Method and system for realizing IPC video object classification search based on AI large model

Abstract

The invention belongs to the technical field of artificial intelligence and smart home, and particularly relates to a method and system for realizing IPC video object classification search based on an AI large model, and the method comprises the steps: collecting video images at regular time; based on the collected video images, adopting an AI large model to identify and detect all the selected objects, obtaining tuple data information of all the selected objects, and performing data storage and archiving; and based on a user request, searching all video clips of the target object in a specified time period in the archived data by adopting a funnel type layer-by-layer positioning method, and splicing to form a video clip set of the target object in the specified time period. According to the method and the device, the user can quickly search and locate the video clip of the selected target object in the massive videos, so that the search efficiency is improved, the inaccuracy of manual search of the user is avoided, and the user experience of the IPC in home security and protection is improved.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence and smart home technology, and specifically relates to a method and system for implementing IPC video object classification and search based on an AI big model. Background Art

[0002] With the development of deep learning and computer vision technology, image recognition and classification have made great progress. Using artificial intelligence (AI) technology, we can accurately identify and classify objects in images. This ability has been widely used in many fields such as autonomous driving, medical image analysis, and security monitoring.

[0003] As an important part of computer vision, object detection algorithms are used to identify target objects in images. The development of this field has experienced a transition from traditional manual feature detection algorithms to object detection algorithms based on AI big models. AI big models are usually based on deep learning algorithms and are trained through large-scale training data sets to build complex models with excellent generalization performance and powerful expression capabilities.

[0004] In the smart home scenario, for the application scenarios of IPC (IP Camera, network camera), currently most of the methods are manually used to find the target object video clips in a large amount of recorded videos, which is complicated, tedious and inefficient. Summary of the invention

[0005] The present invention provides an IPC video object classification search method and system based on an AI big model. By performing timed acquisition of video images, the AI ​​big model is used to identify and detect all target object sub-features, obtain tuple data information of the target object, and then a funnel-type layer-by-layer positioning search is used for all video clips of the target object within a specified time, so that users can quickly find and locate the video clips of the selected target object in massive videos, thereby improving the search efficiency, avoiding the inaccuracy of manual search by users, and improving the user experience of IPC in home security.

[0006] A method for classifying and searching IPC video objects based on an AI big model, comprising:

[0007] Collect video images at regular intervals;

[0008] Based on the collected video images, the AI ​​big model is used to identify and detect all selected objects, obtain the tuple data information of all selected objects, and store and archive the data;

[0009] Based on user requests, a funnel-type layer-by-layer positioning method is used in the archived data to search for all video clips of the target object within a specified period of time, and then the video clips of the target object within the specified period of time are spliced ​​together to form a set.

[0010] By collecting video images at regular intervals, using a large AI model to identify and detect all sub-features of the target object, obtaining the tuple data information of the target object, and then using a funnel-type layer-by-layer positioning to search for all video clips of the target object within a specified time, users can quickly find and locate the video clips of the selected target object in massive videos, thereby improving the search efficiency, avoiding the inaccuracy of manual search by users, and improving the user experience of IPC in home security.

[0011] Furthermore, based on the collected video images, the AI ​​big model is used to identify and detect all selected objects, obtain tuple data information of all selected objects, and store and archive the data, including:

[0012] Based on the collected video images, use them as input for the AI ​​big model;

[0013] Use the AI ​​big model to identify and detect all selected objects, and output object tuple information; the object tuple information includes timestamp, identification object category, identification object feature vector, identification object category and identification object gender;

[0014] Encode and identify object tuple information, and perform data storage and archiving.

[0015] Furthermore, encoding and identifying the object tuple information and performing data storage and archiving include:

[0016] Establish a unique identification code for the object tuple, output the first tuple data, and store and archive it in the memory; the first tuple data includes a timestamp, an identification object category, and an identification object tuple unique identification code;

[0017] The object tuple is Base64 encoded to obtain the encoded string, and the second tuple data is output in combination with the unique identification code of the object tuple, and stored and archived in the database; the second tuple data includes the unique identification code for identifying the object tuple and the Base64 encoding string for identifying the object tuple.

