Container number identification compensation system, individual identification and storage method and device

Abstract

The invention provides a container number identification compensation system, an individual identification and storage method and equipment. The container number identification compensation system comprises an individual feature library part and an individual feature service part. The individual feature library part comprises a key value database and a vector database; the individual feature service part comprises an individual feature Web service module, an individual feature extraction service module, a vector service module and a key value service module. According to the invention, when the OCR container number cannot be identified, the original manual identification operation can be replaced, and the workload of operators can be reduced; the method also has the advantages of high operation performance and second-level execution efficiency, can complete the automatic compensation of OCR box number recognition in real time, improves the operation efficiency of related services, and reduces the operation cost. Identification is carried out aiming at the individual characteristics of the container, the overall identification rate can be improved to 99.0%-99.9%, and the compensation effect of the OCR container number identification technology is fully played.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, specifically to the technical field of container recognition, and particularly to a container number recognition compensation system, an individual recognition and storage method, and a device. Background Art

[0002] Container recognition technology is an important technology in the modern logistics field. It realizes the rapid recognition of containers through advanced technologies such as image recognition and character recognition, and is widely used in business scenarios such as automatic entry of container operation data and automatic confirmation of operation links. This technology has greatly improved logistics efficiency and reduced operating costs, and has broad application prospects and development space.

[0003] The core of container recognition technology is the identification of individual containers. The container number is the unique identifier of an individual container. As long as the container number printed on the container is recognized, the identification of the individual container can be achieved. Alternatively, by extracting, marking, storing, searching, and matching container features, the identification of individual containers can also be achieved. The former is called the container number recognition method, and the latter is called the container feature recognition method.

[0004] For the container number recognition method, the currently commonly used is OCR (Optical Character Recognition) technology, that is, optical character recognition technology. This technology captures images of the container surface through a high-definition camera and uses advanced technologies such as deep learning to process the images, so as to achieve automatic recognition and recording of the container number. However, this technology highly depends on image quality, and in actual scenarios, they are all in outdoor environments, and the image quality is greatly affected by interference factors. For example, the lens distortion problem caused by an unsatisfactory camera installation position, the veiling glare and uneven illumination problems caused by sunlight changes and shadows, and the blurred vision and occlusion problems under special weather conditions (such as heavy fog, heavy rain, ice and snow, etc.), especially the problems of blurred and damaged container numbers caused by the defects of the container itself (rust, dirt, paint peeling, etc.), and even the container numbers are difficult to identify with the naked eye. The problems of blurred and damaged container numbers basically lock the upper limit of the OCR container number recognition rate. The older the container, the more difficult it is to directly recognize the container number through OCR. There are two situations where the OCR container number cannot be recognized. Either the characters cannot be recognized at all, or only some characters are recognized. Therefore, under the condition of permission, a compensation mechanism can be enabled for these two situations to output as exact a container number as possible.

[0005] Traditional container number recognition compensation mechanisms and methods use some characters recognized by OCR to perform fuzzy matching of the container number with the alternative container list provided by an external system, and infer the container number with the closest matching degree. However, this fuzzy matching method does not work well in the actual production environment. The error rate of these characters recognized by OCR is higher than that of normal recognition, and the overall confidence level is insufficient.

[0006] Since shipping companies set the usage period of containers to be phased out after 10 - 20 years, over time, containers are exposed to sun and rain, and the number of loading and unloading operations increases, making the containers increasingly old. Once the container number is blurred or damaged, it is difficult to directly identify the container number through OCR, and the recognition rate is only between 85% - 90%. Summary of the Invention The purpose of the present invention is to solve the problem that commonly exists in the OCR container number recognition technology, that is, due to reasons such as rust, dirt, and paint peeling on the surface of the container, the container becomes increasingly old, and once the container number is blurred or damaged, it is difficult to directly identify the container number through OCR.

