A logistics product positioning data management method and system based on cloud computing

Abstract

The present invention relates to the technical field of logistics data management, and discloses a logistics product positioning data management method and system based on cloud computing. The method comprises obtaining video stream images and positioning images of different types of logistics products, and marking timestamps; performing association processing according to the video stream images and the positioning images to obtain an image association set; performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector; performing a second feature extraction based on a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point; performing category matching on the three-dimensional scale feature point to obtain a matching result; performing parameter extraction according to the matching result to obtain a positioning parameter and a scale parameter; performing parameter matching according to the positioning parameter and the scale parameter to obtain a corresponding logistics product label. The method can improve the accuracy of the positioning and scale parameters of logistics products, and achieve efficient positioning of logistics products.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics data management, and in particular to a logistics product positioning data management method and system based on cloud computing. Background Art

[0002] Logistics product location data management refers to the collection, storage, processing, analysis and management of information such as the location, status and flow direction of logistics products (such as parcels, goods, containers, etc.) during the logistics process, so as to achieve real-time tracking, precise positioning and efficient management of logistics products. This process usually includes data collection, processing and application, aiming to provide timely and accurate location information for each link in the logistics supply chain (such as warehousing, transportation, distribution, etc.) to improve logistics efficiency, optimize resource allocation and reduce operating costs.

[0003] In complex logistics scenarios, logistics product positioning requires processing a large amount of video streams and positioning data. However, traditional logistics product positioning data management methods are mainly based on single image processing and positioning technologies, such as optical cameras, ultrasonic sensors, etc. These methods face problems with positioning accuracy and stability in practical applications.

[0004] In summary, among the existing technologies, traditional technologies such as cameras and sensors can no longer meet the increasingly complex logistics scenarios today, and it is difficult to achieve logistics product positioning quickly and efficiently. Summary of the invention

[0005] The present invention provides a logistics product positioning data management method and system based on cloud computing, which can improve the accuracy of the positioning and scale parameters of logistics products and achieve efficient and accurate positioning of logistics products.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a logistics product location data management method based on cloud computing, comprising:

[0007] Obtain video stream images of different types of logistics products and mark the first timestamp;

[0008] Acquire positioning images of the different types of logistics products and mark them with a second timestamp;

[0009] Performing association processing on the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image;

[0010] The image association set is stored in a cloud storage database based on memory cache, and a first feature extraction is performed on the image association set to obtain a dynamic feature vector and a static feature vector;

[0011] According to the dynamic feature vector and the static feature vector, a second feature extraction is performed based on a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point;

[0012] Performing category matching on the three-dimensional scale feature points to obtain a matching result;

[0013] According to the matching results, parameter extraction is performed to obtain positioning parameters and scale parameters;

[0014] According to the positioning parameters and the scale parameters, parameter matching is performed to obtain a corresponding logistics product label.

[0015] Preferably, the performing association processing according to the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image includes:

[0016] Matching the same timestamp data in the first timestamp and the second timestamp to obtain a video stream image and a positioning image with the same timestamp data;

[0017] The video stream image and the positioning image are associated to obtain an image association set of the video stream image and the positioning image.

[0018] Preferably, the storing of the image association set in a cloud storage database based on memory cache and performing first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector include:

[0019] Performing a first preprocessing on the video stream image based on a convolutional neural network to obtain a dynamic feature vector;

[0020] The positioning image is subjected to a second preprocessing based on a convolutional neural network to obtain a static feature vector.

[0021] Preferably, the first preprocessing of the video stream image based on a convolutional neural network to obtain a dynamic feature vector includes:

[0022] Based on the convolutional neural network, the video stream image is subjected to denoising, image enhancement, image defogging, and image segmentation to obtain a target area of ​​the video stream image;

[0023] The video stream image is converted into a dynamic feature vector based on the convolutional neural network.

[0024] Preferably, the second preprocessing of the positioning image based on a convolutional neural network to obtain a static feature vector includes:

[0025] Based on the convolutional neural network, the positioning image is subjected to denoising, image enhancement, image defogging, and label recognition to obtain a positioning label image;

[0026] The positioning label image is converted into a static feature vector based on the convolutional neural network.

[0027] Preferably, the extracting a second feature based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point comprises:

[0028] Associating the dynamic feature vector with the static feature vector based on an image association set to obtain an association vector;

[0029] The association vector is scaled and translated, and mapped into a three-dimensional space to obtain a three-dimensional scale feature point.

[0030] Preferably, performing category matching on the three-dimensional scale feature points to obtain a matching result includes:

[0031] Describing the three-dimensional scale feature points based on a global feature descriptor to obtain a feature description value;

[0032] According to the feature description value, three-dimensional scale feature point matching is performed based on the NNDR algorithm to obtain a matching result.

