Full-automatic logistics weighing and visual sorting method and system

Abstract

The invention relates to the technical field of logistics management, in particular to a full-automatic logistics weighing and visual sorting method and system.The method comprises the steps that transmission image data and current weighing signals of logistics packages in a weighing area are obtained, and the transmission image data comprise logistics package position information and package feature information; inputting the transmission image data into a preset convolutional neural network, and generating a transmission bump feature of the logistics package; determining a weight compensation coefficient of the logistics package through the transmission jolt characteristic of the logistics package; and the actual weight of the logistics parcel is determined based on the weight compensation coefficient of the logistics parcel and the current weighing signal, and parcel sorting is performed according to the actual weight of the logistics parcel, so that the problem of inaccurate weighing caused by transmission jolt in a traditional weighing technology is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of logistics management, and particularly relates to a full-automatic logistics weighing and visual sorting method and system. BACKGROUND

[0002] In the modern express logistics industry, the sorting efficiency and accuracy of packages directly affect the operation effect of the overall logistics chain. Currently, the sorting of express packages can be determined by measuring the weight of the packages to determine their classification and transportation path. However, in actual operation, the express packages often experience jolts or vibrations when moving on the conveying belt, and these dynamic factors can cause fluctuations and instability in the weighing signal, thereby affecting the accuracy of the weighing result. This inaccuracy can not only lead to incorrect classification of packages, but also cause subsequent delays in logistics and customer complaints.

[0003] In addition, the shape, size and other characteristics of the packages can also affect their motion state during transmission, further exacerbating the fluctuations in the weighing signal. Traditional weighing methods often cannot effectively identify and compensate for these dynamic disturbances, limiting the precision and reliability of the sorting system.

[0004] The above content is only used to assist in understanding the technical solutions of the present application and does not represent an acknowledgement that the above content is prior art. SUMMARY

[0005] The main purpose of the present application is to provide a full-automatic logistics weighing and visual sorting method and system, which aims to solve the technical problem that the current express packages can be sorted by weighing, but the express packages will experience jolts during transmission, causing unstable weighing signals and affecting the sorting effect.

[0006] To achieve the above purpose, the present application provides a full-automatic logistics weighing and visual sorting method, which comprises: acquiring transmission image data of a logistics package in a weighing area and a current weighing signal, the transmission image data including logistics package position information and package characteristic information; inputting the transmission image data into a preset convolutional neural network to generate transmission jolt characteristics of the logistics package; determining a weight compensation coefficient of the logistics package through the transmission jolt characteristics of the logistics package; determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, and sorting the package according to the actual weight of the logistics package.

[0007] Optionally, the determination of the weight compensation coefficient of the logistics package through the transmission jolt characteristics of the logistics package comprises: extracting features of the transmission bumping characteristics of the logistics package to obtain a frequency of periodic displacement of the package and a rotation angle of a bounding box of the package; Based on the frequency of periodic displacement of the package and the rotation angle of the bounding box of the package, the overall bumping intensity of the logistics package is determined by a preset bumping intensity calculation formula. Based on the overall bumping intensity of the logistics package, a weight compensation coefficient of the logistics package is obtained by fitting historical data.

[0008] Optionally, the preset bumping intensity calculation formula is:

[0009] In the formula, is the overall bumping intensity of the logistics package, is the frequency of periodic displacement of the package, is the maximum vibration frequency, is the weight coefficient of the vibration frequency feature in the overall bumping intensity S, is the rotation angle of the bounding box of the package, is the weight coefficient of the tilt angle feature in the overall bumping intensity S, is the maximum tilt angle.

[0010] Optionally, the inputting of the transmission image data into the preset convolutional neural network to generate the transmission bumping characteristics of the logistics package comprises: The transmission image data is subjected to grayscale conversion and normalization processing, and is input into the preset convolutional neural network; Feature extraction is performed through convolution and pooling operations; The extracted features are classified through a fully connected layer, and a plurality of transmission bumping feature labels and corresponding confidence scores are output; The transmission bumping characteristics of the logistics package are determined according to the transmission bumping feature labels and the corresponding confidence scores.

[0011] Optionally, the formula of the preset convolutional neural network is:

[0012] In the formula, is the transmission image data after grayscale conversion and normalization processing, is the transmission bumping characteristics of the logistics package output by the preset neural network, including transmission bumping feature labels and corresponding confidence scores; and respectively represent the weights and biases of the convolutional layer; represents the convolution operation; represents the activation function; represents the pooling operation; Represents the flattening operation, which flattens the multi-dimensional feature map into a one-dimensional vector; and denote the weights and biases of the fully connected layer respectively.

