A method and system for identifying and positioning express parcels based on machine vision

Abstract

The present invention relates to the technical field of logistics transportation, and in particular to a method and system for identifying and positioning express packages based on machine vision, including obtaining three-dimensional information of the express package location, where the three-dimensional information of the express package location includes real-time three-dimensional information and historical three-dimensional information; establishing a relationship between the three-dimensional information of the express package location and the running path of the express package based on the recognition and grasping model to obtain an express package location query model with a position index structure; identifying the real-time three-dimensional information of the express package, and performing position prediction based on the express package location query model to obtain real-time recognition and grasping attitude instructions for different packages; using the recognition and grasping model to achieve position capture and recognition of express packages in the logistics warehouse scenario, and feeding back the express package positioning location. The purpose of the present invention is to solve the problem that the existing method for positioning express packages cannot obtain the accurate object attitude of the express packages, so that the package grasping device cannot determine the suction pose.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics transportation, and particularly to a method and system for identifying and positioning express parcels based on machine vision. Background Art

[0002] In recent years, the online shopping boom has been growing stronger, the transaction scale of the e-commerce market has been increasing, and consumers have higher and higher requirements for logistics speed and quality. At present, the classification of domestic express parcels mainly relies on manual sorting. However, in the face of the strong demand in the domestic market, the manual sorting method not only has low sorting efficiency, high sorting cost, but also is prone to errors, causing double losses to e-commerce and consumers, and cannot meet the needs of the increasingly developing express delivery industry.

[0003] Chinese Patent CN112308915A discloses a method and device for positioning express parcels, including: receiving an express positioning request, where the express positioning request is used to locate the express bill corresponding to the target express parcel from an express image presenting an express parcel, and the express positioning request includes express information; searching for the position information of the express bill corresponding to the express information stored in advance; using the position information to generate an express positioning result image, where the express positioning result image includes a positioning identifier for indicating the express bill corresponding to the target express parcel from the express image; outputting the express positioning result image; however, the existing method for positioning express parcels cannot obtain the accurate object pose of the express parcel, making it impossible for the parcel grabbing device to determine the suction pose. Based on this, we propose a method and system for identifying and positioning express parcels based on machine vision. Summary of the Invention

[0004] Aiming at the deficiencies of the existing algorithms, the present invention aims to solve the problem that the existing method for positioning express parcels cannot obtain the accurate object pose of the express parcel, making it impossible for the parcel grabbing device to determine the suction pose.

[0005] The existing methods for locating express parcels cannot obtain the precise object pose of the express parcels, making it impossible for the parcel grabbing device to determine the suction pose. Based on this, we propose a method for identifying and locating express parcels based on machine vision. The method for identifying and locating express parcels based on machine vision includes: obtaining the three-dimensional information of the express location, where the three-dimensional information of the express location includes real-time three-dimensional information and historical three-dimensional information; establishing the relationship between the three-dimensional information of the express location and the express running path based on the recognition and grabbing model to obtain an express location query model with a position index structure; identifying the real-time three-dimensional information of the express, and making a position prediction based on the express location query model to obtain real-time recognition and grabbing pose instructions for different parcels; using the recognition and grabbing model to achieve the position capture and recognition of the express in the logistics warehouse scenario, and feedback the express positioning location. Through the recognition and grabbing model, this application realizes the position capture and recognition of the express in the logistics warehouse scenario and feedbacks the express positioning location, thus ensuring the recognition accuracy of the express parcels and facilitating the positioning and real-time grabbing of the express parcels.

[0006] The technical solution adopted by the present invention is: a method and system for identifying and locating express parcels based on machine vision, including the following steps:

[0007] Obtaining the three-dimensional information of the express location, where the three-dimensional information of the express location includes real-time three-dimensional information and historical three-dimensional information;

[0008] Establishing the relationship between the three-dimensional information of the express location and the express running path based on the recognition and grabbing model to obtain an express location query model with a position index structure;

[0009] Identifying the real-time three-dimensional information of the express, and making a position prediction based on the express location query model to obtain real-time recognition and grabbing pose instructions for different parcels;

[0010] Using the recognition and grabbing model to achieve the position capture and recognition of the express in the logistics warehouse scenario, and feedback the express positioning location.