[0033] In a second aspect, the present invention provides a cloud computing-based logistics product positioning data management device, comprising: a first image acquisition module, for acquiring video stream images of different types of logistics products and marking them with a first timestamp;

[0034] A second image acquisition module, used to acquire positioning images of the different types of logistics products and mark them with a second timestamp;

[0035] An association processing module, used for performing association processing according to the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image;

[0036] A first feature extraction module, used for storing the image association set in a cloud storage database based on memory cache, and performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector;

[0037] A second feature extraction module, configured to perform second feature extraction based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point;

[0038] A feature matching module, used to perform type matching on the three-dimensional scale feature points to obtain a matching result;

[0039] A parameter extraction module, used to extract parameters according to the matching results to obtain positioning parameters and scale parameters;

[0040] The parameter matching module is used to perform parameter matching according to the positioning parameter and the scale parameter to obtain a corresponding logistics product label.

[0041] In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned cloud computing-based logistics product positioning data management methods.

[0042] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned cloud computing-based logistics product positioning data management methods.

[0043] Compared with the prior art, the present invention has the following beneficial effects: the embodiment of the present invention provides a method and system for managing logistics product positioning data based on cloud computing. The method includes: obtaining video stream images of different types of logistics products and marking them with a first timestamp; obtaining positioning images of the different types of logistics products and marking them with a second timestamp; performing association processing based on the video stream images and the positioning images to obtain an image association set of the video stream images and the positioning images; storing the image association set in a cloud storage database based on memory cache, and performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector; performing a second feature extraction based on a three-dimensional scale feature point recognition model based on the dynamic feature vector and the static feature vector to obtain a three-dimensional scale feature point; performing category matching on the three-dimensional scale feature point to obtain a matching result; performing parameter extraction based on the matching result to obtain a positioning parameter and a scale parameter; performing parameter matching based on the positioning parameter and the scale parameter to obtain a corresponding logistics product label.

[0044] In the present invention, the method can pre-process the video stream image and the positioning image respectively through a convolutional neural network to extract dynamic features and static features. Based on these features, the three-dimensional scale feature points are identified and matched to realize the feature extraction of logistics products. Finally, by matching the positioning parameters and scale parameters, the corresponding logistics product labels are accurately identified. The method improves the accuracy of logistics product positioning by combining dynamic and static features and three-dimensional feature point recognition. At the same time, the method uses cloud computing and memory caching technology to accelerate the processing and storage of large-scale data and realize efficient management of logistics product positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1It is a flow chart of a method for managing logistics product location data based on cloud computing provided by the first embodiment of the present invention;

[0046] Figure 2 It is a structural diagram of a cloud computing-based logistics product positioning data management device provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Reference Figure 1 The first embodiment of the present invention provides a logistics product positioning data management method based on cloud computing, comprising the following steps:

[0049] S11, obtaining video stream images of different types of logistics products and marking the first timestamp;

[0050] S12, obtaining positioning images of the different types of logistics products and marking them with a second timestamp;

[0051] S13, performing association processing according to the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image;

[0052] S14, storing the image association set in a cloud storage database based on memory cache, and performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector;

[0053] S15, performing second feature extraction based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point;

[0054] S16, performing category matching on the three-dimensional scale feature points to obtain a matching result;

[0055] S17, extracting parameters according to the matching results to obtain positioning parameters and scale parameters;

[0056] S18, performing parameter matching according to the positioning parameter and the scale parameter to obtain a corresponding logistics product label.

[0057] It is worth noting that logistics product location data management refers to the collection, storage, processing, analysis and management of information such as the location, status and flow direction of logistics products (such as parcels, goods, containers, etc.) during the logistics process, so as to achieve real-time tracking, precise positioning and efficient management of logistics products. This process usually includes data collection, processing and application, aiming to provide timely and accurate location information for each link in the logistics supply chain (such as warehousing, transportation, distribution, etc.) to improve logistics efficiency, optimize resource allocation and reduce operating costs.

[0058] In order to facilitate the understanding of the present invention, some preferred embodiments of the present invention are further described below.

[0059] In step S11, the video stream image can be acquired using a high-definition camera, a drone, a car camera or other intelligent devices. Depending on the logistics scenario, the camera can be installed on a storage facility, a transport vehicle or a delivery robot, and these devices need to have the ability to transmit in real time. Among them, the high-definition camera can provide a high-resolution video stream, which is helpful for subsequent detail extraction and item tracking. Equipment such as drones can also collect video data outside the logistics center over a larger range. In order to improve the quality of data acquisition, the video stream is usually preliminarily processed by an edge computing device. The edge device can adjust the frame rate, compress, and perform preliminary denoising on the video image to reduce bandwidth usage and ensure real-time data transmission.

[0060] It is worth noting that step S11 will obtain the video stream and positioning image of each existing logistics product, including different types of logistics products, so as to perform type matching in subsequent steps.

[0061] It is worth noting that a video stream is a continuously transmitted digital video data, which refers to streaming media data consisting of multiple video frames. Video streams are usually transmitted and played in real time through a network or storage device. Specifically, a video stream is a series of static images (frames) played quickly and continuously at a certain time interval, thus giving viewers a visual effect of a video.

