Key object identification method and system based on border object relationship

Abstract

The invention discloses a key object recognition method and system based on a border object relationship, and relates to the technical field of data processing, and the method comprises the steps: obtaining historical event information, and obtaining an associated event object set based on the historical event information; obtaining monitoring data and performing frequent item set mining to obtain a final frequent item set; obtaining a monitoring object set and a corresponding time and place based on the monitoring data; obtaining a relationship set of all events based on a space-time law based on the monitoring object set and time and places; generating an object relation graph based on the associated event object set, the final frequent item set and the relation set; obtaining an event level weight, an object relation weight and an occurrence frequency weight based on the object relation graph; obtaining a comprehensive score based on the event level weight, the object relationship weight and the occurrence frequency weight; and determining the key object based on the comprehensive score. The key object can be accurately recognized, and the recognition efficiency of the key object is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more particularly to a method and system for identifying key objects based on border object relations. Background Art

[0002] At present, in the field of border prevention and control, based on border event information and monitoring data of border monitoring equipment and facilities, the characteristic data of objects in the comprehensive events and monitored objects are integrated to establish a relationship analysis model of border objects and identify key objects in object relationships, which is of great help in discovering border event relationships and improving border control efficiency.

[0003] However, traditional association analysis algorithms are mainly used in urban public security control, finance, technology, medical care and other fields, and are less used in border control and prevention fields. They generally use a single support algorithm, which cannot meet the diverse characteristics of border objects and cannot accurately identify key objects in incidents.

[0004] Therefore, how to accurately identify key objects and improve the recognition efficiency of key objects is a problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] In view of this, the present invention provides a key object recognition method and system based on border object relationship, which can accurately recognize key objects and improve the recognition efficiency of key objects.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A key object recognition method based on border object relationship, comprising:

[0008] Acquire historical event information and obtain a set of associated event objects based on the historical event information;

[0009] Obtain monitoring data and perform frequent item set mining to obtain the final frequent item set;

[0010] Obtaining a set of monitored objects and their corresponding time and location based on the monitoring data;

[0011] Based on the monitoring object set and the time and place, a relationship set of all events based on time and space rules is obtained;

[0012] Generate an object relationship graph based on the associated event object set, the final frequent item set and the relationship set;

[0013] Obtaining event level weight, object relationship weight and occurrence frequency weight based on the object relationship graph;

[0014] Obtaining a comprehensive score based on the event level weight, the object relationship weight, and the occurrence frequency weight;

[0015] Based on the comprehensive score, key targets are determined.

[0016] Preferably, the method for obtaining the associated event object set is:

[0017] Based on the historical event information, a border event object information association table including event identifiers and event object identifiers is obtained;

[0018] Obtaining object sets corresponding to a plurality of different events based on the border event object information association table as event object sets;

[0019] Based on the association relationship of the border events, the associated event object sets are merged as an associated object set to obtain the associated event object set.

[0020] Preferably, the final frequent itemset acquisition method is:

[0021] Obtaining a monitoring image set of multiple frames based on the monitoring data;

[0022] Based on all the monitored objects appearing in each frame of the monitored image, a monitored object set is obtained;

[0023] Obtaining a plurality of sub-frequent item sets based on the monitoring object set;

[0024] The final frequent item set is obtained by taking a union based on all the sub-frequent item sets.

[0025] Preferably, a plurality of sub-frequent item sets are obtained based on the monitoring object collection, specifically including:

[0026] S201 traverses each monitoring image in sequence based on the monitoring object set;

[0027] S202 obtains a k-item set of the monitored object set based on the number of monitored objects in the monitored image and the current number of frequent items k;

[0028] S203 obtains the sub-frequent item set corresponding to the frequent item number k based on the support of each item in the k-item set;

[0029] S204 is based on the frequent item number k plus 1, repeating the above S201-S203 until no new frequent items are generated, and obtaining a plurality of sub-frequent item sets corresponding to different frequent item numbers.