[0011] Preferably, the recognition and grabbing model includes:

[0012] A visual acquisition module for identifying and generating a pitch angle to obtain the real-time state of the express parcel;

[0013] A host computer for processing the three-dimensional point cloud information of the express parcel obtained from each perspective.

[0014] Preferably, the visual acquisition module is a Realsense L515 3D camera, and the visual acquisition module is communicatively connected to a host computer.

[0015] Preferably, the training method of the recognition and grabbing model specifically includes:

[0016] Obtaining standard three-dimensional information;

[0017] Segment the given standard three-dimensional information using the jieba segmentation tool, and filter out invalid three-dimensional information to obtain the segmentation result of the standard three-dimensional information;

[0018] Load the GloVe embedding model and convert all the retained standard three-dimensional information into standard three-dimensional vectors;

[0019] Perform initial model training on the standard three-dimensional vector set through the LDA model in the Gensim module.

[0020] Preferably, the training method for identifying the grasping model further includes:

[0021] Using the trained initial grasping model, select the recognition path with the highest probability corresponding to the standard three-dimensional information, and then select the top n paths and their corresponding probabilities under the path with the highest probability;

[0022] Normalize the probability values as the weight information of the n paths.

[0023] Preferably, the method for predicting the position based on the express position query model to obtain the real-time recognition and grasping attitude instructions for different packages specifically includes:

[0024] Concatenate the global three-dimensional vector obtained from the express position query model with the key-point feature vector of the LDA model. After concatenation, the global three-dimensional vector integrates the express attitude features and key-node features of the entire path;

[0025] Establish a functional relationship for the grasping attitude instruction through the LDA model algorithm and historical three-dimensional information parameters to achieve self-calibration of the grasping attitude;

[0026] Perform grasping attitude code transcoding operation based on a preset transcoding protocol to transcode the grasping attitude code into the recognition and grasping attitude instruction.

[0027] Preferably, the method for using the recognition and grasping model to achieve position capture and recognition of express delivery in a logistics warehouse scenario specifically includes:

[0028] Respond to the recognition and grasping attitude instruction, send the recognition and grasping attitude instruction to the visual acquisition module through the recognition and grasping model, and store the corresponding value of the visual acquisition module in the visual acquisition module;

[0029] Locate the optimal comprehensive evaluation index for attitude grasping from the network KPI dataset corresponding to the historical three-dimensional information;

[0030] Generate grasping parameters according to the optimal comprehensive evaluation index for attitude grasping.

[0031] Preferably, the method for capturing and identifying the position of express delivery in the logistics warehouse scenario using the recognition and grasping model specifically further includes:

[0032] Call the host computer corresponding to the grasping parameters generated by the comprehensive evaluation index and the Realsense L515 3D camera, and send them to the multi-channel communication device corresponding to the Realsense L515 3D camera to perform the position capture and identification of the express delivery.

[0033] A machine vision-based express delivery package recognition and positioning system for an express delivery package recognition and positioning method based on machine vision specifically includes:

[0034] A three-dimensional information acquisition module for acquiring the three-dimensional information of the express delivery position, where the three-dimensional information of the express delivery position includes real-time three-dimensional information and historical three-dimensional information;

[0035] A query model acquisition module establishes the relationship between the three-dimensional information of the express delivery position and the running path of the express delivery based on the recognition and grasping model, and obtains an express delivery position query model with a position index structure;

[0036] A position prediction module is used to identify the real-time three-dimensional information of the express delivery, and perform position prediction based on the express delivery position query model to obtain real-time recognition and grasping attitude instructions for different packages.

[0037] Preferably, the machine vision-based express delivery package recognition and positioning system further includes:

[0038] A capture and recognition module uses the recognition and grasping model to achieve the position capture and recognition of the express delivery in the logistics warehouse scenario, and feeds back the express delivery positioning position.