[0007] To overcome the defect that the recognition rate of directly identifying the container number through OCR in the prior art is relatively low; the present invention provides a container number recognition compensation system, an individual recognition and storage method, and a device. The targeted compensation solution proposed by the present invention has the characteristics of strong applicability and effectiveness; it can replace the original manual recognition operation when the OCR container number cannot be recognized, reducing the work burden of operators; moreover, it also has the advantages of high computing performance and second-level execution efficiency, and can automatically complete the compensation of OCR container number recognition in real time, improving the operation efficiency of related services and reducing the operation cost; by identifying the individual characteristics of containers, the overall recognition rate can be increased to between 99.0% - 99.9%, playing the role of compensating for OCR container number recognition.

[0008] The present invention solves the above technical problems through the following technical solutions: The present invention provides a container number recognition compensation system, which includes an individual feature library part and an individual feature service part; The individual feature library part includes: a vector database for storing the individual feature vectors of different containers; a key - value database for storing the container basic information corresponding to the individual feature vectors of each container, and the container basic information includes the container number; The individual feature service part includes: an individual feature Web service module for responding to service requests initiated by an external system; an individual feature extraction service module for responding to a first service request initiated by the individual feature Web service module; the first service request includes: converting the container surface photo included in the incoming data of the external system into a to-be-processed individual feature vector; a vector service module for responding to a second service request initiated by the individual feature Web service module, the second service request includes: searching or storing according to the to-be-processed individual feature vector in combination with the vector database, and obtaining a vector ID; a key-value service module for responding to a third service request initiated by the individual feature Web service module, the third service request includes: matching or storing according to the vector ID in combination with the key-value database, and obtaining a result.

[0009] In the present invention, the vector database is a database specifically used to store and query vectors, and the vectors stored therein are from the vectorization of texts, voices, images, videos, etc. Compared with traditional databases, the vector database can not only complete basic CRUD (add, read query, update, delete) operations, but also perform faster vector similarity searches on vector data. Vector similarity search is a process of comparing a query vector with a database to find the most similar and confident result vector to the query vector. The vector similarity search uses an approximate nearest neighbor search algorithm (ANN) to accelerate the search process. Common ANN algorithms include locality-sensitive hashing (LSH), graph embedding methods (such as HNSW), vector quantization (VQ), etc., which optimize the search process in different ways and play an important role in the field of image recognition.

[0010] In the present invention, the key-value database, also known as a key-value store, is a non-relational database (also known as a NoSQL database). Each unique identifier is stored as a key with a related value, and this data pairing is called a "key-value" pair. In a key-value pair, the key must be unique, and the value associated with the key can be accessed through the key. A single-node key-value database has performance advantages in reading and writing data compared with more general-purpose relational databases, and can be used in combination with a vector database to search for tagged data bound to feature vectors, such as container numbers.

[0011] In some embodiments, the vector service module includes a vector search service module; the key-value service module includes a key-value matching service module; the vector search service module is configured to search for the most similar and confident result vector to the feature vector to be processed in the vector database, and obtain the first vector ID recorded in the vector database; the key-value matching service module is configured to match the container basic information from the key-value database according to the first vector ID.

[0012] In some embodiments, the vector service module includes a vector storage service module; the key-value service module includes a key-value storage service module; the vector storage service module is configured to store the feature vector to be processed into the vector database, and obtain the second vector ID recorded in the vector database; the key-value storage service module is configured to combine the second vector ID with the container basic information included in the incoming data of the external system into a Key-Value key-value pair and store it into the key-value database.

[0013] In some embodiments, the deployment mode of the individual feature service part includes distributed deployment or microservice architecture; the public service is provided externally, and the individual feature library is operated internally.

[0014] In some embodiments, the individual feature Web service module is deployed in the DMZ area of the individual feature service part.

[0015] In some embodiments, the individual feature extraction service module is deployed in the intranet of the individual feature service part.

[0016] In the present invention, the Value value of the key-value database can save not only the container number, but also more information, such as basic information of the container type, load, etc.

[0017] In some embodiments, the key-value database includes an open-source key-value database.