[0030] Preferably, the method for obtaining the k-item set is:

[0031] Based on the comparison between the number m of monitored objects in the monitored image and the current number k of frequent items;

[0032] If m>k, then multiple k item sets are obtained based on the permutations and combinations of the k monitoring objects in the current monitoring image;

[0033] If m=k, all monitoring objects in the current monitoring image are used as a k-item set;

[0034] If m<k, the k-item set is not generated;

[0035] All the monitoring images are traversed, and all the k-item sets obtained form the k-item set.

[0036] Preferably, the method for acquiring the relationship set of all events based on the temporal and spatial laws is:

[0037] Traversing each monitoring object in the monitoring object set;

[0038] Based on the current monitored object and all other monitored objects in the monitored object set, the temporal and spatial overlap judgment is performed based on the time and location;

[0039] When the time-space coincidence condition is met, there is an association relationship, and all association sets of the current monitored object are obtained;

[0040] Otherwise, there is no association relationship;

[0041] After traversing all the monitoring objects, all the association sets are obtained as the relationship sets of all the events based on the spatiotemporal law.

[0042] Preferably, the object relationship graph generation method is:

[0043] Create an initial relationship graph;

[0044] Traversing and determining whether all event objects in the associated object set in the associated event object set exist in the initial relationship graph;

[0045] If not, all event objects in the current associated object set are added to the initial relationship graph, and the first relationship graph is obtained based on connecting the event objects in pairs;

[0046] Traversing and determining whether the monitoring objects in all the sub-frequent item sets in the final frequent item set exist in the first relationship graph;

[0047] If not, all monitoring objects in the current sub-frequent item set are added to the first relationship graph, and the second relationship graph is obtained based on connecting the monitoring objects in pairs;

[0048] Traversing and determining whether all monitoring objects in the association set in the relationship set exist in the second relationship graph;

[0049] If not, all monitoring objects in the current association set are added to the second relationship graph, and the object relationship graph is obtained based on connecting the monitoring objects in pairs;

[0050] In the above judgment process, if it exists, no operation will be performed.

[0051] Preferably, the method for obtaining the event level weight and the object relationship weight is:

[0052] All events in the object relationship graph correspond to corresponding level weights;

[0053] Traversing all events corresponding to each event object in the object relationship graph, and assigning the corresponding event object based on the highest level weight corresponding to all events as the event level weight;

[0054] The number of connections corresponding to the monitoring object with the largest number of connections in the object relationship graph is used as the maximum number of connections;

[0055] All monitoring objects in the object relationship graph are traversed, and the ratio of the number of links of the monitoring object to the maximum number of links is used as the object relationship weight.

[0056] Preferably, the method further includes: obtaining a key area organization based on the key object, specifically:

[0057] S901 creates a new organization based on the unmarked key object in the object relationship graph as the current key object;

[0058] S902: Based on the object relationship graph, select an object that has a connection relationship with the current key object and the number of intermediate connection objects is less than a set value to add to the new organization, and mark the current key object;

[0059] S903 determines whether there is an unmarked key object based on the new organization;

[0060] S904: If it exists, then based on the object relationship graph, select an object that has a connection relationship with the unmarked key object and the number of intermediate connection objects is less than the set value to add to the new organization, and mark the unmarked key object; if it does not exist, execute S6;

[0061] S905 Repeat S903-S904 until all the key objects in the new organization are marked;

[0062] S906 Repeat S901-S905 until all the key objects in the object relationship graph are divided into corresponding organizations, and a plurality of key area organizations are obtained.

[0063] A key object recognition system based on border object relations comprises: a first data processing module, a second data processing module, a third data processing module, a relationship graph generation module, a graph processing module and a key object recognition module;

[0064] The first data processing module is used to obtain historical event information and obtain a set of associated event objects based on the historical event information;

[0065] The second data processing module is used to obtain monitoring data and perform frequent item set mining to obtain final frequent item sets;

[0066] The third data processing module is used to obtain a set of monitored objects and their corresponding time and place based on the monitoring data; and obtain a set of relationships of all events based on time and space rules based on the set of monitored objects and the time and place;

[0067] The relationship graph generation module is used to generate an object relationship graph based on the associated event object set, the final frequent item set and the relationship set;

[0068] The graph processing module is used to obtain event level weights, object relationship weights and occurrence frequency weights based on the object relationship graph;

[0069] The key object identification module is used to obtain a comprehensive score based on the event level weight, the object relationship weight and the occurrence frequency weight; and determine the key object based on the comprehensive score.