[0013] Optionally, before determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, the method further includes: Obtain the segmented slope of the weighing signal line of the logistics package in the weighing area; Comparing the segmented slope of the weighing signal broken line with a preset reference threshold to obtain a time period for logistics weighing compensation; Accordingly, the determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal includes: The actual weight of the logistics package is determined based on the time period of the logistics weighing compensation, the weight compensation coefficient of the logistics package and the current weighing signal.

[0014] Optionally, the preset reference thresholds include a first reference threshold, a second reference threshold, and a third reference threshold, and the step of comparing the segmented slopes of the weighing signal broken line with the preset reference thresholds to obtain the time period for logistics weighing compensation includes: Obtain the slope of the weighing signal line of the current logistics package before and after any moment, and record them as the first slope and the second slope respectively; If the difference between the first slope and the second slope is greater than the first reference threshold, and the second slope is less than the second reference threshold, then this moment is the start moment of the time period; If the difference between the first slope and the second slope is greater than the third reference threshold, and the first slope is less than the second reference threshold, then this moment is the end moment of the time period; The time period for logistics weighing compensation is determined based on the obtained start time and end time.

[0015] In addition, to achieve the above-mentioned purpose, the present invention also provides a fully automatic logistics weighing and visual sorting system, which includes: An image acquisition module is used to obtain the transmission image data of the logistics package in the weighing area and the current weighing signal. The transmission image data includes the logistics package location information and package feature information; A feature extraction module is used to input the transmission image data into a preset convolutional neural network to generate transmission bumpy features of the logistics package; A compensation calculation module, configured to determine a weight compensation coefficient of the logistics package based on a transmission bumpy characteristic of the logistics package; A weight sorting module is configured to determine the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, and to sort the package according to the actual weight of the logistics package.

[0016] In addition, to achieve the above object, the present application also provides a full-automatic logistics weighing and visual sorting device, which comprises a memory, a processor and a full-automatic logistics weighing and visual sorting program stored in the memory and executable on the processor, and the full-automatic logistics weighing and visual sorting program is configured to implement the steps of the full-automatic logistics weighing and visual sorting method according to any one of the above.

[0017] In addition, to achieve the above object, the present application also provides a storage medium, which stores a full-automatic logistics weighing and visual sorting program, and the full-automatic logistics weighing and visual sorting program implements the steps of the full-automatic logistics weighing and visual sorting method according to any one of the above when executed by a processor.

[0018] The present application provides a full-automatic logistics weighing and visual sorting method, which comprises the following steps: acquiring the transmission image data of a logistics package in a weighing area and a current weighing signal, wherein the transmission image data comprises logistics package position information and package feature information; inputting the transmission image data into a preset convolutional neural network to generate transmission bump features of the logistics package; determining the weight compensation coefficient of the logistics package through the transmission bump features of the logistics package; determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, and sorting the package according to the actual weight of the logistics package, thereby solving the problem of inaccurate weighing caused by transmission bump in the conventional weighing technology. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a full-automatic logistics weighing and visual sorting device structure schematic diagram of the hardware running environment involved in the embodiment scheme of the present application; Figure 2 is a flowchart of the first embodiment of the full-automatic logistics weighing and visual sorting method of the present application; Figure 3 is a flowchart of the second embodiment of the full-automatic logistics weighing and visual sorting method of the present application; Figure 4 is a flowchart of the third embodiment of the full-automatic logistics weighing and visual sorting method of the present application; Figure 5 is a structure block diagram of the first embodiment of the full-automatic logistics weighing and visual sorting system of the present application.

[0020] The implementation of the object, functional characteristics and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0021] It should be understood that the specific embodiments described herein are merely exemplary and do not limit the application.

[0022] Referring to Figure 1 , Figure 1 The structural schematic diagram of the full-automatic logistics weighing and visual sorting equipment related to the hardware running environment of the embodiment of the application.

[0023] As Figure 1 shown, the full-automatic logistics weighing and visual sorting equipment can include a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to realize the connection and communication between the components. The user interface 1003 can include a display screen (Display), and the optional user interface 1003 can also include a standard wired interface, a wireless interface, and the wired interface of the user interface 1003 can be a USB interface in the application. The network interface 1004 can optionally include a standard wired interface, a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 1005 can be a high-speed random access memory (RAM) memory, and can also be a stable memory (Non-volatile Memory, NVM), such as a disk memory. The memory 1005 can also be an independent storage device from the aforementioned processor 1001.

[0024] Those skilled in the art can understand Figure 1 that the structure shown in the figure does not constitute a limitation on the full-automatic logistics weighing and visual sorting equipment, and can include more or fewer components than the figure, or combine certain components, or different component arrangements.

[0025] As Figure 1 shown, the memory 1005 as a computer storage medium can include an operating system, a network communication module, a user interface module, and a full-automatic logistics weighing and visual sorting program.