[0018] In a specific embodiment, the key-value database includes a Redis key-value database.

[0019] In some embodiments, the deployment mode of the key-value database includes distributed deployment.

[0020] In some embodiments, the vector database includes an open-source vector database.

[0021] In a specific embodiment, the vector database includes a Milvus vector database.

[0022] In one embodiment, the Milvus vector database includes the HNSW algorithm; the HNSW refers to Hierarchical Navigable Small World graphs, which is a graph-based approximate nearest neighbor search algorithm widely used in the industry and is very suitable for applications that require high-precision ANN search (such as face recognition, semantic search), and can meet the requirements of this technical solution for the confidence of container individual feature recognition. It constructs a multi-layer graph structure, where the number of nodes in each layer gradually decreases, and the number of edges also decreases accordingly. In this way, during the search process from top to bottom, the globally possible nearest nodes can be found first through long links, and then local search can be performed in the lower layer, thereby achieving fast and accurate approximate nearest neighbor search.

[0023] In some embodiments, the deployment method of the vector database includes distributed deployment.

[0024] In a specific embodiment, the individual feature extraction service module includes the SIFT algorithm, which is used to extract individual feature points of the container from the container surface photo included in the incoming data and generate the individual feature vector to be processed.

[0025] Among them, the SIFT (Scale-Invariant Feature Transform) algorithm is a classic computer vision algorithm used to extract feature points from images and can identify target objects in complex scenes. The purpose of the SIFT algorithm is to represent the information in the image in the form of features, so that these features remain invariant under image scaling, rotation, brightness change, and even to a certain extent under affine transformation and other situations. The extraction of target feature points by the SIFT algorithm is essentially to statistically analyze and encode the gradient information in the neighborhood of the feature points, and finally combine them into a descriptor vector to characterize the direction and relative intensity distribution of the local features of the image. The SIFT algorithm has the characteristics of strong robustness and excellent anti-interference ability.

[0026] In the present invention, the container contour, size, and its own painting, as well as the individual unique marks left by the environment such as natural erosion and loading and unloading operations, are used as container individual features for extraction, marking, and storage in the container individual feature library. When the OCR box number cannot be recognized, recognition compensation is performed by searching and matching in the container individual feature library to obtain the box number that has been correctly marked in advance.

[0027] The present invention also provides a container individual recognition method, which uses the above-mentioned container box number recognition compensation system to provide an individual feature Web service and respond to public service requests initiated by external systems; The container individual identification method includes the following steps: S1. Receive the incoming data from the external system; and determine the program to be executed according to the incoming data; S2. Through the individual feature extraction service module, convert the container surface photo included in the incoming data into an individual feature vector to be processed, and execute a search process. The individual feature Web service module calls the vector search service of the vector service module and the key-value matching service of the key-value service module to feedback the searched information to the external system according to the individual feature vector to be processed.

[0028] In some embodiments, the container individual identification method satisfies the following Condition I or Condition II: Condition I: S1. Receive the incoming data from the external system, and the incoming data does not include the container surface photo; feedback from the individual feature Web service module to the external system, indicating a call failure; Condition II: S1. Receive the incoming data from the external system, and the incoming data only includes the container surface photo, and determine to execute a search process; S2. First, the individual feature Web service module calls the individual feature extraction service, and uses the SIFT algorithm to extract individual feature points from the container surface photo to obtain the individual feature vector to be processed; Secondly, the individual feature Web service module calls the vector search service; use the individual feature vector to be processed as a query vector, search for the most similar and confident result vector to the query vector from the vector database, and obtain the first vector ID recorded in the vector database; if the search fails, feedback from the individual feature Web service module to the external system, indicating that no search result is found; Then, the individual feature Web service module calls the key-value matching service; use the first vector ID as a search key value to match the container basic information from the key-value database, and the container basic information includes the container number; if the match is successful, the individual feature Web service module feedbacks the container basic information to the external system; if the match fails, feedback from the individual feature Web service module to the external system, indicating that no search result is found.