[0039] Advantages of the present invention:

[0040] 1. By using the recognition and grasping model to achieve the position capture and recognition of the express delivery in the logistics warehouse scenario, and feeding back the express delivery positioning position, the recognition accuracy of the express delivery package is guaranteed, which facilitates the positioning and real-time grasping of the express delivery package. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a schematic structural diagram of the machine vision-based express delivery package recognition and positioning method of the present invention;

[0042] Figure 2 is a schematic implementation flowchart of the training method of the recognition and grasping model of the present invention;

[0043] Figure 3 is a schematic implementation flowchart of obtaining real-time recognition and grasping attitude instructions for different packages by performing position prediction based on the express delivery position query model of the present invention;

[0044] Figure 4 It is a schematic diagram of the implementation process of using an identification and grasping model to capture and identify the position of express deliveries in a logistics warehouse scenario according to the present invention;

[0045] Figure 5 It is a schematic structural diagram of an express delivery package identification and positioning system based on machine vision according to the present invention,

[0046] In the figure, 100 - three - dimensional information acquisition module, 200 - query model acquisition module, 300 - position prediction module, 400 - capture and identification module. Specific implementation manners

[0047] The following further describes the present invention in conjunction with the accompanying drawings and embodiments. This figure is a simplified schematic diagram, which only illustrates the basic structure of the present invention in a schematic manner. Therefore, it only shows the components related to the present invention.

[0048] Existing methods for positioning express delivery packages cannot obtain the precise object pose of the express delivery packages, making it impossible for the package grasping device to determine the suction pose. Based on this, we propose a method for identifying and positioning express delivery packages based on machine vision. The method for identifying and positioning express delivery packages based on machine vision includes: obtaining three - dimensional information of the express delivery position, where the three - dimensional information of the express delivery position includes real - time three - dimensional information and historical three - dimensional information; establishing the relationship between the three - dimensional information of the express delivery position and the running path of the express delivery based on the identification and grasping model to obtain an express delivery position query model with a position index structure; identifying the real - time three - dimensional information of the express delivery, and performing position prediction based on the express delivery position query model to obtain real - time identification and grasping pose instructions for different packages; using the identification and grasping model to achieve the position capture and identification of express deliveries in the logistics warehouse scenario, and feeding back the express delivery positioning position. Through the identification and grasping model, the present application realizes the position capture and identification of express deliveries in the logistics warehouse scenario and feeds back the express delivery positioning position, thereby ensuring the recognition accuracy of express delivery packages and facilitating the positioning and real - time grasping of express delivery packages.

[0049] As Figure 1 shown, a method and system for identifying and positioning express delivery packages based on machine vision includes the following steps:

[0050] Step S10: Obtain three - dimensional information of the express delivery position, where the three - dimensional information of the express delivery position includes real - time three - dimensional information and historical three - dimensional information.

[0051] Step S20: Establish the relationship between the three - dimensional information of the express delivery position and the running path of the express delivery based on the identification and grasping model to obtain an express delivery position query model with a position index structure.

[0052] Step S30: Identify the real - time three - dimensional information of the express delivery, and perform position prediction based on the express delivery position query model to obtain real - time identification and grasping pose instructions for different packages.

[0053] Step S40: Use the recognition and grasping model to capture and recognize the position of the express delivery in the logistics warehouse scenario, and feedback the positioning position of the express delivery.

[0054] This application realizes the capture and recognition of the position of the express delivery in the logistics warehouse scenario through the recognition and grasping model, and feeds back the positioning position of the express delivery, thus ensuring the recognition accuracy of the express package and facilitating the positioning and real-time grasping of the express package.

[0055] In this embodiment, the recognition and grasping model includes:

[0056] A visual acquisition module, used to recognize and generate a pitch angle, and obtain the real-time status of the express package;

[0057] A host computer, used to process the three-dimensional point cloud information of the express package obtained from various perspectives.

[0058] Exemplarily, the visual acquisition module is a Realsense L515 3D camera, and the visual acquisition module is communicatively connected to a host computer. The host computer is an industrial computer installed with Windows 10 and uses an Intel CORE i7 processor. The Realsense L515 3D camera uses Intel Realsense L515. Intel Realsense L515 is the first lidar camera released by Intel, which can achieve highly accurate depth sensing in a small size. Three different specifications of images can be selected. Here, the point cloud map is used as the input image, and the integrated gimbal (including the host computer and the rotating device) rotates the pitch angle of the two-dimensional laser scanner to obtain the three-dimensional point cloud information of the express package from multiple angles (from various perspectives).

[0059] In this embodiment, the recognition and grasping system further includes a robot component, which mainly consists of three parts: a mobile base, a robotic arm, and the end of the robotic arm. The robotic arm is installed on the mobile base. There is a power supply, a robotic arm control box, an industrial computer, and a suction pump on the mobile base. At the same time, because the method of sucking objects has a higher success rate and a simple and reliable structure, we adopt the method of sucking objects, use the suction pump to provide the adsorption force, and use the solenoid valve to control the on-off of the gas path.

