WO2023124841A1 - Method for storing feature in database, electronic device and computer readable storage medium - Google Patents

Abstract

The invention discloses a method for storing a feature in a database, an electronic device and a storage medium. The method comprises: obtaining a feature to be stored in a database; searching for whether or not the feature to be stored in a database is present in a pre-established buffer queue, and under the condition that the feature to be stored in a database is present in the buffer queue, discarding the feature to be stored in a database; and under the condition that the feature to be stored in a database is not present in the buffer queue, searching for whether or not the feature to be stored in a database is present in a data table; under the condition that the feature to be stored in a database is not present in the data table, storing the feature to be stored in a database into the buffer queue; under the condition that the feature to be stored in a database is present in the data table, discarding the data to be stored in a database; and under the condition that a buffered feature is stored in the buffer queue, storing the feature currently buffered in the buffer queue in a database, so as to cause the currently buffered feature to be stored in the data table.

Description

Feature storage method, electronic device and computer-readable storage medium

related application

This application claims priority to a Chinese patent application with application number 202111681412.2 filed on December 28, 2021, the entire contents of which are incorporated herein by reference.

technical field

The embodiments of the present application relate to the field of image recognition, and in particular, to a feature storage method, electronic equipment, and a computer-readable storage medium.

Background technique

In recent years, with the country's strong support for Artificial Intelligence (AI), the development of AI technology in the security field has also been increasing. Government business. In recent years, the number of police and civilian camera installations has increased dramatically, resulting in massive video data. However, these data are useless data to the data owner. Using the AI analysis system to intelligently analyze video data can greatly reduce the consumption of human resources and improve the response speed to public security incidents.

Face features are the focus of video surveillance systems, and the storage of stranger face features is an important requirement in the security field. The pedestrian trajectory can be generated through the processing of the AI analysis system, and the target object can be found through the pedestrian trajectory and a corresponding alarm can be generated. Then remind the relevant personnel to deal with it in time, which greatly improves the efficiency of using video data to process government affairs. In this process, feature warehousing is particularly important. However, the current implementation method used to solve the problem of repeated warehousing of feature warehousing still has the problem of low efficiency of feature warehousing.

Contents of the invention

The main purpose of the embodiments of the present application is to provide a feature storage method, an electronic device, and a computer-readable storage medium, so that the efficiency of feature storage can be improved while avoiding repeated feature storage.

In order to at least achieve the above purpose, the embodiment of the present application provides a feature storage method, including: obtaining the features to be stored; searching in the pre-established cache queue to see if the features to be stored exist, and when the retrieved In the case where the feature to be stored in the cache queue exists, the feature to be stored in the library is discarded; In the data table, search whether the feature to be stored exists; if the feature to be stored does not exist in the data table, store the feature to be stored in the cache queue In the case where the feature to be stored in the data table is retrieved, discard the feature to be stored in the database; if there is a cached feature in the cache queue, put the cache queue The features currently cached in the database are stored in the database, so that the features currently cached are stored in the data table.

In order to at least achieve the above purpose, an embodiment of the present application further provides an electronic device, including: at least one processor; and a memory connected to the at least one processor in communication; wherein, the memory stores information that can be used by the Instructions executed by at least one processor, the instructions are executed by the at least one processor, so that the at least one processor can execute the above-mentioned feature storage method.

In order to at least achieve the above purpose, the embodiment of the present application also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned feature storage method is implemented.

The feature warehousing method provided by the embodiment of the present application obtains the features to be warehousing, searches whether there are features to be warehousing in the pre-established cache queue, and in the case that there are features to be warehoused in the cache queue. , discard the features to be stored, and if there is no feature to be stored in the cache queue, search whether there is a feature to be stored in the pre-established data table; if there is no feature to be stored in the retrieved data table In the case of the characteristics of storage, store the characteristics to be stored in the cache queue; store the characteristics of the current cache in the cache queue, so that the characteristics of the current cache can enter the data table; that is, the implementation of the present application The example pre-establishes the cache queue to solve the problem of the speed mismatch between the production data and the consumption data. When it is necessary to store features, first search in the cache queue. If there is no feature to be stored in the cache queue, check whether there is a feature to be stored in the data table, so as to avoid targeting all features to be stored. The characteristics of the library must be retrieved in the data table. Since the retrieval in the cache queue is a retrieval in memory, it is much faster than directly searching in the database, that is, the speed of memory retrieval is greater than the speed of database retrieval, so it will greatly improve the retrieval speed. efficiency. At the same time, according to the law of data appearance, the same feature will generally appear in a short period of time, so the probability of being hit by only searching the cache queue is very high, so the retrieval efficiency can be further improved to further improve the storage efficiency. In addition, the use of the cache queue can also reduce the time required to lock and unlock the library when each feature is stored, thereby further improving the efficiency of feature storage.