[0018] Furthermore, based on the user's request, the funnel-type layer-by-layer positioning method is used in the archived data to search for all video clips of the target object within the specified time period, and the video clips set within the specified time period of the target object are spliced ​​together, including:

[0019] Constructing a user request; the user request features include a time sub-feature and a target object sub-feature;

[0020] Based on the user request, the archived data is traversed, and the first-layer timestamp is found and located based on the time sub-feature, and the corresponding layer timestamps are found and located in the first-layer timestamp in the order of the target object sub-features, until the corresponding final layer timestamp is obtained according to the last target object sub-feature, forming a tuple structure object data set;

[0021] Based on the tuple structure object data set and the association relationship between adjacent timestamps, the target object video clip is searched.

[0022] Furthermore, the user request form includes:

[0023] Parameter form, which includes a start timestamp, an end timestamp, a category of the identified object, and a gender of the identified object;

[0024] The image data format includes a start timestamp, an end timestamp, and image data of the object to be identified.

[0025] Further, based on the user request, the archived data is traversed, the first layer timestamp is searched and located based on the time sub-feature, and the corresponding layer timestamps are searched and located in the first layer timestamp in sequence according to the sub-features of the target object, until the corresponding final layer timestamp is obtained according to the last sub-feature of the target object, forming a tuple structure object data set, including:

[0026] Based on user requests, the archived data is traversed, sorted in ascending order based on the timestamp sequence, and the minimum and maximum timestamps corresponding to the time sub-features are searched using the binary search method to obtain the first-level timestamp set for the specified time period;

[0027] Based on the first-layer timestamp set, the timestamps corresponding to the multiple target object sub-features are searched and located in sequence to obtain the timestamp sets of each layer where the target object sub-features appear within the specified time period;

[0028] Until the corresponding timestamp is obtained according to the last target object sub-feature, the final layer timestamp set of all target object sub-features appearing in the specified time period is obtained, and a tuple structure object data set is formed.

[0029] Furthermore, based on the tuple structure object data set and in combination with the association relationship of adjacent timestamps, the search for the target object video clip is completed, including:

[0030] Based on the tuple structure object data set, determine the association relationship between adjacent timestamps;

[0031] When the current timestamp and the next timestamp satisfy an adjacent relationship, the video segments corresponding to the two video stamps are merged;

[0032] When the current timestamp and the next timestamp do not satisfy the adjacent relationship, the current timestamp is used as the midpoint to intercept a video segment within a collection frequency forward and backward respectively;

[0033] Based on all selected video clips, the search for the target object video clip is completed.

[0034] A system for implementing an IPC video object classification search method based on an AI big model, comprising:

[0035] An image acquisition module, which is used to regularly acquire video images;

[0036] The target object recognition module is used to identify and detect all selected objects based on the collected video images using the AI ​​large model to obtain tuple data information of all selected objects;

[0037] A data storage module, which is used to store and archive the acquired tuple data information of all selected objects;

[0038] The data processing module is used to search for all video clips of the target object within a specified period of time in the archived data based on user requests by using a funnel-type layer-by-layer positioning method, and to splice together a set of video clips of the target object within the specified period of time.

[0039] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the above method when executing the computer program.

[0040] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the method described above is implemented.

[0041] The beneficial effects of the present invention are:

[0042] The present invention collects video images at regular intervals, uses an AI large model to identify and detect all target object sub-features, obtains tuple data information of the target object, and then uses a funnel-type layer-by-layer positioning to search for all video clips of the target object within a specified time, so that users can quickly find and locate the video clips of the selected target object in massive videos, thereby improving the search efficiency, avoiding the inaccuracy of manual search by users, and enhancing the user experience of IPC in home security. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the present invention;

[0044] Figure 2 It is a structural diagram of the funnel search and positioning method;

[0045] Figure 3 It is a structural schematic diagram of the system in the present invention;

[0046] Figure 4 A schematic diagram of the structure of a computer device. 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] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0049] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples, and a person of ordinary skill in the art can understand the specific meanings of the above terms in this application under specific circumstances.

[0050] Example 1

[0051] Figure 1 The method is a method for classifying and searching IPC video objects based on an AI big model. By periodically collecting video images, the AI ​​big model is used to identify and detect all sub-features of the target object, and the tuple data information of the target object is obtained. Then, a funnel-type method is used to locate and search for all video clips of the target object within a specified time. This allows users to quickly locate the video clips of the selected target object in massive videos, thereby improving the search efficiency and avoiding the inaccuracy of manual searches by users, thus improving the user experience of IPC in home security. The specific steps are as follows:

[0052] S1: Timed acquisition of video images;

[0053] The video image timing acquisition granularity does not exceed 60 seconds. It should be noted that the acquisition granularity should be set appropriately. Too large an acquisition granularity will result in too low a search accuracy, and too small an acquisition granularity will result in too much resource consumption. In this embodiment, the acquisition granularity is set to t=30s.

