Multi-dimensional logistics package identification and tracking method based on deep learning

Abstract

The invention discloses a multi-dimensional logistics package identification and tracking method based on deep learning. The method comprises the following steps: S1, data acquisition: acquiring related information of a package; s2, data preprocessing: preprocessing the collected data, and dividing the data into a training set, a verification set and a test set; s3, selecting a deep learning model: identifying the parcel through a target detection model, extracting features from the image and detecting the position of the parcel; s4, model training: training the selected deep learning model through training set data; s5, multi-target tracking: adopting a tracking algorithm to combine with the detected parcel position information to realize real-time tracking of a plurality of parcels; and S6, feature extraction and fusion. According to the method, the data which can be acquired by the damaged parcel are identified, and the damaged and lost data are repaired according to the acquired data; and the logistics information of the parcel can be clearly known after the parcel is repaired, so that the parcel can be continuously delivered.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics parcel identification and tracking, and specifically, to a multi-dimensional logistics parcel identification and tracking method based on deep learning. Background Art

[0002] Real-time tracking of express delivery information is particularly important in the modern logistics system. Real-time tracking of express delivery information allows consumers to understand the transportation status and estimated delivery time of parcels at any time, reducing the uncertainty and anxiety during the waiting process, and enhancing the user's shopping experience and satisfaction. Currently, estimating the time of arrival of parcels based on location mainly relies on combining the current location on the map with the preset transportation route, considering factors such as historical average speed and historical road conditions, and roughly predicting the time required for the parcel to reach the destination from the current location through empirical formulas or algorithm models.

[0003] A method and system for real-time tracking of express delivery information disclosed in the existing publication number CN118898435A belongs to the technical field of logistics transportation. The method includes: if it is recognized that the parcel is in the transportation state, determining whether there is a congestion situation; if not, determining the estimated time of arrival of the parcel and the current location of the parcel, and sending them to the user; if it is recognized that the parcel is in the sorting state, determining the sorting duration of the parcel, updating the estimated time of arrival of the parcel and the current location of the parcel, and sending them to the user; if it is recognized that the parcel is in the delivery state, determining the delivery duration, and updating the estimated time of arrival of the parcel according to the delivery duration, updating the current location of the parcel, and sending it to the user; predicting the time of arrival based on the corresponding real-time data in each link, ensuring that the prediction model is calculated based on the latest data, and being able to focus on the characteristics and influencing factors in a specific state, which can improve the prediction accuracy of the time of arrival, and dynamically updating the estimated time of arrival can improve the user experience.

[0004] The existing identification and tracking of logistics parcels generally rely on the labels, barcodes, and two-dimensional codes of the parcels for identification. However, when the parcels are in transit, the labels may be worn or damaged. After wear or damage, the couriers cannot continue with the delivery. To solve the above technical problems, we propose a multi-dimensional logistics parcel identification and tracking method based on deep learning. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-dimensional logistics parcel identification and tracking method based on deep learning to solve the problems in the prior art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A multi-dimensional logistics parcel identification and tracking method based on deep learning, including the following steps:

[0007] S1. Data collection: Collect relevant information of the parcel;

[0008] S2. Data preprocessing: Preprocess the collected data and divide it into a training set, a validation set, and a test set;

[0009] S3. Select a deep learning model: Identify the package through an object detection model, extract features from the image, and detect the position of the package;

[0010] S4. Model training: Train the selected deep learning model with the training set data;

[0011] S5. Multi-object tracking: Use a tracking algorithm combined with the detected package position information to achieve real-time tracking of multiple packages;

[0012] S6. Feature extraction and fusion: Extract the weight, size information, and transportation path of the package, and fuse the weight, size information, and transportation path;

[0013] S7. Model optimization and adjustment; Optimize and adjust the model based on the results of the validation set;

[0014] S8. Real-time collection and tracking: When collecting package data for the first time, the data needs to be stored; Use GPS and Beidou positioning technologies to perform real-time tracking and positioning of the package; Perform key frame sampling and detection on the real-time obtained logistics package video to obtain the position information of the package, and use the multi-object tracking algorithm for continuous tracking;

[0015] S9. Repair and tracking of damaged package information: Identify the data that can be obtained for damaged packages, and repair the damaged and lost data based on the obtained data.

[0016] Preferably, it further includes S10. Model learning and improvement: Continuously train and adjust the model as more data accumulates.

[0017] Preferably, it further includes S11. User interaction: Users can query the status of packages in real time.

[0018] Preferably, in S1, the appearance picture or video stream of the package is captured by a camera; The identity identification information of the package is obtained through an RFID reader; The sensor is used to measure the physical characteristics of the package such as size and weight.

[0019] Preferably, in S2, the data preprocessing includes cropping, resizing, and normalizing of the image.

[0020] Preferably, in S3, identifying the package includes extracting the package ID from barcodes, QR codes, and directly from the text information on the surface of the package; The deep learning model predicts the best transportation route and estimated arrival time of the package.

[0021] Preferably, in S6, by combining the movement and surface feature information of the logistics parcel, the apparent information based on the minimum cosine distance of the deep feature vector and the movement information based on the Mahalanobis distance are weighted and fused.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: When collecting parcel data, the data needs to be stored; the parcel is tracked and positioned in real time through GPS and Beidou positioning technologies; key frame sampling and detection are performed on the real-time obtained logistics parcel video to obtain the position information of the parcel, and a multi-object tracking algorithm is used for continuous tracking; the data that can be obtained for identifying damaged parcels is obtained, and the damaged and lost data is repaired according to the obtained data; after repair, the logistics information of the parcel can be clearly known, so as to continue the distribution of the parcel; the problem that the courier cannot continue the distribution after the label is worn or damaged is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0024] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] To make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention.