[0029] In the present invention, the container surface photo includes photos of 5 surfaces of the container, such as: left, right, front, back, and top surfaces.

[0030] The present invention also provides a container individual storage method, which uses the above-mentioned container number identification compensation system to provide an individual feature Web service and respond to a public service request initiated by an external system; The individual container storage method includes the following steps: S1. Receive the incoming data from the external system; and determine the program to be executed according to the incoming data; S2. Through the individual feature extraction service module, convert the container surface photo included in the incoming data into an individual feature vector to be processed, and execute the storage process. The individual feature Web service module calls the vector storage service of the vector service module and the key-value storage service of the key-value service module to combine the individual feature vector to be processed with the container basic information included in the incoming data into a Key-Value key-value pair, and store it in the key-value database. The container basic information includes the container number.

[0031] In some embodiments, the individual container storage method satisfies the following Condition I or Condition II: Condition I: S1. Receive the incoming data from the external system, and the incoming data does not include the container surface photo; feedback to the external system by the individual feature Web service module to indicate a call failure; Condition II: S1. Receive the incoming data from the external system, the incoming data includes the container surface photo and the container basic information, the container basic information includes the container number, and determine to execute the storage process; S2. First, the individual feature Web service module calls the individual feature extraction service, and uses the SIFT algorithm to extract individual feature points from the container surface photo to obtain the individual feature vector to be processed; Secondly, the individual feature Web service module calls the vector storage service; store the individual feature vector to be processed in the vector database to obtain the second vector ID recorded in the vector database; Then, the individual feature Web service module calls the key-value storage service; combine the second vector ID with the container basic information into a Key-Value key-value pair, and store it in the key-value database; when there is the same Key value in the key-value database, overwrite the old record; and feedback to the external system by the individual feature Web service module to confirm that the individual features of the container have been stored.

[0032] In the present invention, the container surface photo includes photos of 5 faces of the container, such as: left, right, front, back, and top surfaces.

[0033] The present invention also provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned container individual identification method or the above-mentioned container individual storage method.

[0034] The beneficial effects of the present invention are as follows: 1. The technical solution provided by the present invention can be used in combination with the OCR box number recognition technology, and the capabilities of the two are complementary; when the OCR box number cannot be recognized, it can replace the original manual recognition operation, reducing the workload of operators.

[0035] 2. The present invention has high computing performance and a second-level execution efficiency, can automatically compensate for OCR box number recognition in real time, improve the operation efficiency of related services, and reduce operating costs.

[0036] 3. For fuzzy and damaged box numbers, the OCR box number recognition rate is between 85% and 90%. After supplementing with the container individual feature recognition method of the present invention, the overall recognition rate of such containers is significantly improved, reaching between 99.0% and 99.9%.

[0037] 4. The present invention can not only compensate for the OCR box number recognition technology, but also be applied to business scenarios such as automatic entry of container operation data and automatic confirmation of operation links, including but not limited to intelligent tallying, intelligent road crossings, yard loading and unloading, customs inspection, unpacking and repacking, empty container inspection for damage, and safety production in container sea transportation, railway transportation, land transportation, and warehousing, which require container individual recognition and individual information collection (such as: box number recognition and box number information collection). BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic structural diagram of the container box number recognition compensation system according to Embodiment 1 of the present invention; Figure 2 It is a schematic flow diagram of the container individual recognition method and individual storage method according to Embodiment 2 of the present invention; Figure 3 It is a schematic diagram of the search process of the container individual recognition method according to Embodiment 2 of the present invention; Figure 4 It is a schematic diagram of the storage process of the container individual storage method according to Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The recognition system and method of the present invention are a container feature recognition system and method for distinguishing individuals. Using the container contour, size, and its own painting, as well as the individual specific traces left by the environment such as natural erosion and loading and unloading operations, as container individual features for extraction, marking, storage, search, and matching.