[0054] S2: Based on the collected video images, use the AI ​​big model to identify and detect all selected objects, obtain the tuple data information of all selected objects, and store and archive the data;

[0055] S21: Based on the collected video images, use them as input for the AI ​​big model;

[0056] S22: Use the AI ​​big model to identify and detect all selected objects, and output object tuple information; the object tuple information includes timestamp, identification object category, identification object feature vector, identification object category and identification object gender;

[0057] Specifically, when the AI ​​large model fails to recognize the selected target object, it waits for the next recognition request; when the AI ​​large model recognizes the selected target object, it outputs the object tuple information.

[0058] In this embodiment, the object tuple information is represented in the form of: (timestamp, identification object category, identification object feature vector, [identification object gender], [identification object age range]). Among them, the timestamp represents the timestamp when the video image is captured; the identification object category represents the identification object type, which includes PERSON (person), CAT (cat), DOG (dog); the identification object feature vector represents the feature vector value of the identification object; the identification object gender represents the gender of the identification object when the category is person, which includes F (female) and M (male); the identification object age range represents the age range of the identification object when the category is person; [] represents the target object sub-feature.

[0059] S23: Encode and identify the object tuple information, and perform data storage and archiving.

[0060] S231: Create a unique identification code UUID for the object tuple, output the first tuple data, and store and archive it in the memory; the first tuple data includes a timestamp, an identification object category, and an identification object tuple unique identification code UUID;

[0061] The first tuple data is represented in the form of (timestamp, identification category, identification object tuple unique identifier UUID), and is defined as:

[0062] struct tuple{

[0063] long timstamp; / / timestamp

[0064] string type; / / Identify the object category

[0065] string uuid; / / Unique identification code of the object tuple

[0066] }

[0067] Among them, when the first tuple data is stored in the memory, the data structure uses HashMap<key,value> , key represents the timestamp, value represents the tuple structure object, and value includes multiple sub-features of the target object.

[0068] S232: Perform Base64 encoding on the object tuple to obtain the encoded string S, combine it with the unique identification code UUID of the corresponding tuple to output the second tuple data, and store and archive it in the database; the second tuple data includes the unique identification code UUID of the identification object tuple and the Base64 encoding string S of the identification object tuple.

[0069] The second tuple data is represented in the form of: (object tuple unique identifier UUID, object tuple Base64 encoding string S). The second tuple data is stored in the database and the data table is located as:

[0070] create table tuple_schema(

[0071] uuid long,

[0072] b64data text );

[0074] S3: Based on the user's request, a funnel-type layer-by-layer positioning method is used in the archived data to search for all video clips of the target object within the specified time period, and the video clips within the specified time period of the target object are spliced ​​together to form a set.

[0075] S31: construct user request;

[0076] The user request feature includes a time sub-feature and a target object sub-feature. The time sub-feature includes a start timestamp and an end timestamp; the target object sub-feature includes an identification object category and an identification object gender.

[0077] It should be noted that the target sub-object features include not only the identification object category and the identification object gender, which can be one or more, and are adjusted according to the actual needs of users in practical applications.

[0078] Among them, user request forms include:

[0079] The parameter form includes a start timestamp, an end timestamp, a category of the identified object, and a gender of the identified object; the representation form is: (start timestamp, end timestamp, category of the identified object, gender of the identified object).

[0080] The picture data format includes a start timestamp, an end timestamp, and image data of the target object to be found; the representation format is: (start timestamp, end timestamp, Base64enc (picture data of the target object)), where Base64enc (picture data of the target object) represents a Base64-encoded string of binary data of the target object picture data.

[0081] S32: based on the user request, traverse the archived data, locate and obtain the first-layer timestamp based on the time sub-feature, and locate and obtain the corresponding layer timestamp in the first-layer timestamp in accordance with the target object sub-feature sequence, until the corresponding final layer timestamp is obtained according to the last target object sub-feature, thereby forming a tuple structure object data set;

[0082] S321: based on the user request, traverse the archived data, sort them in ascending order based on the timestamp sequence, and use the binary search method to find the minimum timestamp and the maximum timestamp corresponding to the time sub-feature, so as to obtain the first-level timestamp set of the specified time period;

[0083] When the binary search method is used to find the minimum timestamp and the maximum timestamp, the minimum timestamp is not less than the start timestamp in the user request, and the maximum timestamp is not greater than the end timestamp in the user request.