Description of drawings

Fig. 1 is a schematic flow chart of a feature storage method mentioned in the embodiment of the present application;

Fig. 2 is the flow chart that the length of buffer queue mentioned in the embodiment of the present application increases;

Fig. 3 is the flow chart of shortening the length of the cache queue mentioned in the embodiment of the present application;

Fig. 4 is a schematic flow chart of another feature storage method mentioned in the embodiment of the present application;

Fig. 5 is a schematic structural diagram of the electronic device mentioned in the embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. However, those of ordinary skill in the art can understand that in each embodiment of the application, many technical details are provided for readers to better understand the application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can also be realized. The division of the following embodiments is for the convenience of description, and should not constitute any limitation to the specific implementation of the present application, and the embodiments can be combined and referred to each other on the premise of no contradiction.

The inventor of the present application found that, in order to solve the problem of repeated warehousing of features, there are mainly two methods for warehousing facial features, which are introduced below:

Method 1: By configuring the storage time interval of a single face feature. The specific implementation method is: after the system deployment is completed, collect the time when the face features are stored within a period of time (such as 1 week) as the basic data, and take the maximum value of the basic data plus 20% of the maximum value as a single person in the entire system The unit storage time of face feature storage. That is to say, only one feature is allowed to be stored in this unit of time. Even if the storage is completed in advance, the next feature storage operation cannot be performed.

Method 2: Enter the database by locking the table. The specific implementation method is: the video source generates a face picture, and the algorithm model extracts features from the face picture. Each thread that gets the facial features judges whether it can perform storage operations (that is, judges whether the gallery is locked). If the storage operation can be done (that is, the library is not locked), first lock the library operation, and then put the acquired face features into the library (before this thread does not perform the unlock operation, other threads cannot delete the library. Operations other than read operations), after the storage is completed, perform the library unlock operation to release the operation right of the library. Among them, the gallery is the data table in the database.

However, the problems of the above-mentioned method 1 include: low time utilization rate, the configured time interval is no longer applicable when the gallery data changes too much, there will be too much empty waiting time or there is still a phenomenon of repeated storage, which makes the feature input Library efficiency is low. The problems in the above method two include: frequently operating the library, wasting a certain amount of time, and making the feature storage efficiency low.

In the embodiment of the present application, in order to at least solve the technical problem of low efficiency of feature warehousing in the case of avoiding repeated warehousing of features into warehousing, a feature warehousing method is provided, which is applied to an electronic device, and the electronic device may be a server. The application scenarios of the embodiments of the present application may include but not limited to: stream data collection of facial (or humanoid) features into storage, such as suspect chase and escape tracking, generation of pedestrian trajectories, and feature storage in scenarios such as one person with one profile. The schematic flow diagram of the feature storage method in the embodiment of the present application can refer to Figure 1, including:

Step 101: Obtain the features to be stored;

Step 102: Search in the pre-established cache queue whether there are features to be stored in the warehouse, if yes, execute step 103, otherwise execute step 104;

Step 103: discarding the features to be stored;

Step 104: Search in the pre-established data table whether there is a feature to be stored; if yes, execute step 103, otherwise execute step 105;

Step 105: Store the features to be stored in the cache queue;

Step 106: judge whether the cache queue is empty; if yes, continue to execute step 106, otherwise execute step 107;

Step 107: Store the currently cached features in the cache queue into the database, so that the currently cached features are stored in the data table.

In the embodiment of the present application, the cache queue is pre-established to solve the problem of the speed mismatch between the production data and the consumption data. When it is necessary to store features, first search in the cache queue. If there is no feature to be stored in the cache queue, check whether there is a feature to be stored in the data table, so as to avoid targeting all features to be stored. The characteristics of the library must be retrieved in the data table. Since the retrieval in the cache queue is a retrieval in memory, it is much faster than directly searching in the database, that is, the speed of memory retrieval is greater than the speed of database retrieval, so it will greatly improve the retrieval speed. efficiency. At the same time, according to the law of data appearance, the same feature will generally appear in a short period of time, so the probability of being hit by only searching the cache queue is very high, so the retrieval efficiency can be further improved to further improve the storage efficiency. In addition, the use of the cache queue can also reduce the time required to lock and unlock the library when each feature is stored, thereby further improving the efficiency of feature storage.