[0040] Among them, SIFT or other free algorithms are used to extract individual feature points of the container from the photos of the five container surfaces (left, right, front, back, and top), and generate a feature descriptor vector (i.e., a feature vector, which is called the "individual feature vector" of the container in the present invention), which is stored in the open-source vector database Milvus. The vector ID (as the Key value) obtained during storage is stored together with the manually marked container number (as the Value value, which can include other basic container information) in the open-source key-value database Redis. The combination of this vector database and key-value database is called the container "individual feature library".

[0041] In this way, as long as the OCR container number cannot be recognized and cannot be searched in the individual feature library, this technical solution can be enabled to store the individual feature vector of the container and the manually recognized container number into this individual feature library together. Then, when the OCR container number cannot be recognized next time, this technical solution can be enabled to search out the container number in this individual feature library.

[0042] The specific method is to extract the individual feature vector from the incoming container surface photo, perform a vector similarity search in the vector database as a query vector, and search for similar and confident records. Then, use the result vector ID (Key value) of the searched record to match in the key-value database to match the corresponding container number (Value value), realizing the compensation for OCR container number recognition. Normally, as long as the container is not old and the container number is clear, OCR can basically recognize it. At this time, the compensation mechanism does not need to be started, but the individual special vector and the recognized container number can be saved to the individual feature library for use when OCR cannot recognize next time.

[0043] The present invention will be further described below by way of examples, but the present invention is not limited to the scope of the described examples. The experimental methods without specific conditions in the following examples are carried out according to conventional methods and conditions, or according to the product instructions.

[0044] Example 1 This example discloses a container number recognition compensation system; Figure 1 It is a schematic structural diagram of the container number recognition compensation system of this example.

[0045] The container number recognition compensation system is composed of an individual feature service part and an individual feature library part; The individual feature library part is composed of a key-value database and a vector database, providing persistent services of "individual feature vectors" and "container numbers and other basic container information" of the container for the individual feature service; among them, Key-Value Database: The Redis open-source key-value database is adopted and deployed in a distributed manner; it provides the Key-Value pair storage and matching functions for the individual feature service to record the "vector ID" and "container number and other basic container information" in the vector database.

[0046] Vector Database: The Milvus open-source vector database is adopted and deployed in a distributed manner; it provides the storage and search functions for the "individual feature vectors" of the container in the individual feature service, and the result is to record the "vector ID". The search function adopts the HNSW algorithm + L2 (Euclidean distance) index of the database, and the confidence threshold is adjusted to an appropriate value according to the application scenario, such as taking values between 0.85 and 0.92.

[0047] The individual feature service part consists of an individual feature Web service module, an individual feature extraction service module, a key-value service module, and a vector service module; the key-value service module includes a key-value storage service module and a key-value matching service module, and the vector service module includes a vector storage service module and a vector search service module; among them, Individual Feature Web Service Module: Responds to the public service requests initiated by the external system. The data structure of the data passed in by the external system includes "photos of five container surfaces (left, right, front, back, top)" and "container number and other basic container information", and the container surface photos must be included. If only the container surface photos are passed in, the search process is executed, otherwise the storage process is executed. In the former case, what is fed back to the caller is the "container number and other basic container information" found, or a prompt that nothing is found; in the latter case, what is fed back to the caller is the confirmation that the individual features have been stored.

[0048] Individual Feature Extraction Service Module: Responds to the service requests initiated by the internal modules of the system, and the data passed in is "photos of five container surfaces (left, right, front, back, top)". The SIFT algorithm is used to extract the individual feature points of the container from them, generate the "individual feature vector" of the container and return it to the caller.

[0049] Key-Value Storage Service Module: Responds to the service requests initiated by the internal modules of the system: The data passed in is the "vector ID" and "container number and other basic container information" recorded in the vector database, combined into a Key-Value pair and stored in the key-value database. If there is the same Key value, the old record is overwritten.

[0050] Key-Value Matching Service Module: Responds to the service requests initiated by the internal modules of the system: The data passed in is the "vector ID" recorded in the vector database, used as the search Key value, and the "container number and other basic container information" is matched from the key-value database and returned to the caller.