[0062] It is worth noting that the timestamp annotation usually relies on the internal clock or network clock of the device. Among them, when the local clock of the device is used, the device that collects the video (such as a camera, drone, or car camera) is usually equipped with a built-in clock. When a video stream frame is captured, the device automatically generates a timestamp as the timestamp of the frame image. This is the basis for annotating the first timestamp. When the network clock is used, when multiple devices are simultaneously collecting, in order to ensure that all video stream images have a unified timestamp, the clocks of each device can be synchronized using the NTP protocol. In this way, the video frames obtained by each device can share the same global time base, avoiding problems caused by time differences between devices. In order to ensure that the timestamps of video stream images and other data (such as positioning data) can be synchronized, each time the camera captures a frame of image, the device will generate a timestamp at the moment of image capture, called the first timestamp. This timestamp is usually the system time of the device (in milliseconds) and is stored together with the frame image data. If necessary, the system not only needs to annotate the timestamp for the video stream image, but also needs to synchronize with other data sources (such as positioning data provided by GPS, IMU, etc.). In an environment with multiple data sources, all timestamps must be based on a unified time base. Through the NTP protocol or built-in time synchronization mechanism, the timestamp of the video frame can be ensured to be consistent with other data. It is worth noting that the timestamp generally adopts a standard format, such as the UNIX timestamp (that is, the number of seconds since January 1, 1970), because this format is simple and has good cross-platform compatibility. UTC time can also be used to ensure time synchronization worldwide. Regarding the storage of timestamps, they can be stored with each frame of the video stream image, embedded in the video file in the form of metadata, or recorded separately in the database. Whenever it is necessary to trace back or process data, the system can find the corresponding frame image by the timestamp.

[0063] In step S12, the positioning image can be acquired by: an optical camera and an image sensor, wherein optical cameras are widely used to capture positioning images of logistics products to record the current position of each logistics product. Optical images have the advantages of high resolution and clear images, and are suitable for identifying the appearance features of products, packages or goods, and extracting relevant location information through image processing algorithms; fixed camera shooting. In logistics management scenarios, fixed cameras are usually installed at specific locations (such as warehouse entrances, conveyor belts, packaging areas, etc.). These cameras can take a static positioning image when the logistics product passes or stays. Since the movement of the logistics product is intermittent, this static image can accurately record the position of the logistics product at a specific point in time.

[0064] It is worth noting that the positioning image refers to a static image, specifically image data used to determine the static position and status of logistics products. The positioning image does not involve a continuous video stream, but a single frame image acquired at one time.

[0065] The acquisition and format of the second timestamp are substantially the same as those of the first timestamp, and will not be described in detail herein.

[0066] In step S13, association processing is performed based on the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image. Preferably, the association processing is performed based on the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image, including: matching the same timestamp data in the first timestamp and the second timestamp to obtain a video stream image and a positioning image with the same timestamp data; and associating the video stream image with the positioning image to obtain an image association set of the video stream image and the positioning image.

[0067] Specifically, in step S13, the association processing of the video stream image and the positioning image is achieved by timestamp matching. The main purpose of this step is to synchronize the image (dynamic data) in the continuous video stream captured by the camera with the acquired positioning image (static data), so as to form a complete image association set, which can accurately represent the state and position of the logistics product at a certain moment. When acquiring the video stream image, each frame of the video image will be marked with a first timestamp, which indicates the exact time point when the image frame is captured. When acquiring the positioning image, a second timestamp will also be marked for each positioning image, which indicates the time point when the static image (usually a specific location image of a certain cargo) is captured. The first step of the association processing is to match the first timestamp and the second timestamp. When the timestamps of the video stream image and the positioning image are the same or very close, the system believes that the two images reflect the state and position of the logistics product at the same time point, so they can be associated. In this way, the image in the video stream can be synchronized with the positioning image.

[0068] It is worth noting that in the video stream, each frame is marked with a timestamp, so when processing, the system will scan all frames and find those frames whose timestamps are the same as those of the positioning image. These frames may correspond to different states or angles of the logistics product, forming complete dynamic information. The positioning image usually captures the static position information of the logistics product at a certain moment. By matching the timestamp, the system can determine that the static image corresponds to certain frames in the video stream in time. Through the matched timestamp, the system associates the video stream image with the positioning image to generate an image association set. The association set contains information about the logistics product in two dimensions, dynamic and static, at a certain point in time. The video stream image provides the movement trajectory or state change of the logistics product, while the positioning image provides the specific location of the product. In the preferred embodiment, by matching the timestamp, it can be ensured that the video stream image and the positioning image are completely synchronized in time, thereby generating an accurate association set. These image association sets are the basis for subsequent image processing, especially in the links of positioning parameter extraction and three-dimensional feature point recognition, they provide the system with a reference basis in time and space.

