target storage slot allocation scheme is generated using selection and crossover operations according to the shape and category of the packages, the arrival status of the packages, and the occupancy status of the storage slots. The scheme includes randomly generating multiple storage slot allocation schemes, evaluating each scheme based on storage efficiency, access speed, and space utilization, selecting multiple initial schemes based on fitness using a tournament, crossovering the selected multiple initial schemes to generate new storage slot allocation schemes, and repeating the evaluation, selection, and crossover operations until the target storage slot allocation scheme is obtained. Generate intelligent sorting instructions and control the sorting mechanism to send the packages to the corresponding storage cells, including generating sorting instructions according to the target storage cell allocation plan, sending the sorting instructions to the sorting mechanism, and controlling the automated system to sort the packages to the target storage cells according to the target storage cell allocation plan.

4. An intelligent logistics sorting system, comprising: Transmission mechanism, pushing mechanism, sorting mechanism, classification mechanism, identification mechanism and optimization mechanism; The transmission mechanism is used to obtain image information of the package through the collection device of the transmission mechanism and perform preprocessing, and recognize the barcode information of the package based on the deep convolutional generative adversarial network; The identification mechanism is used to determine the barcode recognition confidence level based on the barcode information. If the recognition confidence level is lower than a preset threshold, it is determined that the barcode information recognition of the package is abnormal. If the recognition confidence level is higher than or equal to the preset threshold, it is determined that the barcode information recognition of the package is normal. The pushing mechanism is used to sort the parcels with abnormal barcode information into the collection box; The sorting mechanism is used to classify the parcels with normal barcode information by destination, and distribute the parcels to different conveyor belts according to their destinations; The sorting mechanism is used to detect the size and shape of parcels with the same destination, optimize the allocation of storage cells, and sort the parcels with the same destination into the corresponding storage cells through the sorting mechanism; The optimization mechanism is used to collect sorting data, including the barcode information recognition results of each package, the destination classification results, and the package size and shape detection results, analyze the barcode information recognition rate, destination classification accuracy and package storage efficiency, and optimize and adjust the destination classification and storage grid allocation based on the analysis results; The barcode information of the package identified based on the deep convolutional generative adversarial network includes: Construct a generative network G, input a random noise vector z, and generate a feature map of the barcode through multiple convolutional layers, batch normalization, and activation function ReLU, which is expressed as: ； Where z is a random noise vector, BatchNorm(z) is a batch normalization operation used to standardize the input z, ReLU is an activation function, and Conv2D is a two-dimensional convolution operation used to extract spatial relationships in the input features. Construct the discriminant network D, input the standardized image Inorm, and extract features through multiple convolutional layers and LeakyReLU activation function, which can be expressed as: ； Where Inorm is the normalized image, Conv2D is a two-dimensional convolution operation used to extract spatial information from the input features, LeakyReLU is the activation function, and the Sigmoid function outputs a value between 0 and 1, indicating the probability that the image is a real barcode. The adversarial loss function is constructed based on the loss of real images and the loss of generated images, which is expressed as: ； Among them, Ex~pdata(x) represents the expected value of the real data, Denote the loss of real images, Expressed as the loss of generating images, x represents the real image, pdata(x) is the distribution of real data, Ez~p z (z) represents the expected value of random noise z, G(z) is the image generated by the network output, and p z (z) is the distribution of noise; The discriminant network and the generative network are iteratively trained based on an alternating optimization strategy, including minimizing the loss function L GAN Update the discriminant network and minimize the loss of generated images Update the generative network; The feature map output by the trained generative network is processed through the convolution layer to extract key features, and the extracted features are mapped to the feature vector F using the fully connected layer, which is expressed as: F=[f1,f2,…,fi,…,fn]; Wherein, fi represents the characteristics of the barcode, and the characteristics of the barcode include at least the type, size, and clarity of the barcode; The destination classification of packages includes: The information in the barcode is extracted and recognized through the OCR optical character recognition module, and the recognized barcode information is pre-processed and converted into a feature vector; Input the generated feature vector into the pre-trained multi-layer perceptron MLP classification model; The output of the classification model is processed using the Softmax activation function to calculate the probability distribution of each destination category, which is expressed as: ； Where zi is the unnormalized score output by the classification model, and P(yi) is the probability of predicting category i; Find the index of the maximum value in the probability distribution, determine the destination category with the highest probability, and use that category as the final classification result for the package; The sorting data also includes at least one of the following: package-related information and sorting status; Relevant information about each package, including at least one of the following: barcode information, destination classification result, package size and shape, and storage compartment allocation; Record each sorting situation, including at least one of the following: barcode recognition failure, destination classification error, package size mismatch; Preprocess the collected data, including deduplication, cleaning, and filtering of invalid data; The cross entropy loss function is used to evaluate the multi-layer perceptron MLP classification model, which is expressed as: ； Among them, y is the true label of the i-th sample, is the model's predicted probability for the i-th sample, and N is the number of samples; The collected sorting data is added to the training set to form an incremental learning data set, and the model is continuously updated and retrained, which is expressed as: ； Among them, θ new is the updated model parameter, θ old is the current model parameter, θ is the model parameter, η is the learning rate, is the gradient of the loss function with respect to the parameters; The fitness function is established based on the collected sorting data, which is expressed as: ； Among them, storage efficiency Eff= , S used is the used storage space, S total is the total capacity of the storage space, access speed , ∑Ts is the sum of the access time of the package, N is the total number of accessed packages, α, β, γ are weight coefficients; The weight coefficients α, β, and γ in the fitness function are adjusted through iterative learning to optimize the storage grid allocation.

5. A smart logistics sorting device, comprising: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, together with the at least one processor, enable the apparatus to perform the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 3 when executed by a processor.

Images