[0051] Vector storage service module: Respond to service requests initiated by internal modules of the system. The input data is the "individual feature vector" of the container, which is stored in the vector database. The "vector ID" of the vector database record is obtained and returned to the caller.

[0052] Vector search service module: Respond to service requests initiated by internal modules of the system. The input data is the "individual feature vector" of the container, which is used as a query vector to search for the most similar and confident result vector to the query vector from the vector database. The "vector ID" of the vector database record is obtained and returned to the caller, or it is prompted that the search fails.

[0053] Embodiment 2 This embodiment discloses a method for identifying individual containers and a method for storing individuals, which uses the container number identification compensation system of Embodiment 1; Figure 2 It is a schematic flow diagram of the method for identifying individual containers and the method for storing individuals in this embodiment.

[0054] The method for identifying individual containers and the method for storing individuals include the following steps: S1. Provide an individual feature Web service to respond to public service requests initiated by external systems; S1.1. Receive the input data from the external system, which includes "photos of five container surfaces (left, right, front, back, top)" and "container number and other basic container information"; S1.2. The input data from the external system must include photos of five container surfaces. Otherwise, the individual feature Web service module feeds back to the external system, prompting the call to fail; the provided photos of container surfaces are "photos of five container surfaces (left, right, front, back, top)"; S1.3. If the input data from the external system only includes photos of container surfaces, then execute the search process. Figure 3 It is a schematic diagram of the search process of the method for identifying individual containers in this embodiment. The specific search process is as follows: S1.3.1. The individual feature Web service module calls the individual feature extraction service, inputs "photos of five container surfaces (left, right, front, back, top)", and uses the SIFT algorithm to extract individual feature points of the container to obtain the "individual feature vector" of the container; S1.3.2. The individual feature Web service module calls the vector search service, inputs the "individual feature vector" of the container as a query vector, searches for the most similar and confident result vector to the query vector from the vector database, obtains the "vector ID" of the vector database record. If the search fails, the individual feature Web service feeds back to the external system, prompting that the search fails. S1.3.3. The individual feature Web service module calls the key-value matching service, passes in the "vector ID" of the vector database record as the search key value, and matches the "container number and other basic container information" from the key-value database. If the match fails, the individual feature Web service feeds back to the external system, indicating that the search fails. S1.3.4. The individual feature Web service module feeds back the searched "container number and other basic container information" to the external system. S1.4. If the input data from the external system includes "container number and other basic container information" in addition to the "container surface photo", then execute the storage procedure. Figure 4 The following is a schematic diagram of the storage procedure of the container individual storage method of this embodiment. The specific storage procedure is as follows: S1.4.1. The individual feature Web service module calls the individual feature extraction service, passes in the "five container surface photos of the container (left, right, front, back, top)", and uses the SIFT algorithm to extract the individual feature points of the container to obtain the "individual feature vector" of the container. S1.4.2. The individual feature Web service module calls the vector storage service, passes in the "individual feature vector" of the container and stores it in the vector database to obtain the vector database record "vector ID". S1.4.3. The individual feature Web service module calls the key-value storage service, passes in the vector database record "vector ID" and the "container number and other basic container information", combines them into a Kay-Value key-value pair, and stores it in the key-value database. If there is the same key value, the old record is overwritten. S1.4.4. The individual feature Web service module feeds back to the external system to confirm that the individual features have been stored.

[0055] The above steps S1.2, S1.3, and S1.4 are different programs according to the input data, which can play a compensatory role for the OCR container number recognition technology.

[0056] This embodiment can replace the original manual recognition operation, reduce the work burden of the operators; and has high computing performance, second-level execution efficiency, can automatically complete the compensation of OCR container number recognition in real time, improve the operation efficiency of related services, and reduce the operation cost; further, adopting the method of the present invention can significantly improve the overall recognition rate of such containers, reaching between 99.0% and 99.9%.