[0026] In Figure 1The network interface 1004 is mainly used for connecting a background server and communicating data with the background server in the full-automatic logistics weighing and visual sorting device shown; the user interface 1003 is mainly used for connecting an external device; the full-automatic logistics weighing and visual sorting device calls a full-automatic logistics weighing and visual sorting program stored in the memory 1005 through the processor 1001, and executes the full-automatic logistics weighing and visual sorting method provided by the embodiment of the application.

[0027] Based on the above hardware structure, the embodiment of the full-automatic logistics weighing and visual sorting method of the application is provided.

[0028] Reference Figure 2 , Figure 2 The flowchart of the first embodiment of the full-automatic logistics weighing and visual sorting method of the application is provided.

[0029] In the first embodiment, the full-automatic logistics weighing and visual sorting method comprises the following steps: S10: acquiring transmission image data of a logistics parcel in a weighing area and a current weighing signal, the transmission image data comprising logistics parcel position information and parcel feature information.

[0030] It should be noted that the execution subject of the embodiment can be a comprehensive full-automatic logistics weighing and visual sorting system, which can integrate hardware devices, software systems and execution processes, and realize high-precision weighing and automatic sorting of logistics parcels through collaborative work of multiple modules, or other electronic devices capable of realizing the above functions, and the embodiment does not limit this.

[0031] It should be noted that the logistics parcel refers to a parcel or cargo that needs to be sorted and transported in the logistics transportation process, and usually has different shapes, sizes, weights and materials. The weighing area refers to a specific area on the logistics conveying belt, which is equipped with a weighing sensor and a high-precision camera for weighing and image acquisition of the parcel passing through the area. The transmission image data refers to the image information of the logistics parcel in the weighing area collected by the camera, including the appearance, shape, size, color and other visual features of the parcel. The logistics parcel position information refers to the specific position coordinates and motion trajectory of the parcel in the weighing area, which is usually extracted through image processing technology, and is used to determine the real-time position of the parcel on the conveying belt. The parcel feature information refers to the visual features of the parcel, such as shape, size, surface texture, color, etc., which can be extracted through image analysis technology for subsequent parcel identification and classification. The current weighing signal refers to the weight data of the parcel collected by the weighing sensor in real time, which can be output in the form of an electrical signal, reflecting the weight information of the parcel in the weighing area.

[0032] In a specific implementation, high-precision cameras and weighing sensors are installed in the weighing area of the logistics conveyor belt, and the equipment is ensured to be in normal working condition. Initialize the image acquisition module and the weighing signal acquisition module, set the sampling frequency and data format. When the logistics package enters the weighing area, the camera starts to collect the transmission image data of the package. The collected image data includes the front, side and top views of the package to ensure that complete package feature information is obtained. The weighing sensor collects the weight signal of the package in real time and uploads the data to the signal processing module. The collection of the weighing signal is synchronized with the collection of the image data to ensure the time consistency of the data. The collected image data is analyzed using image processing algorithms to extract the position information and feature information of the package. Through target detection and tracking algorithms, the center coordinates and motion trajectory of the package in the weighing area are determined. Through image segmentation and feature extraction algorithms, the visual features such as shape, size and color of the package are obtained. The collected transmission image data (including position information and feature information) and the current weighing signal are synchronized and stored in the database of the system for subsequent processing. The processed image data and weighing signal are transmitted to the deep learning module and dynamic compensation module of the system to generate transmission bump features and calculate weight compensation coefficients.

[0033] S20: input the transmission image data into a preset convolutional neural network to generate transmission bump features of the logistics package.

[0034] S30: determine the weight compensation coefficient of the logistics package through the transmission bump features of the logistics package.

[0035] It should be noted that the preset convolutional neural network (CNN) is a deep learning model specially designed for processing image data. Through multiple layers of convolution and pooling operations, CNN can extract high-level features from images for identifying and analyzing the transmission state of the package. Transmission bump features refer to dynamic features of the logistics package caused by bumps during transmission, including the amplitude, frequency, direction and other information of the bumps. These features reflect the motion state of the package during transmission. The weight compensation coefficient is a coefficient for correcting the weighing signal error. By analyzing the transmission bump features, the system can calculate the weighing error caused by bumps and generate the corresponding compensation coefficient to improve the weighing accuracy.