In step 101, the features to be stored are reference features that are expected to be stored in the data table for subsequent feature matching, and may be: picture features, sound features, and the like. The feature to be stored is the unique representation of the collected data calculated by the algorithm.

For example, in a face recognition scenario, the features to be stored can be face features (ie features of face pictures), and the face features stored in the data table are used for subsequent face feature matching. In the fingerprint recognition scenario, the features to be stored may be fingerprint features (ie features of fingerprint pictures), and the fingerprint features stored in the data table are used for subsequent fingerprint feature matching. In the voiceprint recognition scenario, the features to be stored may be voiceprint features, and the voiceprint features stored in the data table are used for subsequent voiceprint feature matching. It can be understood that, when applied to different scenarios, the features to be loaded into the library may be set according to actual needs, which is not specifically limited in this embodiment.

In one embodiment, the features to be stored are extracted from the video stream or picture stream provided by the camera, that is, the server can obtain the video stream or picture stream taken by the camera, and extract the picture features in the video stream or picture stream; wherein , the camera may be a security camera. For the video stream or picture stream provided by the camera, the picture features of the same person are generally generated in one time period, which greatly increases the possibility of the picture features of the same person being hit in the cache queue, that is, in It is easy to quickly detect whether the feature to be stored in the cache queue is stored in the cache queue within a period of time, so that a judgment can be quickly made on whether to discard the feature to be stored, so as to further improve the retrieval efficiency and storage efficiency.

In step 102, the pre-established cache queue is used to cache the features to be stored, which is equivalent to data producers (features to be stored) and data consumers (features to be Storage) transmission buffer channel between. When the production speed of the data producer is greater than the consumption speed of the data consumer, it is difficult to achieve the synchronization of feature production and feature consumption. Cache into the cache queue, and then gradually store the features cached in the cache queue. The store operation can be understood as transferring the features cached in the cache queue to the data table in the database. The feature enqueue and feature dequeue in the cache queue may be performed at the same time, which can also improve the efficiency of feature warehousing to a certain extent.

The server can search whether there is a feature that is the same as the feature to be stored in the currently cached features in the cache queue. If there is a feature that is the same as the feature to be stored in the cache queue, it means that the cache queue has already stored the feature to be stored In order to avoid repeated storage, the server will perform step 103, that is, discard the feature to be stored; if there is no feature identical to the feature to be stored in the cache queue, it means that the cache queue has not yet stored the feature to be stored. If the feature to be stored, step 104 can be executed to continue searching in the data table whether the feature to be stored exists.

In step 104, the data table is a data table in a database where the features to be stored may be stored. For example, the database includes: a gallery, a sound library, etc., and the gallery and the sound library can be two data tables stored in the database respectively. .

The server can search whether there is the same feature as the feature to be stored in the feature stored in the data table. If there is a feature that is the same as the feature to be stored in the data table, it means that the feature to be stored has already been stored in the data table. In order to avoid repeated storage, the server will execute step 103, that is, discard the feature to be stored; if the same feature as the feature to be stored does not exist in the data table, it means that the feature to be stored has not been stored in the data table. For the features to be stored, step 105 can be performed to store the features to be stored in the cache queue.

In step 106, the server judges whether the cache queue is empty, that is, whether there is a cache feature in the cache queue. When the cache queue of the server is empty, that is, there is no cache characteristic in the cache queue, continue to execute step 106 . When the cache queue of the server is not empty, that is, there is a cache feature in the cache queue, go to step 107 .

In step 107, the server performs a warehouse-in operation on the characteristics of the current cache in the cache queue, so that the characteristics of the current cache in the cache queue are stored in the data table; wherein, the warehouse-in operation can also be understood as a dequeue operation. If it is empty, the features cached in the cache queue are dequeued in the order of first-in-first-out, and stored in the data table in turn. That is to say, if the cache queue is not empty, features can be continuously fetched from the cache queue and put into storage. Wherein, since the data table is stored in the database, storing in the data table can be understood as storing in the database.