[0057] Embodiment 3 This embodiment discloses an electronic device, which includes a memory, a processor, and a computer program stored in the memory and configured to run on the processor. When the processor executes the computer program, it implements the container individual identification method and the individual storage method provided in the above-mentioned Embodiment 2. The electronic device is merely an example and should not impose any limitations on the functions and scope of use of the embodiments of the present disclosure.

[0058] The electronic device may be presented in the form of a general-purpose computing device. For example, it may be a server device. The components of the electronic device may include, but are not limited to: the above-mentioned at least one processor, the above-mentioned at least one memory, and a bus connecting different system components (including the memory and the processor).

[0059] The bus includes a data bus, an address bus, and a control bus.

[0060] The memory may include volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read-only memory (ROM).

[0061] The memory may also include program tools (or utilities) having a set (at least one) of program modules. Such program modules include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. The implementation of a network environment may be included in each or some combination of these examples.

[0062] By running the computer program stored in the memory, the processor executes various functional applications and data processing, such as the container individual identification method and the individual storage method provided in the above-mentioned Embodiment 2.

[0063] The electronic device may also communicate with one or more external devices (such as a keyboard, a pointing device, etc.). Such communication may be carried out through an input / output (I / O) interface. Moreover, the electronic device may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter. As shown in the figure, the network adapter communicates with other modules of the electronic device through the bus. It should be understood that other hardware and / or software modules may be used in combination with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (redundant array of independent disks) systems, tape drives, and data backup storage systems, etc.

[0064] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / modules. Conversely, the features and functions of one unit / modules described above can be further divided and embodied by multiple units / modules.

Claims

1. A container number recognition and compensation system, characterized in that: It includes an individual feature library part and an individual feature service part; The individual feature library part includes: A vector database, used to store individual feature vectors of different containers; A key-value database, used to store basic container information corresponding to the individual feature vector of each container, wherein the basic container information includes a container number; The individual feature service part includes: Individual feature Web service module, used to respond to service requests initiated by external systems; The individual feature extraction service module is used to respond to a first service request initiated by the individual feature Web service module; the first service request includes: converting the container surface photo contained in the incoming data of the external system into an individual feature vector to be processed; A vector service module, used to respond to a second service request initiated by the individual feature Web service module, wherein the second service request includes: searching or storing according to the individual feature vector to be processed in combination with the vector database, and obtaining a vector ID; The key-value service module is used to respond to the third service request initiated by the individual feature Web service module, and the third service request includes: matching or storing according to the vector ID in combination with the key-value database, and obtaining the result.

2. The container number recognition and compensation system according to claim 1, characterized in that: The vector service module includes a vector search service module; the key value service module includes a key value matching service module; The vector search service module is used to search the vector database for the most similar and confident result vector to the feature vector to be processed, and obtain the first vector ID recorded in the vector database; The key-value matching service module is used to match the first vector ID from the key-value database to obtain basic container information.

3. The container number recognition and compensation system according to claim 1, characterized in that: The vector service module includes a vector storage service module; the key value service module includes a key value storage service module; The vector storage service module is used to store the feature vector to be processed in the vector database, and obtain the second vector ID recorded in the vector database; The key-value storage service module is used to combine the second vector ID and the basic container information included in the input data of the external system into a Key-Value pair, and store it in the key-value database.

4. The container number recognition and compensation system according to claim 1, characterized in that: The container number identification and compensation system meets one or more of the following conditions: ① The deployment method of the individual feature service part includes distributed deployment or microservice architecture; ② The individual feature Web service module is deployed in the DMZ area of ​​the individual feature service part; ③ The individual feature extraction service module is deployed in the intranet of the individual feature service part; ④ The key-value database includes an open source key-value database; ⑤ The deployment method of the key-value database includes distributed deployment; ⑥ The vector database includes an open source vector database; ⑦ The deployment method of the vector database includes distributed deployment.