[0084] S322: Based on the first-layer timestamp set, sequentially search and locate timestamps corresponding to multiple target object sub-features to obtain timestamp sets of each layer where the target object sub-features appear within a specified time period;

[0085] S323: until the corresponding timestamp is obtained according to the last target object sub-feature, a final layer timestamp set of all target object sub-features appearing in the specified time period is obtained, and a tuple structure object data set is formed.

[0086] S33: Based on the tuple structure object data set and in combination with the association relationship between adjacent timestamps, the target object video segment is searched.

[0087] S331: Based on the tuple structure object data set, determine the association relationship between adjacent timestamps;

[0088] S3311: When the current timestamp and the next timestamp satisfy an adjacent relationship, the video segments corresponding to the two video stamps are merged;

[0089] S3312: When the current timestamp and the next timestamp do not satisfy the adjacent relationship, the current timestamp is used as the midpoint to respectively intercept a video segment within a collection frequency forward and backward;

[0090] S332: Based on all selected video clips, the search for the target object video clip is completed.

[0091] Example 2

[0092] In this embodiment, a method for classifying and searching IPC video objects based on an AI big model is provided, which specifically includes the following steps:

[0093] T1: Timed acquisition of video images; the granularity of timed acquisition is t;

[0094] T2: Based on the collected video images, use the AI ​​big model to identify and detect all selected objects, obtain the tuple data information of all selected objects, and store and archive the data;

[0095] T21: Based on the collected video images, use them as input for the AI ​​big model;

[0096] T22: Use AI big model to identify and detect all selected objects:;

[0097] T221: When the AI ​​large model fails to recognize the selected target object, it waits for the next recognition request;

[0098] T222: When the AI ​​big model recognizes the selected target object, it outputs the object tuple information, and the object tuple information is expressed in the form of: (timestamp, recognition object category, recognition object feature vector, [recognition object gender], [recognition object age range]);

[0099] T23: Encode and identify object tuple information, and store and archive data.

[0100] T231: Create a unique identification code UUID for the object tuple, output the first tuple data, and store and archive it in the memory; the first tuple data is represented in the form of: (timestamp, identification category, identification object tuple unique identification code UUID);

[0101] T232: Perform Base64 encoding on the object tuple to obtain the encoded string S, combine the unique identification code UUID of the corresponding tuple to output the second tuple data, and store and archive it in the database; the second tuple data is represented as: (object tuple unique identification code UUID, object tuple Base64 encoding string S);

[0102] T3: Based on user requests, a funnel-type layer-by-layer positioning method is used in the archived data to search for all video clips of the target object within a specified period of time, and then the video clips of the target object within the specified period of time are assembled.

[0103] T31: Construct user requests based on different user request forms;

[0104] T311: When the user request form is in parameter form, the user request representation form is: (start timestamp, end timestamp, identification object category, identification object gender);

[0105] T312: When the user request is in the form of image data, the user request representation is: (start timestamp, end timestamp, Base64enc (image data of the target object))

[0106] T32: Based on the user request, traverse the archived data, locate and obtain the first-layer timestamp based on the time sub-feature, and locate and obtain the corresponding layer timestamp in the first-layer timestamp in the order of the target object sub-feature, until the corresponding final layer timestamp is obtained according to the last target object sub-feature, forming a tuple structure object data set;

[0107] T32l: Based on user requests, traverse the archive data structure HashMap<key,value> , sort in ascending order based on the timestamp key sequence, and use the binary search method to find the minimum timestamp K corresponding to the time sub-feature m and the maximum timestamp K n , get the first layer timestamp set P1 of the specified time period = (K m , K m+1 , …, K q , …, K n ); q represents each timestamp;

[0108] T322: Based on the first-layer timestamp set P1, search and locate multiple target object sub-features V in sequence i The corresponding timestamps are obtained to obtain the timestamp set P of each layer where the sub-features of the target object appear within the specified time period. i (i=2,3,…,j);

[0109] T323: Until the last target object sub-feature V j Get the corresponding timestamp and get the final layer timestamp set P of all target object sub-features appearing in the specified time period j , and constitute a tuple structure object data set.

[0110] T33: Based on the tuple structure object data set and the association relationship between adjacent timestamps, the target object video clip is searched.