[0069] In step S14, the image association set is stored in a cloud storage database based on memory cache, and a first feature extraction is performed on the image association set to obtain a dynamic feature vector and a static feature vector. Preferably, the image association set is stored in a cloud storage database based on memory cache, and a first feature extraction is performed on the image association set to obtain a dynamic feature vector and a static feature vector, including: performing a first preprocessing on the video stream image based on a convolutional neural network to obtain a dynamic feature vector; performing a second preprocessing on the positioning image based on a convolutional neural network to obtain a static feature vector. Preferably, the first preprocessing on the video stream image based on a convolutional neural network to obtain a dynamic feature vector includes: performing denoising, image enhancement, image defogging, and image segmentation on the video stream image based on the convolutional neural network to obtain a target area of ​​the video stream image; converting the video stream image into a dynamic feature vector based on the convolutional neural network. Preferably, the second preprocessing of the positioning image based on the convolutional neural network to obtain a static feature vector includes: denoising, image enhancement, image defogging, and label recognition on the positioning image based on the convolutional neural network to obtain a positioning label image; and converting the positioning label image into a static feature vector based on the convolutional neural network.

[0070] Specifically, in step S14, the system needs to store the generated image association set (formed by video stream images and positioning images through timestamp matching) in a cloud storage database based on memory cache, and perform feature extraction on these images, thereby providing a basis for subsequent logistics product positioning and analysis. The feature extraction process involves processing the video stream images and positioning images separately to extract their dynamic feature vectors and static feature vectors. The core of this process is to use convolutional neural networks (CNNs) to extract useful feature information from images.

[0071] Specifically, the image association set is stored in a cloud storage database based on memory cache, which is an efficient data storage method that uses memory cache to accelerate data access. Cloud storage is scalable and can handle large amounts of data, including video streams and image data. Memory cache is used to temporarily store data to improve system response speed. After the image association set is matched by timestamp, it will be stored in this database to ensure that subsequent data can be quickly accessed and processed. The video stream and positioning data of the logistics product are uploaded to the cloud storage system, processed and analyzed through the cloud computing platform, and efficient data storage and processing are achieved. This method provides more accurate and real-time data support for the accurate positioning of logistics products by extracting, processing and analyzing video streams and positioning data. Accurate positioning of logistics products plays a very important role in the optimization of economic activities and supply chains, and is of great significance for improving logistics efficiency and reducing costs.

[0072] Specifically, the goal of the first feature extraction is to extract the dynamic features of the video stream image and the static features of the positioning image from the image association set. The key to this step is to complete the image preprocessing and feature extraction through the convolutional neural network (CNN).

[0073] The video stream images contain continuous motion pictures of logistics products. By processing the video stream images with a convolutional neural network, dynamic information in these pictures (such as product movement, rotation, etc.) can be identified. The preprocessing steps include: denoising, which uses image processing technology to remove noise in video images to make the images clearer and reduce interference in subsequent feature extraction; image enhancement, which improves the brightness and contrast of the image, so that the details of the image are more prominent and convenient for subsequent recognition and processing; image defogging, in some complex environments, video images may become blurred due to weather or other factors, and the defogging step can clarify the image; image segmentation, which separates different parts of the image, especially the target area (i.e., logistics products) from the background. This step can help the convolutional neural network more accurately identify the dynamic information of logistics products. The preprocessed video stream images will be input into the convolutional neural network, and the network will extract dynamic features in the image. Dynamic feature vectors are high-dimensional representations of these features, which are used to describe the movement, changes, and trajectory information in video stream images. These dynamic feature vectors will be used to locate and track logistics products in subsequent steps.

[0074] The positioning image is a single-frame static image that represents the static state of the logistics product at a certain moment. Convolutional neural networks can also be used to process these images and extract static features. The preprocessing steps include: denoising, similar to video stream images, the denoising step can remove interference factors in the image to make the image clearer; image enhancement, by improving the details of the image, such as color, contrast, etc., to help identify the features of objects in the image; image defogging, making the image clearer and easier to identify labels; label recognition, the positioning image usually contains labels or logos of logistics products, and CNN will identify these labels to ensure that the system can match the image with specific logistics products. After preprocessing, the positioning image will extract its static features, such as color, shape, texture, etc. through the convolutional neural network. These features are represented by static feature vectors, which are descriptions of the location and state of logistics products at a specific point in time.

[0075] In step S15, a second feature extraction is performed based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point. Preferably, the second feature extraction is performed based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point, including: associating the dynamic feature vector and the static feature vector based on an image association set to obtain an association vector; scaling and translating the association vector and mapping it to a three-dimensional space to obtain a three-dimensional scale feature point.

[0076] Specifically, step S15 describes how to extract three-dimensional scale feature points from the video stream and the positioning image in the image association set of the logistics product through the three-dimensional scale feature point recognition model, so as to more accurately locate and identify the logistics product. This process combines the dynamic feature vectors extracted from the video stream image and the static feature vectors extracted from the positioning image, and maps them to the three-dimensional space, and then performs feature matching.