5. The container number recognition and compensation system according to claim 4, characterized in that: The container number identification and compensation system meets one or more of the following conditions: ① The key-value database includes a Redis key-value database; ② The vector database includes the Milvus vector database; ③ The individual feature extraction service module includes a SIFT algorithm, and the SIFT algorithm is used to extract individual feature points of the container from the container surface photos contained in the incoming data to generate the individual feature vector to be processed.

6. A container individual identification method, characterized in that: It adopts the container number identification and compensation system as described in any one of claims 1 to 5 to provide individual feature Web services and respond to public service requests initiated by external systems; The container individual identification method comprises the following steps: S1. Receive the incoming data from the external system; and determine the program to be executed according to the incoming data; S2. The container surface photo contained in the incoming data is converted into an individual feature vector to be processed through the individual feature extraction service module, and a search process is performed. The individual feature Web service module calls the vector search service of the vector service module and the key-value matching service of the key-value service module to feed back the searched information to the external system according to the individual feature vector to be processed.

7. The container individual identification method according to claim 6, characterized in that: The container individual identification method meets the following conditions I or II: Condition I: S1. receiving the incoming data from the external system, wherein the incoming data does not include a container surface photo; the individual feature Web service module feeds back to the external system, indicating that the call failed; Condition II: S1, receiving the incoming data from the external system, wherein the incoming data only includes a photo of the container surface, and determining to execute a search process; S2. First, the individual feature Web service module calls the individual feature extraction service, and uses the SIFT algorithm to extract individual feature points from the container surface photo to obtain the individual feature vector to be processed; Secondly, the individual feature Web service module calls the vector search service; uses the individual feature vector to be processed as a query vector, searches the vector database for the most similar and confident result vector to the query vector, and obtains the first vector ID recorded in the vector database; if the search fails, the individual feature Web service module feeds back to the external system, indicating that the search failed; Then, the individual feature Web service module calls the key-value matching service; uses the first vector ID as the search key value, and matches the basic information of the container from the key-value database, wherein the basic information of the container includes the container number; if the match is successful, the individual feature Web service module feeds back the basic information of the container to the external system; if the match fails, the individual feature Web service module feeds back to the external system, indicating that the search was not found.

8. A method for storing individual containers, characterized in that: It adopts the container number identification and compensation system as described in any one of claims 1 to 5 to provide individual feature Web services and respond to public service requests initiated by external systems; The container individual storage method comprises the following steps: S1. Receive the incoming data from the external system; and determine the program to be executed according to the incoming data; S2. The container surface photo contained in the incoming data is converted into an individual feature vector to be processed through the individual feature extraction service module, and a storage process is executed. The individual feature Web service module calls the vector storage service of the vector service module and the key value storage service of the key value service module to combine the individual feature vector to be processed with the basic container information contained in the incoming data into a Key-Value pair, and store them in a key value database. The basic container information includes the container number.

9. The method for storing individual containers according to claim 8, characterized in that: The container individual storage method meets the following conditions I or II: Condition I: S1. receiving the incoming data from the external system, wherein the incoming data does not include a container surface photo; the individual feature Web service module feeds back to the external system, indicating that the call failed; Condition II: S1, receiving the incoming data from the external system, the incoming data including the container surface photo and the basic information of the container, the basic information of the container including the container number, and determining to execute the stored procedure; S2. First, the individual feature Web service module calls the individual feature extraction service, and uses the SIFT algorithm to extract individual feature points from the container surface photo to obtain the individual feature vector to be processed; Secondly, the individual feature Web service module calls the vector storage service; stores the individual feature vector to be processed in the vector database, and obtains the second vector ID recorded in the vector database; Then, the individual feature Web service module calls the key-value storage service; combines the second vector ID and the container basic information into a Key-Value pair, and stores the pair in the key-value database; When the same key value exists in the key-value database, the old record is overwritten; and the individual feature Web service module feeds back to the external system to confirm the individual features of the stored container.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the individual container identification method according to claim 6 or 7 or the individual container storage method according to claim 8 or 9 is implemented.

Images