[0111] T33l: Based on the tuple structure object data set, determine the association relationship between adjacent timestamps;

[0112] S3311: When the current timestamp T q With the next timestamp T q+1 Satisfy the adjacent relationship, that is, T q +1 = T q+1 , the video clips corresponding to the two video stamps are merged;

[0113] S3312: When the current timestamp T q With the next timestamp T q+1 Does not satisfy the adjacent relationship, that is, T q +1≠T q+1 , the current timestamp T q As the midpoint, a video segment within the acquisition granularity t is captured forward and backward respectively;

[0114] T332: Based on all selected video clips, complete the search for the target object video clip.

[0115] In this embodiment, if Figure 2 As shown, when the user request is in parameter form and is represented as: (K1, K2, PERSON, F), the process of using the funnel-type layer-by-layer positioning method to find all video clips of the target object within the specified time period in the archived data includes:

[0116] R1: Based on user requests, traverse the archive data structure HashMap<key,value> , sort in ascending order based on the timestamp key sequence, and use the binary search method to find the minimum timestamp K corresponding to the time sub-feature m and the maximum timestamp K n , get the first layer timestamp set P1 of the specified time period (K1, K2) = (K m , K m+1 , …, K q , …, K n ); q represents each timestamp;

[0117] R2: Based on the first-layer timestamp set P1, find the timestamp when the target object sub-feature V1 is equal to the identification object category and the identification object category is PERSON (person), and obtain the second-layer timestamp P2 when the identification object category is PERSON in the specified time period (K1, K2);

[0118] R2: Based on the second-level timestamp set P2, find the timestamp when the target object sub-feature V2 is equal to the gender of the identified object and the gender of the identified object is F (female), and obtain the third-level timestamp P3 in which the identified object category is PERSON and the identified gender is F within the specified time period (K1, K2), forming a tuple structure object data set.

[0119] Example 3

[0120] like Figure 3 As shown, this embodiment also provides an IPC video object classification and search system based on an AI large model, which includes an image acquisition module, a target object recognition module, a data storage module, and a data processing module.

[0121] Specifically, an image acquisition module, which is used to regularly acquire video images;

[0122] Specifically, the target object recognition module is used to identify and detect all selected objects based on the collected video images using the AI ​​large model to obtain tuple data information of all selected objects;

[0123] Specifically, a data storage module is used to store and archive the acquired tuple data information of all selected objects;

[0124] Specifically, the data processing module is used to search for all video clips of the target object within a specified time period in the archived data based on user requests by using a funnel-type layer-by-layer positioning method, and to splice together a set of video clips of the target object within the specified time period.

[0125] Example 4

[0126] Based on the same technical concept, the embodiment of the present application also provides a computer device, including a memory 1 and a processor 2, such as Figure 4 As shown, the memory 1 stores a computer program, and the processor 2 implements any of the above-mentioned methods when executing the computer program.

[0127] Among them, the memory 1 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 1 can be an internal storage unit of the IPC video object classification search system based on the AI ​​large model, such as a hard disk. In other embodiments, the memory 1 can also be an external storage device of the IPC video object classification search system based on the AI ​​large model, such as 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), etc. Further, the memory 1 can also include both an internal storage unit and an external storage device of the IPC video object classification search system based on the AI ​​large model. The memory 1 can not only be used to store application software and various types of data installed in the IPC video object classification search system based on the AI ​​large model, such as the code of the IPC video object classification search system program based on the AI ​​large model, but also can be used to temporarily store data that has been output or is to be output.

[0128] In some embodiments, processor 2 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in memory 1 or process data, such as executing an IPC video object classification search system program based on an AI large model.

[0129] The disclosed embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to perform the steps of the method described in the above method embodiment. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0130] The computer program product of the application page content refresh method provided in the disclosed embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the steps of the method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0131] The disclosed embodiment of the present invention also provides a computer program, which implements any one of the methods of the aforementioned embodiments when executed by a processor. The computer program product can be implemented in hardware, software or a combination thereof. In an optional embodiment, the computer program product is specifically embodied as a computer storage medium, and in another optional embodiment, the computer program product is specifically embodied as a software product, such as a software development kit (SDK) and the like.

[0132] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0133] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0134] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0135] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0136] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0137] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0138] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0139] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0140] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for classifying and searching IPC video objects based on an AI big model, characterized in that: include: Collect video images at regular intervals; Based on the collected video images, the AI ​​big model is used to identify and detect all selected objects, obtain the tuple data information of all selected objects, and store and archive the data; Based on user requests, a funnel-type layer-by-layer positioning method is used in the archived data to search for all video clips of the target object within a specified period of time, and then the video clips of the target object within the specified period of time are spliced ​​together to form a set.