[0060] The 3D camera is also installed at the end of the robotic arm. The 3D camera is a Realsense L515 3D camera. It uses proprietary MEMS mirror scanning technology, which can achieve higher laser power efficiency compared with other time-of-flight technologies. The depth stream power consumption of the Intel RealSense LiDAR camera L515 is less than 3.5W, making it the most energy-efficient high-resolution LiDAR camera.

[0061] The embodiment of the present invention provides a method for training the recognition and grasping model, as Figure 2As shown, the method for training an identification and grasping model specifically includes:

[0062] Step S101, obtain standard three-dimensional information;

[0063] Step S102, segment the given standard three-dimensional information using the jieba segmentation tool and filter out invalid three-dimensional information to obtain the segmentation result of the standard three-dimensional information;

[0064] Step S103, load the GloVe embedding model and convert all the remaining standard three-dimensional information into standard three-dimensional vectors;

[0065] Step S104, perform initial grasping model training on the standard three-dimensional vector set through the LDA model in the Gensim module;

[0066] Step S105, use the trained initial grasping model to select the recognition path with the highest probability corresponding to the standard three-dimensional information, and then select the top n paths and their corresponding probabilities under the path with the highest probability;

[0067] Step S106, normalize the probability values as the weight information of the n paths.

[0068] Exemplarily, the neural network structure model of the GloVe embedding model is divided into 2 independent GloVes. The input sequences are input into the 2 GloVe neural networks in forward and reverse order respectively for feature extraction, and the 2 output vectors (i.e., the extracted feature vectors) are used for initial grasping model training of the standard three-dimensional vector set. The model design concept of the GloVe embedding model is to enable the feature data obtained at time t to have information between the past and the future at the same time. Experiments have proved that this neural network structure model has better extraction efficiency and performance for the standard three-dimensional vector set than a single GloVe structure model. It is worth mentioning that the parameters of the 2 GloVe neural networks in the GloVe are independent of each other.

[0069] The embodiment of the present invention provides a method for predicting positions based on an express delivery position query model to obtain real-time identification and grasping posture instructions for different packages, as Figure 3 As shown, the method for predicting positions based on an express delivery position query model to obtain real-time identification and grasping posture instructions for different packages specifically includes:

[0070] Step S201, splice the global three-dimensional vector obtained based on the express delivery position query model with the key point feature vector of the LDA model. After splicing, the global three-dimensional vector integrates the express delivery posture features and key node features of the entire path;

[0071] Step S202, establish a functional relationship of the grasping posture instruction through the LDA model algorithm and historical three-dimensional information parameters to achieve self-calibration of the grasping posture.

[0072] Step S203: Perform transcoding operation on the grasping posture code based on a preset transcoding protocol to transcode the grasping posture code into an instruction for recognizing the grasping posture.

[0073] The embodiment of the present invention provides a method for realizing the position capture and recognition of express delivery in a logistics warehouse scenario by using a recognition and grasping model. As Figure 4 shown, the method for realizing the position capture and recognition of express delivery in a logistics warehouse scenario by using a recognition and grasping model specifically includes:

[0074] Step S301: Respond to the instruction for recognizing the grasping posture, send the instruction for recognizing the grasping posture to the visual acquisition module through the recognition and grasping model, and store the corresponding value of the visual acquisition module in the visual acquisition module.

[0075] Step S302: Locate the optimal comprehensive evaluation index for posture grasping from the network KPI dataset corresponding to the historical three-dimensional information.

[0076] Step S303: Generate grasping parameters according to the optimal comprehensive evaluation index for posture grasping.

[0077] Step S304: Call the upper computer and the Realsense L515 3D camera corresponding to the generated grasping parameters according to the comprehensive evaluation index, and send them to the multi-channel communication device corresponding to the Realsense L515 3D camera to perform the position capture and recognition of the express delivery.

[0078] The embodiment of the present invention provides an express package recognition and positioning system based on machine vision. As Figure 5 shown, the express package recognition and positioning system based on machine vision specifically includes:

[0079] A three-dimensional information acquisition module 100, which is used to acquire the three-dimensional information of the express delivery position, where the three-dimensional information of the express delivery position includes real-time three-dimensional information and historical three-dimensional information.