[0026] Please refer to Figure 1 , in an embodiment of the present invention, a multi-dimensional logistics parcel identification and tracking method based on deep learning includes the following steps:

[0027] S1. Data collection: Collect relevant information of the parcel;

[0028] S2. Data preprocessing: Preprocess the collected data, and divide the data into a training set, a validation set and a test set;

[0029] S3. Select a deep learning model: Identify the parcel through an object detection model, extract features from the image and detect the position of the parcel;

[0030] S4. Model Training: Train the selected deep learning model with the training set data;

[0031] S5. Multi-object Tracking: Use the tracking algorithm combined with the detected package location information to achieve real-time tracking of multiple packages;

[0032] S6. Feature Extraction and Fusion: Extract the weight, size information, and transportation path of the package, and fuse the weight, size information, and transportation path;

[0033] S7. Model Optimization and Adjustment: Optimize and adjust the model based on the results of the validation set;

[0034] S8. Real-time Data Collection and Tracking: When collecting package data for the first time, the data needs to be stored; Use GPS and Beidou positioning technologies to perform real-time tracking and positioning of the package; Perform key frame sampling and detection on the real-time obtained logistics package video to obtain the location information of the package, and use the multi-object tracking algorithm for continuous tracking;

[0035] S9. Repair and Tracking of Damaged Package Information: Identify the data that can be obtained for damaged packages, and repair the damaged and lost data based on the obtained data.

[0036] Preferably, it further includes S10. Model Learning and Improvement: Continuously train and adjust the model as more data accumulates.

[0037] Preferably, it further includes S11. User Interaction: Users can query the status of packages in real time.

[0038] Preferably, in S1, capture the appearance picture or video stream of the package through a camera; Obtain the identity identification information of the package through an RFID reader; The sensor is used to measure the physical characteristics of the package such as size and weight.

[0039] Preferably, in S2, the data preprocessing includes cropping, resizing, and normalizing the image.

[0040] Preferably, in S3, the identification of the package includes barcodes, QR codes, and extracting the package ID directly from the text information on the surface of the package; The deep learning model predicts the best transportation route and estimated arrival time of the package.

[0041] Preferably, in S6, combine the motion and surface feature information of the logistics package, and perform weighted fusion of the appearance information based on the minimum cosine distance of the deep feature vector and the motion information based on the Mahalanobis distance.

[0042] The working principle of the present invention is as follows: collect relevant information of the package; preprocess the collected data and divide the data into a training set, a validation set and a test set; identify the package through a target detection model, extract features from the image and detect the position of the package; train the selected deep learning model with the training set data; adopt a tracking algorithm combined with the detected package position information to achieve real-time tracking of multiple packages; extract the weight, size information and transportation path of the package, and fuse the weight, size information and transportation path; optimize and adjust the model according to the results of the validation set; store the data when the package data is collected for the first time; perform real-time tracking and positioning of the package through GPS and Beidou positioning technologies; perform key frame sampling and detection on the real-time obtained logistics package video, obtain the position information of the package, and continuously track it using a multi-target tracking algorithm; identify the data that can be obtained for damaged packages, and repair the damaged and lost data according to the obtained data; continuously train and adjust the model with the accumulation of more data; users can query the status of the package in real time.

[0043] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacement on some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-dimensional logistics package identification and tracking method based on deep learning, characterized in that: The following steps are involved: S1. Data collection: collect relevant information of the package; S2, data preprocessing: preprocess the collected data and divide the data into training set, validation set and test set; S3. Select a deep learning model: Use the object detection model to identify the package, extract features from the image and detect the location of the package; S4, model training: training the selected deep learning model using training set data; S5, Multi-target tracking: Use tracking algorithms combined with detected package location information to achieve real-time tracking of multiple packages; S6, feature extraction and fusion: extract the weight, size information, and transportation path of the package, and fuse the weight, size information, and transportation path; S7. Model optimization and adjustment: Optimize and adjust the model based on the results of the validation set; S8, Real-time collection and tracking: When the package data is collected for the first time, the data needs to be stored; the package is tracked and located in real time through GPS and Beidou positioning technology; key frame sampling and detection are performed on the real-time logistics package video to obtain the location information of the package, and continuous tracking is performed using a multi-target tracking algorithm; S9. Damaged package information repair tracking: Identify the data that can be obtained from the damaged package and repair the damaged and lost data based on the obtained data.

2. According to claim 1, a multi-dimensional logistics package identification and tracking method based on deep learning is characterized in that: Also includes S10, model learning and improvement: continuously training and adjusting the model as more data is accumulated.

3. According to the multi-dimensional logistics package identification and tracking method based on deep learning in claim 1, it is characterized in that: It also includes S11, user interaction: users can check the status of the package in real time.

4. According to the multi-dimensional logistics package identification and tracking method based on deep learning in claim 1, it is characterized in that: In S1, a camera is used to capture a picture or video stream of the package's appearance; an RFID reader is used to obtain the package's identity information; and a sensor is used to measure the package's physical properties such as size and weight.

5. According to the multi-dimensional logistics package identification and tracking method based on deep learning in claim 1, it is characterized in that: The data preprocessing in S2 includes image cropping, resizing, and normalization.

6. The multi-dimensional logistics package identification and tracking method based on deep learning according to claim 1 is characterized in that: The package identification in S3 includes extracting the package ID from the barcode, QR code and text information on the package surface directly; the deep learning model predicts the best transportation route and estimated arrival time of the package.

7. The multi-dimensional logistics package identification and tracking method based on deep learning according to claim 1 is characterized in that: In the S6, the movement and surface feature information of the logistics package are combined to weightedly fuse the appearance information based on the minimum cosine distance of the depth feature vector and the movement information based on the Mahalanobis distance.

Images