[0036] In a specific implementation, the transmission image data collected from the weighing area, including the location information and feature information of the parcel, is preprocessed, such as image normalization, denoising and enhancement, to ensure the quality of the input data. The preprocessed transmission image data is input into a preset convolutional neural network (CNN). The CNN model has been trained and can identify the dynamic behavior of the parcel during transmission. The CNN model extracts the transmission jolt features of the parcel from the image data through multiple layers of convolution and pooling operations. These features can include jolt amplitude, jolt frequency and motion direction, where jolt amplitude refers to the motion amplitude of the parcel in the vertical direction. Jolt frequency refers to the number of jolts per unit time of the parcel. Motion direction refers to the motion trend of the parcel in the horizontal direction, and the extracted features reflect the dynamic state of the parcel during transmission. According to the extracted transmission jolt features, combined with the kinematic model of the parcel, the weighing signal error caused by jolting is calculated. Through a preset compensation algorithm, a weight compensation coefficient of the logistics parcel is generated. The calculation of the compensation coefficient can include the following steps: analyzing the influence of jolt amplitude and frequency on the weighing signal; adjusting the compensation coefficient according to the motion direction of the parcel; dynamically updating the compensation coefficient in combination with historical data and real-time data. The generated transmission jolt features and weight compensation coefficient are output to the dynamic compensation module of the system for subsequent correction of the weighing signal. At the same time, the relevant data is stored in the database of the system for subsequent analysis and optimization.

[0037] S40: determining the actual weight of the logistics parcel based on the weight compensation coefficient of the logistics parcel and the current weighing signal, and performing parcel sorting according to the actual weight of the logistics parcel.

[0038] Specifically, the weight signal of the logistics package is collected in real time from the weighing sensor; the weight compensation coefficient generated based on the transmission bump feature is read from the dynamic compensation module. The weighing signal is corrected using the weight compensation coefficient, and the actual weight of the logistics package is calculated. The formula is: actual weight = current weighing signal x weight compensation coefficient, and the calculation process is ensured to be real-time to meet the efficiency requirements of logistics sorting. The actual weight calculated is verified to check whether it is within a reasonable range (such as matching the size and type of the package). If an abnormal value (such as too large or too small weight) is found, a data correction mechanism is started to re-collect the weighing signal or adjust the compensation coefficient. According to the actual weight of the logistics package, combined with the pre-set sorting rules, a sorting strategy is developed. For example: the packages are divided into light, medium and heavy categories according to weight, and different sorting paths or destinations are selected. The sorting rules can be flexibly adjusted according to the logistics requirements, such as prioritizing the processing of overweight packages or special packages. The sorting strategy is converted into specific sorting instructions and sent to automated sorting equipment (such as mechanical arms, sorting robots or sorting slides). The accuracy and real-time nature of the sorting instructions are ensured to avoid package accumulation or mis-sorting. The automated sorting equipment sorts the logistics package to the corresponding transport path or destination according to the sorting instructions. During the sorting process, the system monitors the sorting status in real time to ensure that each package is correctly sorted. The actual weight of the logistics package, the sorting result and the sorting time are recorded in the database of the system for subsequent analysis and optimization. If an abnormality occurs during the sorting process (such as unsorted or mis-sorted packages), the system will generate an alarm and feedback to the operator for processing. According to the sorting data and feedback information, the weight compensation algorithm and the sorting strategy are optimized to further improve the sorting accuracy and efficiency of the system. The system parameters and models are updated regularly to adapt to changes in logistics requirements and the emergence of new package types.

[0039] The embodiment provides a full-automatic logistics weighing and visual sorting method, which comprises the following steps: acquiring transmission image data of a logistics package in a weighing area and a current weighing signal, wherein the transmission image data comprises position information and feature information of the logistics package; inputting the transmission image data into a preset convolutional neural network to generate transmission bump features of the logistics package; determining a weight compensation coefficient of the logistics package through the transmission bump features of the logistics package; and determining an actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, and sorting the package according to the actual weight of the logistics package, thereby solving the problem of inaccurate weighing caused by transmission bump in the traditional weighing technology.

[0040] Reference Figure 3 , Figure 3 It is a flowchart of a second embodiment of the full-automatic logistics weighing and visual sorting method, wherein the weight compensation coefficient of the logistics package is determined through the transmission bump features of the logistics package, which comprises the following steps: S301: Feature extraction of the transport jolt characteristics of the logistics package to obtain the frequency of package periodic displacement and the rotation angle of the package bounding box; S302: Based on the frequency of the package periodic displacement and the rotation angle of the package bounding box, determine the comprehensive jolt intensity of the logistics package through a preset jolt intensity calculation formula; S303: Based on the comprehensive jolt intensity of the logistics package, obtain the weight compensation coefficient of the logistics package through historical data fitting.