[0077] Specifically, the dynamic feature vector comes from the video stream image, which represents the motion characteristics of the logistics product (such as position change, rotation, direction of movement, etc.). These features reflect the dynamic information of the logistics product at different time points. The static feature vector comes from the positioning image, which represents the static characteristics of the logistics product at a certain moment (such as shape, color, texture, etc.). These features are used to describe the position and appearance state of the product in space. The video stream image and the positioning image are image association sets formed by timestamp matching. In this association set, the dynamic feature vector and the static feature vector are extracted separately by the convolutional neural network. Next, the system associates the dynamic feature vector with the static feature vector. This association processing means combining the dynamic information (from the video stream) and the static information (from the positioning image) at the same time point to form a complete association vector.

[0078] Specifically, the purpose of scaling is to uniformly process the data from different image sources (video stream images and positioning images), because the image data from different sources may have differences in scale or unit. For example, video stream images may be collected by different cameras with different resolutions or image ratios, resulting in different representations of the same object in different images. In the association vector, the scales of dynamic features and static features may not be consistent. Scaling is to unify the feature vectors from these different sources to the same scale or ratio through mathematical operations, so that the system can process them uniformly. For example, the size of all feature vectors is standardized to the range of 0 to 1, or scaled at a certain ratio so that the size of the same logistics product in different images is consistent. The purpose of translation transformation is to correct the position of the association vector. Even after scaling, the positions of logistics products in different images may still be different. For example, the positions of logistics products in video stream images and positioning images do not completely overlap, resulting in a shift in the position information represented in the feature space. Translation transformation can handle various forms of deviations. For example, the logistics products in the image may be offset on the X-axis, Y-axis or Z-axis. By performing translation operations on the corresponding axes, the positions of the logistics products in different images can be aligned. Three-dimensional space mapping refers to mapping the association vectors after scaling and translation transformation into three-dimensional space. This step converts the two-dimensional image information into a representation in three-dimensional space, which facilitates the subsequent identification and matching of three-dimensional feature points. By projecting the association vectors into three-dimensional space, the position and shape of the logistics products in space can be accurately described, especially the changes in height, width and depth. These three-dimensional scale feature points can be used to accurately locate logistics products and match them with other three-dimensional feature points. Finally, after scaling, translation transformation and three-dimensional space mapping, the system will generate a series of three-dimensional scale feature points, which can represent the key position information of logistics products in three-dimensional space. These three-dimensional feature points can be used for tracking, positioning and further analysis of logistics products.

[0079] In step S16, the three-dimensional scale feature points are matched by type to obtain a matching result; wherein the second three-dimensional scale feature points are three-dimensional scale feature points obtained by the same operation of other logistics products. Preferably, the three-dimensional scale feature points are matched by type to obtain a matching result, including: describing the three-dimensional scale feature points based on a global feature descriptor to obtain a feature description value; and matching the three-dimensional scale feature points based on the NNDR algorithm according to the feature description value to obtain a matching result.

[0080] Specifically, in step S16, the system needs to match the first three-dimensional scale feature point with the second three-dimensional scale feature point to determine the similarity or identity of the logistics products. The first three-dimensional scale feature point refers to the three-dimensional feature of the current logistics product, while the second three-dimensional scale feature point is the three-dimensional feature point of other logistics products (products that have undergone the same feature extraction process). By matching these two three-dimensional feature points, the system can determine whether the products are the same or similar, or confirm whether a logistics product is confused with other products.

[0081] Specifically, before matching the three-dimensional feature points, the system will describe each three-dimensional feature point. This process is called using a global feature descriptor. The global feature descriptor is a feature extraction method used to describe three-dimensional feature points. It can convert the complex information of three-dimensional points (such as shape and position) into a fixed numerical vector. The numerical vector is used to represent the uniqueness of this feature point. In order to compare and match these three-dimensional scale feature points, the system uses an algorithm called the Nearest Neighbor Distance Ratio (NNDR). The purpose of this algorithm is to determine the similarity of two feature points by calculating the distance between their feature description values. Through the above process, the system can obtain the matching results between each first three-dimensional scale feature point and the second three-dimensional scale feature point. If the nearest neighbor distance ratio (NNDR) is less than a certain threshold, the two three-dimensional feature points are considered to be matched successfully. The matching result indicates which logistics products are similar, or confirms whether a product is confused with other products. Through matching, the uniqueness and consistency of products in the logistics process can be ensured.

[0082] In step S17, according to the matching results, parameter extraction is performed to obtain positioning parameters and scale parameters. The main purpose of performing parameter extraction according to the matching results is to extract specific positioning parameters and scale parameters from the matched three-dimensional feature points, which are used to accurately describe the spatial position and size of the logistics product. In step S16, the system has obtained matching results through three-dimensional feature point matching. These matching results represent the similarity between the three-dimensional scale feature points of the current logistics product and the three-dimensional feature points of other logistics products. Each matching result includes a pair of successfully matched three-dimensional feature points and their relative position relationship in space.