[0070] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a key object recognition method and system based on border object relationship, which has the following beneficial effects:

[0071] 1. The present invention proposes a key object identification method based on a border object relationship map of event levels and monitored object frequencies. The relationship map makes the key object identification interpretable, which is conducive to visual display, thereby improving the recognition efficiency of key objects.

[0072] 2. The algorithm of the present invention, which calculates frequent items based on monitoring data and analyzes object relationships based on spatiotemporal rules, can scientifically reflect the relationship set between objects and thus accurately identify key objects.

[0073] 3. The present invention performs weighted calculation on the importance of objects by comprehensively considering factors such as event level, object connection, and frequency of occurrence. The calculation result can reflect the importance of an object in the relationship map, and organize and divide the object set based on key objects, so as to obtain border object organization efficiently and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0075] Figure 1 A flow chart of a key object recognition method based on border object relations provided by the present invention.

[0076] Figure 2 A flow chart of a method for obtaining multiple sub-frequent item sets based on a collection of monitored objects provided by the present invention.

[0077] Figure 3 This is a flow chart of the object relationship graph generation method provided by the present invention.

[0078] Figure 4 A schematic diagram of the structure of a key object recognition system based on border object relations provided by the present invention. DETAILED DESCRIPTION

[0079] 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.

[0080] Example 1

[0081] like Figure 1 As shown, the embodiment of the present invention discloses a key object recognition method based on border object relationship, including:

[0082] Obtain historical event information and obtain a set of associated event objects based on the historical event information;

[0083] Obtain monitoring data and perform frequent item set mining to obtain the final frequent item set;

[0084] Obtaining a set of monitored objects and their corresponding time and location based on the monitoring data;

[0085] Based on the monitoring object set and time and place, we can obtain the relationship set of all events based on the time and space rules;

[0086] Generate an object relationship graph based on the associated event object set, the final frequent item set and the relationship set;

[0087] Based on the object relationship graph, event level weight, object relationship weight and occurrence frequency weight are obtained;

[0088] A comprehensive score is obtained based on the event level weight, object relationship weight, and occurrence frequency weight;

[0089] Identify key targets based on the comprehensive score.

[0090] Example 2

[0091] The embodiment of the present invention discloses a key object recognition method based on border object relationship, comprising:

[0092] Obtain historical event information and obtain a set of associated event objects based on the historical event information.

[0093] Preferably, the method for obtaining the associated event object set is:

[0094] Acquire a border event object information association table including event identifiers and event object identifiers based on historical event information;

[0095] Based on the border event object information association table, a set of objects corresponding to multiple different events is obtained as an event object set;

[0096] Based on the association relationship of the border events, the associated event object sets are merged as an associated object set to obtain an associated event object set.

[0097] Preferably, this embodiment obtains border objects such as people, vehicles, ships, and livestock in each border event based on the acquired historical event information;

[0098] Based on the correspondence between border objects and border events, event object identifiers and event identifiers are obtained to form a border event object information association table.

[0099] Preferably, border event object information association table , Indicates the nth record in the object information association table, including the event identifier and the object that appears in the event; counts all event objects in each event to form a set of objects under the same event , where C m Represents the set of event objects corresponding to the mth event.

[0100] Preferably, the border event information includes event name, event level and event time.

[0101] Preferably, based on the association relationship of border events, the event object sets corresponding to the related events are merged, indicating that the event objects in multiple related events are associated.

[0102] Obtain monitoring data and perform frequent item set mining to obtain the final frequent item set.

[0103] Preferably, in border object relationship analysis, in addition to being able to use the above-mentioned historical event information data for object analysis, surveillance data can also be used for object relationship analysis. By obtaining objects (people, vehicles, livestock, ships, etc.) that appear simultaneously in surveillance images or videos and using frequent item set mining on them, the relationship between objects can be analyzed.