[0041] It should be noted that the frequency of the package periodic displacement refers to the frequency of the periodic up-and-down or left-and-right motion of the logistics package due to jolt during transportation. It reflects the vibration characteristics of the package on the conveying belt, and is measured in hertz (Hz). The rotation angle of the package bounding box refers to the rotation angle of the package due to jolt during transportation. Through image processing technology, the rotation angle of the package bounding box can be extracted, reflecting the attitude change of the package during transportation. The comprehensive jolt intensity is a quantitative index used to describe the overall jolt degree of the logistics package during transportation. It combines the frequency of the package periodic displacement and the rotation angle of the bounding box, reflecting the jolt state of the package. The preset jolt intensity calculation formula is a mathematical formula used to calculate the comprehensive jolt intensity based on the frequency of the package periodic displacement and the rotation angle of the bounding box. Historical data fitting refers to establishing the mapping relationship between jolt intensity and weight compensation coefficient based on historical weighing data and jolt characteristic data through statistical or machine learning methods.

[0042] In specific implementation, by analyzing the motion trajectory of the package on the conveying belt, the frequency of its up-and-down or left-and-right vibration is calculated. Through image processing technology, the rotation angle of the package bounding box is extracted, reflecting the attitude change of the package. The extracted frequency of the package periodic displacement and the rotation angle of the bounding box are input into the preset jolt intensity calculation formula to calculate the comprehensive jolt intensity of the logistics package. Based on historical weighing data and jolt characteristic data, a mapping relationship between the comprehensive jolt intensity and the weight compensation coefficient is established through statistical or machine learning methods. For example, linear regression, support vector machine (SVM) or neural network model is used to fit the relationship between jolt intensity and compensation coefficient. The calculated comprehensive jolt intensity is input into the fitting model to generate the weight compensation coefficient of the logistics package. The compensation coefficient usually ranges from 0.8 to 1.2, which is used to correct the weighing error caused by jolt. The generated weight compensation coefficient is verified to check whether it is within a reasonable range. If an abnormal value is found, the jolt intensity calculation formula or the fitting model is adjusted to optimize the accuracy of the compensation coefficient. The generated weight compensation coefficient is output to the dynamic compensation module for subsequent weighing signal correction. At the same time, the comprehensive jolt intensity and the compensation coefficient are stored in the database of the system for subsequent analysis and optimization.

[0043] The preset bump strength calculation formula is:

[0044] Where, is the comprehensive bumpy strength of the logistics package, is the frequency of the package periodic displacement, is the maximum vibration frequency, is the weight coefficient of the vibration frequency characteristic in the comprehensive bump intensity S, is the rotation angle of the wrapping bounding box, is the weight coefficient of the tilt angle feature in the comprehensive bump intensity S, Maximum tilt angle.

[0045] It should be noted that S is the comprehensive bumpiness intensity of the logistics package. It is a dimensionless quantitative indicator used to describe the overall bumpiness of the package during transportation. The value range is usually between 0 and 1. The larger the value, the higher the bumpiness intensity. The frequency of the periodic displacement of the package refers to the frequency of the periodic up and down or left and right vibrations caused by the bumps in the package during transmission. It can be extracted by analyzing the motion trajectory of the package on the conveyor belt using Fourier transform or other frequency analysis techniques. The maximum vibration frequency is a reference value preset by the system, indicating the maximum vibration frequency that the package may reach during transmission. It can be determined based on historical data or experimental data, and is usually the maximum vibration frequency observed during system operation. θ is the weight coefficient of the vibration frequency feature in the comprehensive bump intensity S. It is used to adjust the contribution of vibration frequency to the comprehensive bump intensity. It can be determined through experiments or optimization algorithms (such as grid search or genetic algorithms) to ensure a balanced contribution of vibration frequency and rotation angle to the comprehensive bump intensity. θ is the rotation angle of the package bounding box, which refers to the rotation angle caused by the bumps during transportation. The rotation angle of the package bounding box can be extracted using image processing techniques (such as edge detection and contour analysis). The maximum tilt angle is a reference value preset by the system, indicating the maximum rotation angle that the package may reach during transportation. It can be determined based on historical data or experimental data, and is usually the maximum rotation angle observed during system operation. is the weight coefficient of the tilt angle feature in the comprehensive bump intensity S, which is used to adjust the contribution of the rotation angle to the comprehensive bump intensity. It can be determined through experiments or optimization algorithms to ensure that the contributions of the vibration frequency and rotation angle to the comprehensive bump intensity are balanced.

[0046] Reference Figure 4 , Figure 4A flowchart of a third embodiment of the full-automatic logistics weighing and visual sorting method of the present application is shown in FIG. 3. In the third embodiment, the transmission image data is input into a preset convolutional neural network to generate the transmission jolt features of the logistics package, which includes: S201: The transmission image data is subjected to grayscale conversion and normalization processing, and is input into a preset convolutional neural network; and feature extraction is performed through convolution and pooling operations; S202: The extracted features are classified through a fully connected layer, and a plurality of transmission jolt feature labels and corresponding confidence scores are output; S203: The transmission jolt features of the logistics package are determined according to the transmission jolt feature labels and corresponding confidence scores.