[0080] A query model acquisition module 200, which establishes the relationship between the three-dimensional information of the express delivery position and the express delivery operation path based on the recognition and grasping model to obtain an express delivery position query model with a position index structure.

[0081] A position prediction module 300, which is used to recognize the real-time three-dimensional information of the express delivery and perform position prediction based on the express delivery position query model to obtain real-time recognition and grasping posture instructions for different packages.

[0082] A capture and recognition module 400, which uses the recognition and grasping model to realize the position capture and recognition of the express delivery in the logistics warehouse scenario and feeds back the express delivery positioning position.

[0083] The present invention also provides a readable storage medium with a computer program stored thereon. When the computer program is executed by a processor, the processor performs the steps of the method for identifying and positioning express parcels based on machine vision.

[0084] Specifically, the execution steps are as follows:

[0085] Step S10: Obtain the three-dimensional information of the express location, where the three-dimensional information of the express location includes real-time three-dimensional information and historical three-dimensional information.

[0086] Step S20: Based on the recognition and grasping model, establish the relationship between the three-dimensional information of the express location and the express operation path, and obtain an express location query model with a position index structure.

[0087] Step S30: Identify the real-time three-dimensional information of the express, and perform position prediction based on the express location query model to obtain the real-time recognition and grasping attitude instructions for different parcels.

[0088] Step S40: Use the recognition and grasping model to achieve the position capture and recognition of the express in the logistics warehouse scenario, and feedback the express positioning location.

[0089] This application realizes the position capture and recognition of the express in the logistics warehouse scenario through the recognition and grasping model, and feedbacks the express positioning location, thereby ensuring the recognition accuracy of the express parcel and facilitating the positioning and real-time grasping of the express parcel.

[0090] Exemplarily, the vision acquisition module is a Realsense L515 3D camera, and the vision acquisition module is communicatively connected to a host computer. The host computer is an industrial computer installed with Windows10 and uses an Intel CORE i7 processor. The RealsenseL515 3D camera uses an Intel Realsense L515. The Intel Realsense L515 is the first lidar camera released by Intel, which can achieve highly accurate depth sensing in a small size. Three different specifications of images can be selected. Here, the point cloud map is used as the input image, and the integrated cloud platform (including the host computer and the rotating device) rotates the pitch angle of the two-dimensional laser scanner to obtain the three-dimensional point cloud information of the express parcel from multiple angles (under each perspective).

[0091] In this embodiment, the recognition and grasping system further includes a robot component, which mainly consists of three parts: a mobile base, a robotic arm, and the end of the robotic arm. The robotic arm is installed on the mobile base. There is a power supply, a robotic arm control box, an industrial computer, and a suction pump on the mobile base. At the same time, because the method of sucking objects has a higher success rate and a simple and reliable structure, we adopt the method of sucking objects, use the suction pump to provide the adsorption force, and use the solenoid valve to control the on / off of the gas path.

[0092] Exemplarily, a computer program may be divided into one or more modules. One or more modules are stored in a memory and executed by a processor to implement the present invention. One or more modules may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device. For example, the above computer program may be divided into execution units or modules of the machine vision-based express package recognition and positioning method provided by the above various system embodiments.

[0093] Those skilled in the art can understand that the description of the above terminal device is only an example and does not constitute a limitation on the terminal device. It may include more or fewer components than the above description, or combine certain components, or different components. For example, it may include input / output devices, network access devices, buses, etc.

[0094] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The above processor is the control center of the above terminal device, and connects various parts of the entire user terminal through various interfaces and lines.

[0095] The above-mentioned memory can be used to store computer programs and / or modules. By running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory, the above-mentioned terminal device can implement various functions. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as information collection template display function, product information release function, etc.); the data storage area can store the data created by using the machine vision-based express package recognition and positioning method (such as product information collection templates corresponding to different product types, product information to be released by different product providers, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disks, memory, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices.

[0096] If the modules / units integrated in the terminal device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the modules / units in the above-mentioned embodiment system of the present invention, it can also be completed by instructing relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the functions of the above-mentioned various system embodiments. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device that can carry computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0097] In summary, the present invention provides a machine vision-based express package recognition and positioning method. The machine vision-based express package recognition and positioning method provided by the present invention preprocesses the acquired point cloud data through a 3D camera, such as filtering and segmentation. For the complex scene of the acquired express delivery stack, a method of cascading a neural network and a traditional point pair feature matching algorithm is applied to complete the similarity matching and recognition and positioning of different features.