In this embodiment, when the service is started, a cache queue is established in the memory, and when features need to be stored in the database, they are first retrieved in the cache queue. Since the retrieval of the cache queue is in-memory retrieval, it is faster than directly searching the database Many, so it will greatly improve the retrieval efficiency of features. If the same feature is retrieved, the feature warehousing process will be terminated. Otherwise, the data table will be retrieved. If the same feature is retrieved, the warehousing process will be terminated. Otherwise, the feature will be stored in the data table. Avoid storing identical features in the database. Using the cache queue, compared with the method of locking the table into the database, in this embodiment, it does not need to spend the time of locking and unlocking the table, which greatly improves the efficiency of feature storage. At the same time, according to the law of data appearance, the same feature will generally appear in a short period of time, so the probability of being hit by only searching the cache queue is very high, which can further improve the retrieval efficiency and storage efficiency.

In one embodiment, the retrieval process (that is, the process of retrieving whether there is a feature to be stored in the cache queue and the data table) and the storage process (that is, the process of entering the feature cached in the cache queue into the data table) are two parallel The processes are respectively implemented by two independent threads, that is, steps 106-107 and steps 101-105 are two parallel processes, so that the efficiency of feature storage can be greatly improved.

In one embodiment, the cache queue has a preset initial length, and the feature storage method further includes: when the cache length occupied by the currently cached feature in the cache queue is greater than the first preset length, the queue of the cache queue The length is increased to a second preset length; wherein, the first preset length is smaller than the initial length. The first preset length and the second preset length can be set according to actual needs. It can be understood that since the queue length of the cache queue is increased to the second preset length, the second preset length is greater than the initial length . In the case that the cache length occupied by the currently cached features in the cache queue is greater than the first preset length, since the first preset length is smaller than the initial length of the cache queue, it means that the cache queue is about to be full at this time but not yet full, The speed at which features are stored in the cache queue (which can be understood as the production speed of feature producers) is greater than the speed at which features are stored in storage (which can be understood as the consumption speed of feature consumers), so that there is a large amount of data in the cache queue that has not been stored in the warehouse. The queue length of the cache queue is increased to the second preset length, that is, the length of the cache queue is expanded so that it can adapt to the scenario where the production speed of the feature producer is greater than the consumption speed of the feature consumer, avoiding the The problem of feature loss caused by the mismatch between the speed and the consumption speed of the feature consumer.

In one embodiment, the first preset length is greater than 0.5n and less than n, the second preset length is greater than n and less than or equal to 1.5n, and n is the initial length. The first preset length and the second preset length are denoted by T1 and T2 respectively, and the relationship satisfied by T1 and T2 is respectively: 0.5n<T1<n; n<T2≤1.5n. 0.5n<T1<n can better measure the length state of the cache queue that is about to be full but not yet full, and n<T2≤1.5n can better ensure that the expanded cache queue can store more waiting items The features of the library are well adapted to the scenario where the production speed of feature producers is greater than the consumption speed of feature consumers. At the same time, T2≤1.5n can also avoid excessive expansion of the length of the cache queue and waste of memory resources.

In one embodiment, the first preset length is 0.8n, and the second preset length is 1.2n. That is, when the cache length occupied by the currently cached features in the cache queue is greater than 0.8n, the queue length of the cache queue is increased to 1.2n. 0.8n can well represent the state that is about to be full but not yet full. At this time, the queue length of the cache queue is increased to 1.2n, which is equivalent to providing 0.4n for the data to be stored in the cache queue next. storage space. By properly expanding the length of the cache queue at the right time, it is beneficial to make reasonable use of resources while avoiding the problem of repeated storage of the same feature due to the mismatch between the production speed of the feature producer and the consumption speed of the feature consumer. question.

In one embodiment, the flow chart of increasing the length of the buffer queue involved in the feature storage method can refer to FIG. 2, including:

Step 201: Determine whether the cache length occupied by the currently cached features in the cache queue is greater than 0.8n, if yes, execute step 202, otherwise the process ends;

Step 202: Increase the queue length of the cache queue to 1.2n.

In one embodiment, the feature warehousing method further includes: when it is detected k consecutive times within the preset time period that the length occupied by the currently cached feature in the cache queue is less than the third preset length, storing the cache queue The length of the queue is reduced to a fourth preset length; wherein, the third preset length is smaller than the first preset length, and the k is a natural number greater than 1. The third preset length and the fourth preset length can be set according to actual needs. It can be understood that, since the queue length of the cache queue is reduced to the fourth preset length, the fourth preset length can be smaller than the initial length. The preset time period can be set according to actual needs, which is not specifically limited in this embodiment. If it is detected that the length occupied by the currently cached feature in the cache queue is less than the third preset length for k consecutive times within the preset period of time, and because the third preset length is less than the first preset length, it means that the buffer is cached for a period of time The amount of data cached in the queue continues to be small. At this time, reducing the queue length of the cache queue to a fourth preset length is beneficial to avoid waste of resources while meeting the cache requirements of the current cache queue. Especially in a multi-process scenario, it can effectively save resources and maximize resource utilization.