[0047] It should be noted that grayscale conversion refers to the process of converting a color image into a grayscale image. A grayscale image only contains brightness information and does not contain color information, which can simplify the computational load of image processing. Normalization processing refers to scaling the pixel values of the image data to a uniform range (such as between 0 and 1) to facilitate the processing and convergence of the neural network. Convolution operation refers to a kind of operation in convolutional neural network (CNN), which extracts local features of the image by sliding the convolution kernel on the image. Convolution operation can capture edge, texture and other information in the image. Pooling operation also refers to a kind of operation in convolutional neural network, which reduces the size of the feature map through downsampling while preserving the main features. Common pooling operations include max pooling and average pooling. The fully connected layer is a layer structure in the convolutional neural network, which connects all neurons of the previous layer with all neurons of the current layer, and is used for classification or regression of the extracted features. The transmission jolt feature label is used to describe the specific category of the transmission jolt features of the logistics package, such as "slight jolt", "moderate jolt", "severe jolt", etc. The confidence score represents the prediction confidence of the neural network for a certain transmission jolt feature label, which is usually a value between 0 and 1, and the larger the value, the more reliable the prediction.

[0048] In a specific implementation, the collected transmission image data is converted from a color image to a grayscale image to reduce computational complexity. The pixel values of the grayscale image are scaled to between 0 and 1 to facilitate processing by the neural network. The preprocessed image data is input to a preset convolutional neural network (CNN). By sliding a convolution kernel over the image, local features of the image (such as edges, textures, etc.) are extracted. The feature map obtained by convolution is down-sampled to reduce the size of the feature map while retaining the main features. The features extracted by convolution and pooling operations are input to a fully connected layer for feature classification. The fully connected layer outputs several transmission jolt feature labels, such as "mild jolt", "moderate jolt", and "severe jolt". At the same time, a confidence score corresponding to each transmission jolt feature label is output, indicating the prediction confidence of the neural network for each label. According to the confidence score, the transmission jolt feature label with the highest confidence is selected as the transmission jolt feature of the logistics package. If the confidence scores of multiple labels are close, a comprehensive judgment can be made in combination with multiple labels, or the final transmission jolt feature is determined by weighted averaging. The determined transmission jolt feature is output to the subsequent modules of the system for calculation of the weight compensation coefficient.

[0049] wherein the formula of the preset convolutional neural network is:

[0050] In the formula, is the transmission image data after grayscale conversion and normalization processing, is the transmission jolt feature of the logistics package output by the preset neural network, including a transmission jolt feature label and a confidence score corresponding thereto; and represent the weights and biases of the convolution layer, respectively; represents a convolution operation; represents an activation function; represents a pooling operation; represents a flattening operation, which flattens a multi-dimensional feature map into a one-dimensional vector; and represent the weights and biases of the fully connected layer, respectively.

[0051] It should be noted that, is the transmission image data after grayscale conversion and normalization processing, which is the input of the convolutional neural network. is the transmission jolt feature of the logistics package output by the preset convolutional neural network, including a transmission jolt feature label and a confidence score corresponding thereto. is the weight of the convolution layer, which is used to extract features of the image in the convolution operation and can be optimized through a neural network training process (such as a backpropagation algorithm). Bias for the convolutional layer, used to adjust the results of the convolution operation. It can be optimized through the training process of the neural network. Convolution operation, which extracts local features of the image by sliding the convolution kernel over the image. Activation function, used to introduce nonlinearity and enhance the expressive power of the neural network. Common activation functions include ReLU, Sigmoid, and Tanh. Pooling operation, which reduces the size of the feature map through downsampling while preserving the main features. Common pooling operations include max pooling and average pooling. Flattening operation, which flattens the multi-dimensional feature map into a one-dimensional vector to input into the fully connected layer. Weights for the fully connected layer, used to classify or regress the flattened features. They can be optimized through the training process of the neural network. Bias for the fully connected layer, used to adjust the results of the fully connected layer. It can be optimized through the training process of the neural network.

[0052] Further, in the present embodiment, before determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, further comprising: Obtaining the segment slope of the weighing signal fold line of the logistics package in the weighing area; Comparing the segment slope of the weighing signal fold line with the preset reference threshold to obtain the time period of logistics weighing compensation; Correspondingly, the actual weight of the logistics package is determined based on the weight compensation coefficient of the logistics package and the current weighing signal, comprising: Determining the actual weight of the logistics package based on the time period of logistics weighing compensation, the weight compensation coefficient of the logistics package, and the current weighing signal.