[0104] Preferably, the final frequent item set acquisition method is:

[0105] Obtaining a monitoring image set of multiple frames based on the monitoring data;

[0106] Based on all the monitoring objects appearing in each frame of monitoring image, a monitoring object set is obtained;

[0107] Get multiple sub-frequent item sets based on the set of monitored objects;

[0108] Based on the union of all sub-frequent item sets, the final frequent item set is obtained.

[0109] Preferably, Figure 2 As shown in the figure, multiple sub-frequent item sets are obtained based on the collection of monitored objects, including:

[0110] S201 traverses each monitoring image in turn based on the monitoring object set;

[0111] S202 obtains a set of k items of a monitoring object set based on the number of monitoring objects in the monitoring image and the current number of frequent items k;

[0112] S203 obtains a sub-frequent item set corresponding to the frequent item number k based on the support of each item in the k-item set;

[0113] S204 is based on the frequent item number k plus 1 as the new current frequent item number, and the above S201-S203 are repeated to obtain sub-frequent item sets corresponding to different frequent item numbers of the monitored object set, until no new frequent items are generated, and multiple sub-frequent item sets corresponding to different frequent item numbers are obtained.

[0114] Preferably, in this embodiment, the initial value of the frequent item number k is 2.

[0115] Preferably, based on the monitoring object set , traverse each monitoring image T in turn n; Based on the number m of monitored objects in the monitored image and the current number k of frequent items, the k-item set R of the monitored object set is obtained k , and the k-item set R k The number of times each item appears in w indicates that there are w monitoring images in the monitoring object set T including R k .

[0116] Preferably, the method for obtaining the k-item set is:

[0117] Based on the comparison between the number of monitored objects m in the monitored image and the current number of frequent items k;

[0118] If m>k, multiple k-item sets are obtained based on the permutations and combinations of the k monitoring objects in the current monitoring image;

[0119] If m=k, all monitored objects in the current monitoring image are regarded as a k-item set;

[0120] If m<k, no k-item set is generated;

[0121] Traverse all monitoring images, and obtain all k-item sets to form a k-item set.

[0122] Preferably, when m>k, the following is obtained based on the permutations and combinations of k monitoring objects in the current monitoring image: k item sets.

[0123] Preferably, the sub-frequent item set corresponding to the frequent item number k obtained based on the support of each k-item set in the k-item set includes:

[0124] Based on the k-item set R k For each k-item set A in :

[0125] ;

[0126] in, Indicates that A is in R k The number of times it appears in Represents R k How many frequent items are there in total? It is a preset value, usually not exceeding 10. Usually, if a person appears a certain number of times on a car, then the car must be related to him.

[0127] Based on the obtained support and minimum support By contrast, if the support is greater than the minimum support It indicates that the k-item set corresponding to the support is a sub-frequent item set, otherwise it does not belong to it.

[0128] Optimum support is a preset value. The larger k is, The smaller.

[0129] Based on the monitoring data, a set of monitoring objects and their corresponding time and location are obtained.

[0130] Preferably, according to the monitoring data, a monitoring object set can be obtained , Represents the sth monitored object, as well as the appearance time and location of each monitored object.

[0131] Based on the set of monitored objects and time and place, a set of relationships among all events based on temporal and spatial laws is obtained.

[0132] Preferably, the method for obtaining the relationship set of all events based on the temporal and spatial laws is:

[0133] Traverse each monitoring object in the monitoring object set;

[0134] Based on the current monitoring object, the temporal and spatial coincidence judgment is performed based on time and location with all other unjudged monitoring objects in the monitoring object set;

[0135] When the time-space coincidence condition is met, there is an association relationship, and all the association sets of the current monitored object are obtained;

[0136] Otherwise, there is no association relationship;

[0137] After traversing all monitored objects, all associated sets are obtained as the relationship sets of all events based on spatiotemporal laws.

[0138] Preferably, traverse the set of monitored objects, is the currently traversed monitoring object, j=1, is the unjudged monitoring object, h=2.