[0098] Inspired by the above-described ideal embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for identifying and positioning express parcels based on machine vision, characterized in that, Including: Obtain the three-dimensional information of the express location, where the three-dimensional information of the express location includes real-time three-dimensional information and historical three-dimensional information; Based on the recognition and grasping model, establish the relationship between the three-dimensional information of the express location and the express operation path, and obtain an express location query model with a position index structure; Identify the real-time three-dimensional information of the express, and based on the express location query model, perform position prediction to obtain real-time recognition and grasping attitude instructions for different packages; Specifically including: The global three-dimensional vector obtained based on the express location query model is spliced with the key point feature vector of the LDA model. After splicing, the global three-dimensional vector integrates the express attitude features and key node features of the entire path; Establish a functional relationship of the grasping attitude instruction through the LDA model algorithm and the historical three-dimensional information parameters to realize the self-calibration of the grasping attitude; Based on the preset transcoding protocol, perform grasping attitude code transcoding operation to transcode the grasping attitude code into an identification and grasping attitude instruction; Use the recognition and grasping model to realize the position capture and recognition of the express in the logistics warehouse scenario, specifically including: Respond to the recognition and grasping attitude instruction, send the recognition and grasping attitude instruction to the visual acquisition module through the recognition and grasping model, and store the corresponding value of the visual acquisition module in the visual acquisition module; Locate the optimal attitude grasping comprehensive evaluation index from the network KPI dataset corresponding to the historical three-dimensional information; Generate grasping parameters according to the optimal attitude grasping comprehensive evaluation index; Call the host computer corresponding to the grasping parameters generated by the comprehensive evaluation index and the Realsense L515 3D camera, and send them to the multi-channel communication device corresponding to the Realsense L515 3D camera to perform the position capture and recognition of the express; Use the recognition and grasping model to realize the position capture and recognition of the express in the logistics warehouse scenario, and feedback the express positioning position.

2. The method for identifying and positioning an express parcel based on machine vision according to claim 1, wherein The recognition and grasping model includes: A visual acquisition module for recognizing and generating a pitch angle to obtain the real-time state of the express package; A host computer for processing the three-dimensional point cloud information of the express package obtained from each perspective.

3. The method for identifying and positioning an express parcel based on machine vision according to claim 2, wherein, The visual acquisition module is a Realsense L515 3D camera, and the visual acquisition module is communicatively connected to a host computer.

4. The method for identifying and positioning express parcels based on machine vision according to claim 3, wherein, The training method of the recognition and grasping model specifically includes: Obtain standard three-dimensional information; Based on the jieba segmentation tool, segment the given standard three-dimensional information and filter out invalid three-dimensional information to obtain the segmentation result of the standard three-dimensional information; Load the GloVe embedding model and convert all the remaining standard three-dimensional information into standard three-dimensional vectors; Perform grasping initial model training on the standard three-dimensional vector set through the LDA model in the Gensim module.

5. The method for identifying and positioning an express parcel based on machine vision according to claim 4, wherein The training method of the recognition and grasping model also includes: Using the trained grasping initial model, select the recognition path with the highest probability corresponding to the standard three-dimensional information, and then select the top n paths and their corresponding probabilities under the path with the highest probability; Normalize the probability values as the weight information of the n paths.

6. The system for the machine vision-based express package identification and positioning method according to any one of claims 1-5, characterized in that, Including: A three-dimensional information acquisition module for obtaining the three-dimensional information of the express location, where the three-dimensional information of the express location includes real-time three-dimensional information and historical three-dimensional information; A query model acquisition module that establishes the relationship between the three-dimensional information of the express delivery location and the express delivery operation path based on the recognition and grasping model, and obtains an express delivery location query model with a position index structure; A position prediction module that is used to identify the real-time three-dimensional information of the express delivery, and perform position prediction based on the express delivery location query model to obtain real-time recognition and grasping posture instructions for different packages.

7. The system of a machine vision-based express package recognition and positioning method according to claim 6, characterized in that, The system further includes: A capture and recognition module that uses the recognition and grasping model to achieve position capture and recognition of express deliveries in the logistics warehouse scenario, and feedback the express delivery positioning location.

Images