[0053] It should be noted that the weighing sensor is used to collect the weighing signal of the logistics package in the weighing area in real time to generate the weighing signal fold line. The weighing signal fold line is divided into multiple small time periods, and the slope of each time period is calculated. According to historical data or experimental data, a reference threshold is set to determine whether the logistics package enters or leaves the weighing area.

[0054] The preset reference threshold includes a first reference threshold, a second reference threshold, and a third reference threshold, and the comparison of the segment slope of the weighing signal fold line with the preset reference threshold to obtain the time period of logistics weighing compensation comprises: Obtaining the slope before and after any time in the weighing signal fold line of the current logistics package, and recording them as the first slope and the second slope, respectively; If the difference between the first slope and the second slope is greater than the first reference threshold value, and the second slope is less than the second reference threshold value, the time is the time period starting time; If the difference between the first slope and the second slope is greater than the third reference threshold value, and the first slope is less than the second reference threshold value, the time is the time period ending time; According to the obtained starting time and ending time, the time period of the logistics weighing compensation is determined.

[0055] It should be understood that, when the logistics package just enters the weighing area, the slope of the weighing signal fold line can be regarded as the first slope because the logistics package has not completely entered the weighing area, and when the logistics package completely enters the weighing area, the slope of the weighing signal fold line basically tends to zero, which can be regarded as the second slope. At the moment when the logistics package completely enters the weighing area, the difference between the first slope and the second slope is large, and at this moment, a reference threshold value can be set to determine the moment when the logistics package completely enters the weighing area, that is, the first reference threshold value. At the same time, after the logistics package completely enters the weighing area, the second slope is small, and the second reference threshold value is set to determine that the time is the time period starting time.

[0056] Similarly, at the moment when the logistics package just starts to leave the weighing area, the difference between the first slope and the second slope is large, and at this moment, a reference threshold value can be set to determine the moment when the logistics package just starts to leave the weighing area, that is, the third reference threshold value. At the same time, before the logistics package just starts to leave the weighing area, the first slope is small, and at this moment, the first slope is less than the second reference threshold value to determine that the time is the time period ending time. The time period of the logistics weighing compensation is determined according to the obtained starting time and ending time, so that the residence time of the logistics package in the weighing area is accurately identified, and the weight compensation of the logistics package is facilitated.

[0057] In addition, an embodiment of the present application also provides a storage medium, and the storage medium stores a full-automatic logistics weighing and visual sorting program. The full-automatic logistics weighing and visual sorting program is executed by a processor to realize the steps of the full-automatic logistics weighing and visual sorting method.

[0058] In addition, with reference to Figure 5 , the embodiment of the present application also provides a full-automatic logistics weighing and visual sorting system, which comprises: An image acquisition module 10 is configured to acquire transmission image data of the logistics package in the weighing area and a current weighing signal, and the transmission image data comprises logistics package position information and package feature information. A feature extraction module 20 is configured to input the transmission image data into a preset convolutional neural network to generate transmission bump features of the logistics package. The compensation calculation module 30 is configured to determine a weight compensation coefficient of the logistics package according to the transmission bumping feature of the logistics package. The weight sorting module 40 is configured to determine an actual weight of the logistics package according to the weight compensation coefficient of the logistics package and the current weighing signal, and perform package sorting according to the actual weight of the logistics package.

[0059] Other embodiments or specific implementations of the full-automatic logistics weighing and visual sorting system according to the present application can refer to the above-mentioned method embodiments, which will not be described here again.

[0060] It should be noted that in this document, the terms "comprise", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or system. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of another identical element in the process, method, article or system that includes the element.

[0061] The above-mentioned embodiment numbers of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. In the system unit claims listed in several systems, several of these systems can be embodied by the same hardware item. The use of the words first, second, and third does not represent any order, and these words can be interpreted as names.

[0062] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by software and the necessary general hardware platform, of course, they can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory image (ROM) / random access memory (RAM), a magnetic disk, an optical disk), and includes a plurality of instructions for making a terminal user device (which can be a mobile phone, a computer, a server, an air conditioner, or a network user device) execute the methods described in the embodiments of the present application.

[0063] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation made by using the content of the present application specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A fully automatic logistics weighing and visual sorting method, characterized in that: The method comprises: Acquire the transmission image data of the logistics package in the weighing area and the current weighing signal, wherein the transmission image data includes the logistics package location information and package feature information; Inputting the transmission image data into a preset convolutional neural network to generate transmission bumpy features of the logistics package; Determining a weight compensation coefficient of the logistics package based on the transmission bumpy characteristics of the logistics package; The actual weight of the logistics package is determined based on the weight compensation coefficient of the logistics package and the current weighing signal, and the packages are sorted according to the actual weight of the logistics package.