[0139] based on and The time and place of occurrence are judged for time and space overlap. If the time and space overlap conditions are met, the object and There is a connection; otherwise, there is no connection;

[0140] Let h=h+1, repeat the above steps, and finally get The set of all associations of ;

[0141] Let j=j+1, h=j+1, repeat the above steps, and finally obtain the association set of all monitored objects as the relationship set S of all events based on spatiotemporal laws.

[0142] Preferably, based on the monitoring data, statistics are collected on the combination of objects that occur simultaneously and regularly in the same area within a period of time. For example, if object a appears for a period of time and object b also appears in the same area, and this happens a certain number of times, then objects a and b have a certain relationship.

[0143] In border monitoring data, in addition to objects that appear at the same time, objects that appear at different times and locations may also have potential connections. Based on the above, the time-space coincidence conditions in this embodiment are specifically as follows:

[0144] If the monitored object B appears at the same place within a certain time interval after the monitored object A appears, this phenomenon occurs. times, the monitored objects a and b are related;

[0145] If the monitored object a is at a fixed time of the day Appear in one place, and the monitored object B appears at another fixed time of the day Appearing in the same place, this phenomenon occurs times, the monitored objects a and b are related;

[0146] If the monitored object a appears in multiple locations within a period of time g, and the monitored object b also appears in these places during this period of time g, then the monitored objects a and b are related;

[0147] If the time and place of two monitored objects meet any of the above time and space coincidence conditions, it means that they are related.

[0148] An object relation graph is generated based on the associated event object set, the final frequent item set and the relationship set.

[0149] Preferably, Figure 3 As shown, the object relationship graph generation method is:

[0150] Create an initial relationship graph;

[0151] Traverse and determine whether the event objects in all associated object sets in the associated event object set exist in the initial relationship graph;

[0152] If not, all event objects in the current associated object set are added to the initial relationship graph, and the first relationship graph is obtained based on connecting the event objects in pairs;

[0153] Traversing and judging whether the monitoring objects in all sub-frequent item sets in the final frequent item set exist in the first relationship graph;

[0154] If it does not exist, all monitoring objects in the current sub-frequent item set are added to the first relationship graph, and the second relationship graph is obtained based on the connection between the monitoring objects.

[0155] Traversing and determining whether the monitoring objects in all the associated sets in the relationship set exist in the second relationship graph;

[0156] If not, all monitored objects in the current association set are added to the second relationship graph, and based on the connection between the monitored objects, an object relationship graph is obtained;

[0157] In the above judgment process, if it exists, no operation will be performed.

[0158] Event level weights, object relationship weights and occurrence frequency weights are obtained based on the object relationship graph.

[0159] Preferably, the method for obtaining the event level weight and the object relationship weight is:

[0160] All events in the object relationship graph correspond to corresponding level weights;

[0161] Traverse all the events corresponding to each event object in the object relationship graph, and assign the corresponding event object as the event level weight based on the corresponding highest level weight among all events .

[0162] Preferably, the number of connections corresponding to the monitored object with the largest number of connections in the object relationship graph is used as the maximum number of connections. ;

[0163] Traverse all monitoring objects in the object relationship graph, based on the number of connections of the monitoring object t Maximum number of connections The ratio of is taken as the object relationship weight of the monitoring object t :

[0164] .

[0165] Preferably, the method for obtaining the frequency weight is:

[0166] Traverse all objects in the object relationship graph in turn and count the number of object occurrences, including the number of event occurrences and monitoring data occurrences ;

[0167] Get the maximum number of events that can be counted among all objects in the object relationship graph and Maximum number of monitoring data occurrences ;

[0168] The frequency weight of the object for:

[0169] .

[0170] A comprehensive score is obtained based on event level weight, object relationship weight and occurrence frequency weight.

[0171] Preferably, based on event level weight , object relationship weight and frequency weight Weighted comprehensive score :

[0172] ;

[0173] in, , and Indicates the corresponding coefficients of event level weight, object relationship weight, and occurrence frequency weight, which can be customized according to actual needs.