In one embodiment, the third preset length is greater than 0 and less than or equal to 0.5n, and the fourth preset length is greater than 0.5n and less than n. The third preset length and the fourth preset length are denoted by T3 and T4 respectively, and the relationship satisfied by T3 and T4 is respectively: 0<T3≤0.5n; 0.5n<T4<n. The third preset length is greater than 0 and less than or equal to 0.5n, which can better measure the state of less data stored in the cache queue. The fourth preset length is greater than 0.5n and less than n, which can better ensure that the reduced length The cache queue can also meet the current cache requirements of the cache queue, and is also beneficial to avoid waste of resources.

In one embodiment, the third preset length is 0.5n, and the fourth preset length is 0.6n. That is, when the length occupied by the currently cached features in the cache queue is detected k times in a preset period of time is less than 0.5n, the queue length of the cache queue is reduced to 0.6n, and if it is less than 0.5n, it is less than the original length Half, it can well represent the state of the cache queue with less long-term cache. At this time, the queue length of the cache queue is reduced to 0.6n, which is equivalent to providing at least 0.1n for the data to be stored in the cache queue next. storage space. By properly shortening the length of the cache queue at the right time, it is beneficial to make reasonable use of resources while avoiding the problem of repeated storage of the same feature due to the mismatch between the production speed of the feature producer and the consumption speed of the feature consumer. question.

In one embodiment, the flow chart of shortening the length of the cache queue involved in the method for storing features can refer to FIG. 3 , including:

Step 301: Determine whether the length occupied by the currently cached feature in the cache queue is less than 0.5n; if yes, execute step 302, otherwise the process ends;

Step 302: accumulating the number of times that the length of the currently cached feature in the cache queue detected continuously within a preset time period is less than 0.5n;

Step 303: judge whether the accumulated number of times is greater than k; if yes, execute step 304, otherwise execute step 305;

Step 304: reducing the queue length of the cache queue to 0.6n;

Step 305: Set the accumulated times to 0.

In one embodiment, in the flow chart of feature storage similar to that shown in FIG. 1 , the process of increasing the length of the cache queue may be included, such as the process similar to that in FIG. 2 , and the process of shortening the length of the cache queue may also be included. For example, similar to the process in FIG. 3 , it may also include the process of increasing and shortening the length of the cache queue at the same time. That is, in the embodiment of the present application, a dynamically adjustable cache queue is used, and the length of the queue can be adaptively changed as the production speed of feature data changes, which effectively balances the space complexity and storage efficiency.

In one embodiment, the schematic flow chart of the feature storage method can refer to Figure 4, including:

Step 401: Initialize a cache queue q with a length of n;

Step 402: Obtain the picture and extract the feature f of the picture; for example, to obtain the picture collected by the security camera, the feature f is the feature to be stored;

Step 403: Determine the relationship between the cache length m and n occupied by the features in the current cache queue;

Step 404: If m>0.8n, modify the length of the cache queue to 1.2n;

Step 405: If m<0.5n is detected for k consecutive times within a period of time, modify the length of the cache queue to 0.6n;

Step 406: If m≤0.8n or if m<0.5n is not detected for k consecutive times within a period of time, then do not change the cache queue length n;

Step 407: Search whether the feature f exists in the cache queue q, if it exists, execute step 408, otherwise execute step 409;

Step 408: Discard the feature f;

Step 409: Search whether the feature f exists in the data table, if it exists, execute step 408, otherwise execute step 410;

Step 410: Insert feature f into cache queue q;

Repeat the steps from step 401 to step 410 above;

Step 411: If the cache queue q is not empty, continuously extract features from q and put them into the library.

It should be noted that although step 411 is written as the last step in FIG. 4 , it does not mean that step 411 is not the last step executed, that is, step 411 can be executed as long as there is a cached feature in the cache queue q.

In this embodiment, since the retrieval of the cache queue is a retrieval in memory, the retrieval efficiency of the same feature will be greatly improved. Moreover, the length of the cache queue can be dynamically changed (it can be shortened or expanded), which fundamentally solves the problem that the production speed of the feature producer is inconsistent with the consumption speed of the feature consumer, that is, it does not match. Moreover, for the video stream or picture stream provided by the security camera, the picture features of the same person are generally generated in one time period, which greatly increases the possibility of the picture features of the same person being hit in the cache queue. Thus, the retrieval efficiency is further improved. In addition, in this embodiment, the two processes of retrieval and storage can be performed concurrently, which is beneficial to further improve storage efficiency.