2. The fully automatic logistics weighing and visual sorting method according to claim 1, characterized in that: The determining of the weight compensation coefficient of the logistics package according to the transmission bumpy characteristics of the logistics package includes: Extracting the transmission bump characteristics of the logistics package to obtain the frequency of the package's periodic displacement and the rotation angle of the package's bounding box; Based on the frequency of the periodic displacement of the package and the rotation angle of the package boundary box, the comprehensive bump intensity of the logistics package is determined by a preset bump intensity calculation formula; Based on the comprehensive bumpiness strength of the logistics package, the weight compensation coefficient of the logistics package is obtained by fitting historical data.

3. The fully automatic logistics weighing and visual sorting method according to claim 2, characterized in that: The preset bump strength calculation formula is: Where, is the comprehensive bumpy strength of the logistics package, is the frequency of the package periodic displacement, is the maximum vibration frequency, is the weight coefficient of the vibration frequency characteristic in the comprehensive bump intensity S, is the rotation angle of the wrapping bounding box, is the weight coefficient of the tilt angle feature in the comprehensive bump intensity S, is the maximum tilt angle.

4. The fully automatic logistics weighing and visual sorting method according to claim 1, characterized in that: The step of inputting the transmission image data into a preset convolutional neural network to generate transmission bumpy features of the logistics package includes: grayscale conversion and normalization processing are performed on the transmitted image data, and the data is input into a preset convolutional neural network; Feature extraction through convolution and pooling operations; The extracted features are classified through the fully connected layer, and several transmission bump feature labels and their corresponding confidence scores are output; The transmission bump feature of the logistics package is determined according to the transmission bump feature label and its corresponding confidence score.

5. The fully automatic logistics weighing and visual sorting method according to claim 4, characterized in that: The formula of the preset convolutional neural network is: Where, is the transmitted image data after grayscale conversion and normalization. The transmission bump characteristics of the logistics package output by the preset neural network, including the transmission bump feature label and its corresponding confidence score; and Represent the weights and biases of the convolutional layer respectively; Represents the convolution operation; represents the activation function; Represents a pooling operation; Represents the flattening operation, which flattens the multi-dimensional feature map into a one-dimensional vector; and denote the weights and biases of the fully connected layer respectively.

6. The fully automatic logistics weighing and visual sorting method according to claim 1, characterized in that: Before determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, the method further includes: Obtain the segmented slope of the weighing signal line of the logistics package in the weighing area; Comparing the segmented slope of the weighing signal broken line with a preset reference threshold to obtain a time period for logistics weighing compensation; Accordingly, the determining the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal includes: The actual weight of the logistics package is determined based on the time period of the logistics weighing compensation, the weight compensation coefficient of the logistics package and the current weighing signal.

7. The fully automatic logistics weighing and visual sorting method according to claim 6, characterized in that: The preset reference thresholds include a first reference threshold, a second reference threshold, and a third reference threshold. Comparing the segmented slopes of the weighing signal broken line with the preset reference thresholds to obtain the time period for logistics weighing compensation includes: Obtain the slope of the weighing signal line of the current logistics package before and after any moment, and record them as the first slope and the second slope respectively; If the difference between the first slope and the second slope is greater than the first reference threshold, and the second slope is less than the second reference threshold, then this moment is the start moment of the time period; If the difference between the first slope and the second slope is greater than the third reference threshold, and the first slope is less than the second reference threshold, then this moment is the end moment of the time period; The time period for logistics weighing compensation is determined based on the obtained start time and end time.

8. A fully automatic logistics weighing and visual sorting system, characterized in that: The fully automatic logistics weighing and visual sorting system includes: An image acquisition module is used to obtain the transmission image data of the logistics package in the weighing area and the current weighing signal. The transmission image data includes the logistics package location information and package feature information; A feature extraction module is used to input the transmission image data into a preset convolutional neural network to generate transmission bumpy features of the logistics package; A compensation calculation module, configured to determine a weight compensation coefficient of the logistics package based on a transmission bumpy characteristic of the logistics package; The weight sorting module is used to determine the actual weight of the logistics package based on the weight compensation coefficient of the logistics package and the current weighing signal, and sort the packages according to the actual weight of the logistics package.

9. A fully automatic logistics weighing and visual sorting equipment, characterized in that: The device includes: a memory, a processor, and a fully automatic logistics weighing and visual sorting program stored in the memory and runnable on the processor, wherein the fully automatic logistics weighing and visual sorting program is configured to implement the steps of the fully automatic logistics weighing and visual sorting method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a fully automatic logistics weighing and visual sorting program, which, when executed by the processor, implements the steps of the fully automatic logistics weighing and visual sorting method as described in any one of claims 1 to 7.

Images