[0083] It is worth noting that the parameter extraction method can be: first, the feature description value of the matching result is clustered into three-dimensional scale feature points. In the logistics product positioning system, the matching result contains a series of three-dimensional scale feature points. In order to simplify the processing and improve the calculation efficiency, these three-dimensional scale feature points are first clustered. Using a clustering algorithm (such as K-means or other clustering methods), the feature points are divided into multiple categories (C1, C2,…, Cn). Each category corresponds to the feature points of a certain part of the logistics product, such as different faces, corners or edge areas. After the clustering is completed, the center point is extracted from each cluster. The center point can be calculated by the centroid formula, that is, the geometric mean of all points in each category is calculated. By extracting the center point of each cluster (P1, P2,…, Pn), the key part of the logistics product can be more accurately located. Then, through the cluster matching algorithm, the center of the extracted feature point is matched with the positioning label (static feature) of the logistics product. This process combines the dynamic three-dimensional feature point information with the static positioning label information to ensure that the identified logistics products are accurate in spatial positioning and size. Through this matching, the system can extract the positioning parameters (X, Y, Z spatial coordinates) and scale parameters (such as length, width, height) of logistics products, thereby providing accurate data for subsequent logistics management and product tracking.

[0084] Specifically, the positioning parameter describes the location of the logistics product in three-dimensional space, usually including the location information on the three coordinate axes of X, Y, and Z. This parameter tells the system the specific location of the logistics product in the warehouse or during transportation. The scale parameter describes the spatial dimension information of the logistics product, usually including height, width, depth, etc. These parameters help the system understand the physical size of the product so as to make reasonable storage and transportation planning.

[0085] Specifically, the extracted positioning parameters and scale parameters can be used in logistics management for the following applications: product tracking and positioning. The system can use positioning parameters to track the location of products in different scenarios such as warehousing and transportation in real time, thereby ensuring the transparency and efficiency of the logistics process. For example, the system can know which shelf in the warehouse a specific product is currently on, or whether the position of the product has shifted during transportation. Automated storage and transportation planning. The scale parameters help the system automatically plan the placement of products in warehousing or transportation based on their size. For example, in a warehouse with limited space, the system can reasonably arrange the storage location of each product based on the size and shape of the product to maximize the use of storage space. Similarly, transport vehicles can also plan the loading sequence based on the size of the product to ensure safe and efficient loading and unloading.

[0086] In a preferred embodiment, a clustering method can be applied to further optimize the extraction process. The three-dimensional scale feature points obtained by matching must first be subjected to cluster analysis. Because in actual applications, the feature points on different logistics products will have a certain degree of dispersion, and directly using these feature points to extract parameters may lead to inaccuracy. Through clustering, the feature points can be reasonably grouped to ensure that the feature points in each cluster are highly related or close. The system divides them into several clusters based on the distance between the feature points, and each cluster contains several similar or close feature points. For example, one cluster may represent a plane area of ​​a logistics product, and another cluster represents the edge or corner of the product. Through clustering, the system can identify the feature points of different parts of the product surface. After clustering is completed, the system will analyze each cluster and extract the center point of the cluster. The center point refers to the geometric center of all feature points in a cluster, which can usually represent the overall position of the cluster. The center point plays a key role in positioning and scale parameter extraction. For each cluster cluster, the system can calculate the position of the product in space through these center points. For example, if a cluster represents a face of a product, the center point of the cluster can represent the center position of the face. By combining the center points of multiple clusters, the system can accurately determine the overall location of the product. By calculating the distance between different clusters, the system is able to derive the size information of the logistics product. For example, the distance between two cluster centers can represent the width, length or height of the product.

[0087] In step S18, parameter matching is performed according to the positioning parameters and the scale parameters to obtain the corresponding logistics product label. The system further obtains the positioning label data of the logistics product through the positioning parameters and scale parameters extracted above, and stores these data in the positioning product database in the cloud storage database. This step ensures that the location information and size data of the logistics product can be saved in real time and managed through cloud storage, thereby providing data support for subsequent tracking, management and analysis.

[0088] It is worth noting that the system obtains the spatial positioning parameters of the logistics product from the extracted three-dimensional feature points and matches them with the standard positioning data of the logistics products already in the database. This database stores the standard size and position data of all logistics products. During matching, the system compares the positioning parameters of the current logistics product with the reference positioning parameters in the database to ensure that the position of the product in the warehouse or during transportation is accurate. Specifically, during the matching process, the system uses the scale parameters extracted from the three-dimensional feature points (such as the height, width, and depth of the product) and compares these parameters with the standard size information in the database. If the actual size of the logistics product is within the set threshold range with the standard size in the database, the product size is considered to be correct and can be identified as the correct product. When the positioning parameters and scale parameters are matched successfully, the system will generate a unique logistics product label for the logistics product, namely the positioning label data.

[0089] Specifically, the positioning tag data usually includes the basic information of the logistics product (such as product ID, type, etc.) and the spatial information of the product (positioning parameters and scale parameters). These data are combined to form a complete positioning tag, which is used to identify and track each logistics product. Once the positioning tag data is generated, the system will store this data in the cloud storage database. The role of the cloud storage database is to provide efficient and real-time storage and management capabilities, so that the system can quickly access and update this data.