It should be noted that the above examples in the embodiments of the present application are all illustrations for the convenience of understanding, and do not limit the technical solution of the present invention.

The step division of the above various methods is only for the sake of clarity of description. During implementation, it can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. ; Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but not changing the core design of the algorithm and process are all within the scope of protection of this patent.

The embodiment of the present application also provides an electronic device, as shown in FIG. 5 , including: at least one processor 501; and a memory 502 communicatively connected to at least one processor 501; The instructions executed by the processor 501 are executed by at least one processor 501, so that the at least one processor 501 can execute the method for storing features in the foregoing embodiments.

Wherein, the memory 502 and the processor 501 are connected by a bus, and the bus may include any number of interconnected buses and bridges, and the bus connects one or more processors 501 and various circuits of the memory 502 together. The bus may also connect together various other circuits such as peripherals, voltage regulators, and power management circuits, all of which are well known in the art and therefore will not be further described herein. The bus interface provides an interface between the bus and the transceivers. A transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing means for communicating with various other devices over a transmission medium. The data processed by the processor 501 is transmitted on the wireless medium through the antenna, and further, the antenna also receives the data and transmits the data to the processor 501 .

Processor 501 is responsible for managing the bus and general processing, and may also provide various functions including timing, peripheral interface, voltage regulation, power management and other control functions. And the memory 502 may be used to store data used by the processor 501 when performing operations.

The embodiment of the present application also provides a computer-readable storage medium storing a computer program. The above method embodiments are implemented when the computer program is executed by the processor.

That is, those skilled in the art can understand that all or part of the steps in the method of the above-mentioned embodiments can be completed by instructing related hardware through a program, the program is stored in a storage medium, and includes several instructions to make a device ( It may be a single-chip microcomputer, a chip, etc.) or a processor (processor) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. .

Those of ordinary skill in the art can understand that the above-mentioned embodiments are specific examples for realizing the present invention, and in practical applications, various changes can be made to it in form and details without departing from the spirit and spirit of the present invention. scope.

Claims (10)

A feature storage method, including: Obtain the characteristics to be stored; Retrieve whether the feature to be stored in the pre-established cache queue exists, and discard the feature to be stored in the case that the feature to be stored in the cache queue is retrieved; When it is retrieved that the feature to be stored does not exist in the cache queue, search whether the feature to be stored exists in the pre-established data table; When it is retrieved that the feature to be stored does not exist in the data table, store the feature to be stored in the cache queue; In the case that the feature to be stored in the data table is retrieved, discarding the feature to be stored in the database; If there are cached features in the cache queue, store the currently cached features in the cache queue, so that the currently cached features are stored in the data table. The feature warehousing method according to claim 1, wherein the cache queue has a preset initial length, and the method further comprises: In the case that the cache length occupied by the currently cached feature in the cache queue is greater than the first preset length, increase the queue length of the cache queue to a second preset length; wherein, the first preset The length is less than the initial length. The feature warehousing method according to claim 2, wherein the first preset length is greater than 0.5n and less than n, the second preset length is greater than n and less than or equal to 1.5n, and the n is the initial length. The feature storage method according to claim 3, wherein the first preset length is 0.8n, and the second preset length is 1.2n. The feature storage method according to any one of claims 1 to 4, wherein the method further comprises: When it is detected k times in a preset time period that the length occupied by the currently cached feature in the cache queue is less than the third preset length, the queue length of the cache queue is reduced to a fourth preset length; wherein, The third preset length is smaller than the first preset length, and the k is a natural number greater than 1. The feature storage method according to claim 5, wherein the third preset length is greater than 0 and less than or equal to 0.5n, and the fourth preset length is greater than 0.5n and less than n. The feature storage method according to claim 6, wherein the third preset length is 0.5n, and the fourth preset length is 0.6n. The feature storage method according to any one of claims 1 to 4, wherein the features to be stored are extracted from video streams or picture streams provided by cameras. An electronic device, comprising: at least one processor; and, a memory communicatively coupled to the at least one processor; wherein, The memory is stored with instructions executable by the at least one processor, the instructions are executed by the at least one processor, so that the at least one processor can perform any one of claims 1 to 8 The feature storage method. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the feature storage method described in any one of claims 1 to 8 is implemented.

Images