2. According to claim 1, a method for classifying and searching IPC video objects based on an AI large model is characterized in that: Based on the collected video images, the AI ​​big model is used to identify and detect all selected objects, obtain tuple data information of all selected objects, and store and archive the data, including: Based on the collected video images, use them as input for the AI ​​big model; Use the AI ​​big model to identify and detect all selected objects, and output object tuple information; the object tuple information includes timestamp, identification object category, identification object feature vector, identification object category and identification object gender; Encode and identify object tuple information, and perform data storage and archiving.

3. According to claim 1, a method for classifying and searching IPC video objects based on an AI large model is characterized in that: The encoding and identification of the object tuple information and data storage and archiving include: Establish a unique identification code for the object tuple, output the first tuple data, and store and archive it in the memory; the first tuple data includes a timestamp, an identification object category, and an identification object tuple unique identification code; The object tuple is Base64 encoded to obtain the encoded string, and the second tuple data is output in combination with the unique identification code of the object tuple, and stored and archived in the database; the second tuple data includes the unique identification code for identifying the object tuple and the Base64 encoding string for identifying the object tuple.

4. According to claim 1, a method for classifying and searching IPC video objects based on an AI large model is characterized in that: Based on the user's request, the funnel-type layer-by-layer positioning method is used in the archived data to search for all video clips of the target object within the specified time period, and splice together to form a set of video clips of the target object within the specified time period, including: Constructing a user request; the user request features include a time sub-feature and a target object sub-feature; Based on the user request, the archived data is traversed, and the first-layer timestamp is found and located based on the time sub-feature, and the corresponding layer timestamps are found and located in the first-layer timestamp in the order of the target object sub-features, until the corresponding final layer timestamp is obtained according to the last target object sub-feature, forming a tuple structure object data set; Based on the tuple structure object data set and the association relationship between adjacent timestamps, the target object video clip is searched.

5. According to claim 4, a method for classifying and searching IPC video objects based on an AI large model is characterized in that: The user request forms include: Parameter form, which includes a start timestamp, an end timestamp, a category of the identified object, and a gender of the identified object; The image data format includes a start timestamp, an end timestamp, and image data of the object to be identified.

6. According to claim 4, a method for classifying and searching IPC video objects based on an AI large model is characterized in that: Based on the user request, the archived data is traversed, and the first-layer timestamp is found and located based on the time sub-feature, and the corresponding layer timestamp is found and located in the first-layer timestamp in the order of the target object sub-feature until the corresponding final layer timestamp is obtained according to the last target object sub-feature, forming a tuple structure object data set, including: Based on user requests, the archived data is traversed, sorted in ascending order based on the timestamp sequence, and the minimum and maximum timestamps corresponding to the time sub-features are searched using the binary search method to obtain the first-level timestamp set for the specified time period; Based on the first-layer timestamp set, the timestamps corresponding to the multiple target object sub-features are searched and located in sequence to obtain the timestamp sets of each layer where the target object sub-features appear within the specified time period; Until the corresponding timestamp is obtained according to the last target object sub-feature, the final layer timestamp set of all target object sub-features appearing in the specified time period is obtained, and a tuple structure object data set is formed.

7. According to claim 4, a method for classifying and searching IPC video objects based on an AI large model is characterized in that: Based on the tuple structure object data set and the association relationship between adjacent timestamps, the target object video clip is searched, including: Based on the tuple structure object data set, determine the association relationship between adjacent timestamps; When the current timestamp and the next timestamp satisfy an adjacent relationship, the video segments corresponding to the two video stamps are merged; When the current timestamp and the next timestamp do not satisfy the adjacent relationship, the current timestamp is used as the midpoint to intercept a video segment within a collection frequency forward and backward respectively; Based on all selected video clips, the search for the target object video clip is completed.

8. A system for implementing the IPC video object classification search method based on the AI ​​big model as claimed in claim 1, characterized in that: include: An image acquisition module, which is used to regularly acquire video images; The target object recognition module is used to identify and detect all selected objects based on the collected video images using the AI ​​large model to obtain tuple data information of all selected objects; A data storage module, which is used to store and archive the acquired tuple data information of all selected objects; The data processing module is used to search for all video clips of the target object within a specified period of time in the archived data based on user requests by using a funnel-type layer-by-layer positioning method, and to splice together a set of video clips of the target object within the specified period of time.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images