[0090] The following describes the working process of the present invention by taking a relatively common scenario as an example. The working process is as follows:

[0091] In a large warehouse, multiple cameras are installed in different locations to track and record the location and status of each logistics product passing through the warehouse entrance and exit and conveyor belt. The entire process uses a cloud computing platform to manage and store this data.

[0092] When a logistics product enters the warehouse, the camera installed at the entrance captures its video stream image and marks it with a first timestamp. At the same time, the camera at a specific location captures a static positioning image of the product and marks it with a second timestamp.

[0093] The system associates the video stream images and positioning images through timestamp matching to form an image association set containing dynamic and static information of logistics products.

[0094] The image association set is stored in a cloud storage database based on memory cache, and feature extraction is performed through a convolutional neural network to obtain dynamic feature vectors and static feature vectors.

[0095] The system further extracts three-dimensional scale feature points of logistics products, including a combination of dynamic and static information.

[0096] Match the 3D feature points of the current logistics product with the 3D feature points of other logistics products in the database to determine whether it is similar to other products.

[0097] Based on the matching results, the product's positioning parameters (the product's spatial position) and scale parameters (the product's size) are extracted.

[0098] Through these parameters, a positioning label of the logistics product is generated, and the label data is stored in the positioning product database.

[0099] In summary, the present invention provides an inverter operation control method, comprising: obtaining video stream images of different types of logistics products and marking them with a first timestamp; obtaining positioning images of the different types of logistics products and marking them with a second timestamp; performing association processing according to the video stream images and the positioning images to obtain an image association set of the video stream images and the positioning images; storing the image association set in a cloud storage database based on memory cache, and performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector; performing a second feature extraction based on a three-dimensional scale feature point recognition model according to the dynamic feature vector and the static feature vector to obtain a three-dimensional scale feature point; performing category matching on the three-dimensional scale feature points to obtain a matching result; performing parameter extraction according to the matching result to obtain a positioning parameter and a scale parameter; performing parameter matching according to the positioning parameter and the scale parameter to obtain a corresponding logistics product label.

[0100] In the present invention, the method improves the accuracy of logistics product positioning by combining dynamic and static features and three-dimensional feature point recognition. At the same time, the method uses cloud computing and memory caching technology to accelerate the processing and storage of large-scale data and realize efficient management of logistics product positioning.

[0101] Reference Figure 2 The second embodiment of the present invention provides a logistics product positioning data management device based on cloud computing, including:

[0102] A first image acquisition module, used to acquire video stream images of different types of logistics products and mark them with a first timestamp;

[0103] A second image acquisition module, used to acquire positioning images of the different types of logistics products and mark them with a second timestamp;

[0104] An association processing module, used for performing association processing according to the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image;

[0105] A first feature extraction module, used for storing the image association set in a cloud storage database based on memory cache, and performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector;

[0106] A second feature extraction module, configured to perform second feature extraction based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point;

[0107] A feature matching module, used to perform type matching on the three-dimensional scale feature points to obtain a matching result;

[0108] A parameter extraction module, used to extract parameters according to the matching results to obtain positioning parameters and scale parameters;

[0109] The parameter matching module is used to perform parameter matching according to the positioning parameter and the scale parameter to obtain a corresponding logistics product label.

[0110] In an optional implementation, the second image acquisition module is specifically used to:

[0111] Matching the same timestamp data in the first timestamp and the second timestamp to obtain a video stream image and a positioning image with the same timestamp data;

[0112] The video stream image and the positioning image are associated to obtain an image association set of the video stream image and the positioning image.

[0113] In an optional implementation, the first feature extraction module is specifically used to:

[0114] Performing a first preprocessing on the video stream image based on a convolutional neural network to obtain a dynamic feature vector;

[0115] The positioning image is subjected to a second preprocessing based on a convolutional neural network to obtain a static feature vector.

[0116] In an optional implementation, the first feature extraction module is specifically used to:

[0117] Based on the convolutional neural network, the video stream image is subjected to denoising, image enhancement, image defogging, and image segmentation to obtain a target area of ​​the video stream image;

[0118] The video stream image is converted into a dynamic feature vector based on the convolutional neural network.

[0119] In an optional implementation, the first feature extraction module is specifically used to:

[0120] Based on the convolutional neural network, the positioning image is subjected to denoising, image enhancement, image defogging, and label recognition to obtain a positioning label image;

[0121] The positioning label image is converted into a static feature vector based on the convolutional neural network.

[0122] In an optional implementation, the second feature extraction module is specifically used to:

[0123] Associating the dynamic feature vector with the static feature vector based on an image association set to obtain an association vector;

[0124] The association vector is scaled and translated, and mapped into a three-dimensional space to obtain a three-dimensional scale feature point.

[0125] In an optional implementation, the second feature extraction module is specifically used to:

[0126] Describing the three-dimensional scale feature points based on a global feature descriptor to obtain a feature description value;

[0127] According to the feature description value, three-dimensional scale feature point matching is performed based on the NNDR algorithm to obtain a matching result.

[0128] It should be noted that a cloud computing-based logistics product positioning data management device provided in an embodiment of the present invention is used to execute all the process steps of a cloud computing-based logistics product positioning data management method in the above embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0129] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a cloud computing-based logistics product positioning data management method program. When the processor executes the computer program, the steps in the above-mentioned cloud computing-based logistics product positioning data management method embodiments are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the second feature extraction module.

[0130] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0131] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0132] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0133] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0134] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0135] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0136] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A logistics product positioning data management method based on cloud computing, characterized in that: include: Obtain video stream images of different types of logistics products and mark the first timestamp; Acquire positioning images of the different types of logistics products and mark them with a second timestamp; Performing association processing on the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image; The image association set is stored in a cloud storage database based on memory cache, and a first feature extraction is performed on the image association set to obtain a dynamic feature vector and a static feature vector; According to the dynamic feature vector and the static feature vector, a second feature extraction is performed based on a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point; Performing category matching on the three-dimensional scale feature points to obtain a matching result; According to the matching results, parameter extraction is performed to obtain positioning parameters and scale parameters; According to the positioning parameters and the scale parameters, parameter matching is performed to obtain a corresponding logistics product label; The step of performing association processing on the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image comprises: matching the same timestamp data in the first timestamp and the second timestamp to obtain a video stream image and a positioning image with the same timestamp data; associating the video stream image with the positioning image to obtain an image association set of the video stream image and the positioning image; The step of extracting a second feature based on the dynamic feature vector and the static feature vector and based on a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point comprises: associating the dynamic feature vector and the static feature vector based on an image association set to obtain an association vector; scaling and translating the association vector and mapping it to a three-dimensional space to obtain a three-dimensional scale feature point; The performing type matching on the three-dimensional scale feature points to obtain a matching result includes: describing the three-dimensional scale feature points based on a global feature descriptor to obtain a feature description value; performing three-dimensional scale feature point matching based on an NNDR algorithm according to the feature description value to obtain a matching result; Among them, the first three-dimensional scale feature point represents the three-dimensional feature of the current logistics product, and the second three-dimensional scale feature point is the three-dimensional feature point of other logistics products that have undergone the same feature extraction process; the first three-dimensional scale feature point is matched with the second three-dimensional scale feature point, and the complex information of the three-dimensional feature point is converted into a fixed numerical vector using a global feature descriptor, and then the distance between the feature description values ​​of the two feature points is calculated using the nearest neighbor distance ratio algorithm. When the nearest neighbor distance ratio is less than the threshold, the two three-dimensional feature points are considered to be matched successfully.

2. The cloud computing-based logistics product location data management method according to claim 1, characterized in that: The storing of the image association set in a cloud storage database based on memory cache and performing first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector include: Performing a first preprocessing on the video stream image based on a convolutional neural network to obtain a dynamic feature vector; The positioning image is subjected to a second preprocessing based on a convolutional neural network to obtain a static feature vector.

3. The method for managing logistics product location data based on cloud computing according to claim 2, characterized in that: The first preprocessing of the video stream image based on the convolutional neural network to obtain a dynamic feature vector includes: Based on the convolutional neural network, the video stream image is subjected to denoising, image enhancement, image defogging, and image segmentation to obtain a target area of ​​the video stream image; The video stream image is converted into a dynamic feature vector based on the convolutional neural network.

4. The method for managing logistics product location data based on cloud computing according to claim 2, characterized in that: The second preprocessing of the positioning image based on the convolutional neural network to obtain a static feature vector includes: Based on the convolutional neural network, the positioning image is subjected to denoising, image enhancement, image defogging, and label recognition to obtain a positioning label image; The positioning label image is converted into a static feature vector based on the convolutional neural network.

5. A cloud computing-based logistics product positioning data management device, characterized in that: A method for managing logistics product location data based on cloud computing as claimed in any one of claims 1 to 4, comprising: A first image acquisition module, used to acquire video stream images of different types of logistics products and mark them with a first timestamp; A second image acquisition module, used to acquire positioning images of the different types of logistics products and mark them with a second timestamp; An association processing module, used for performing association processing according to the video stream image and the positioning image to obtain an image association set of the video stream image and the positioning image; A first feature extraction module, used for storing the image association set in a cloud storage database based on memory cache, and performing a first feature extraction on the image association set to obtain a dynamic feature vector and a static feature vector; A second feature extraction module, configured to perform second feature extraction based on the dynamic feature vector and the static feature vector and a three-dimensional scale feature point recognition model to obtain a three-dimensional scale feature point; A feature matching module, used to perform type matching on the three-dimensional scale feature points to obtain a matching result; A parameter extraction module, used to extract parameters according to the matching results to obtain positioning parameters and scale parameters; The parameter matching module is used to perform parameter matching according to the positioning parameter and the scale parameter to obtain a corresponding logistics product label.

6. An electronic device, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the cloud computing-based logistics product positioning data management method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the cloud computing-based logistics product positioning data management method as described in any one of claims 1 to 4.

Images