[0174] Identify key targets based on the comprehensive score.

[0175] Preferred, based on comprehensive rating It is compared with the set threshold. If it is greater than the set threshold, it is determined to be a key object.

[0176] Preferably, the method further includes: obtaining key area organizations based on key objects, specifically:

[0177] S901 creates a new organization R, based on the unmarked key object in the object relationship graph as the current key object ;

[0178] S902 Select the current key object based on the object relationship graph Objects with connection relationships and the number of intermediate connection objects less than the set value p are added to the new organization and marked as the current key objects ;

[0179] S903 Determine whether there are unmarked key objects based on the new organization ;

[0180] S904 If it exists, select the unmarked key object based on the object relationship graph Objects with connection relationships and the number of intermediate connection objects less than the set value p are added to the new organization, and unmarked key objects are marked ; If it does not exist, execute S906;

[0181] S905 Repeat S903-S904 until all key objects in the new organization R are marked;

[0182] S906 Repeat S901-S905 until all key objects in the object relationship graph are divided into corresponding organizations, and multiple key area organizations are obtained.

[0183] Preferably, in actual situations, each member of the organization has a direct or indirect relationship with the core member, i.e., the key target. The number of intermediaries in the indirect relationship is generally no more than p. The members of the organization are divided with the key target as the core, thereby obtaining the key area organization.

[0184] Preferably, the set value p is less than 4 in this embodiment.

[0185] Preferably, the present invention forms a graph of border objects such as people, vehicles, ships, and livestock by performing data fusion mining and analysis on objects in border events and identified objects in monitoring data, and identifies key members based on the weighted model. It mainly solves the following technical problems:

[0186] 1) Based on border event information and border monitoring data, event object relationship analysis, monitoring object frequent item mining analysis, monitoring object relationship analysis within time and space, etc. are carried out to form a border object relationship map; an association analysis algorithm is proposed, which is relatively simple to implement and takes into account the diversity of border objects; comprehensive consideration of border event data and monitoring data makes the obtained object relationship set more comprehensive.

[0187] 2) A method for identifying key members based on the border object relationship map of event level and monitored object frequency was creatively proposed. The relationship map makes the identification of key members explainable and conducive to visualization. There has been no related research before.

[0188] 3) Based on this, a method for dividing the organizational relationship of border objects with key members as the core is proposed, providing an innovative algorithm for the organizational division of border objects.

[0189] Example 3

[0190] like Figure 4 As shown, a key object recognition system based on border object relations comprises: a first data processing module, a second data processing module, a third data processing module, a relationship graph generation module, a graph processing module and a key object recognition module;

[0191] A first data processing module, used to obtain historical event information and obtain a set of associated event objects based on the historical event information;

[0192] The second data processing module is used to obtain monitoring data and perform frequent item set mining to obtain the final frequent item set;

[0193] The third data processing module is used to obtain a set of monitored objects and their corresponding time and place based on the monitoring data; and obtain a set of relationships of all events based on time and space rules based on the set of monitored objects and the time and place;

[0194] A relationship graph generation module is used to generate an object relationship graph based on the associated event object set, the final frequent item set and the relationship set;

[0195] A graph processing module, used to obtain event level weights, object relationship weights and occurrence frequency weights based on the object relationship graph;

[0196] The key object identification module is used to obtain a comprehensive score based on the event level weight, object relationship weight and occurrence frequency weight; and determine the key object based on the comprehensive score.

[0197] Preferably, the function implementation method of each functional module in this embodiment corresponds to the content of the above method one by one, and will not be repeated here.

[0198] Example 4

[0199] Based on the same inventive concept, the present invention also provides a computer device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0200] Memory, used to store computer programs;

[0201] The processor, when used to execute the program stored in the memory, can implement a key object recognition method based on border object relationship as in embodiment 1 or 2.

[0202] The electronic device may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. The processor may call the logic instructions in the memory to execute a key object recognition method based on border object relationship in embodiment 1 or 2.

[0203] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0204] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a key object recognition method and system based on border object relationship, which has the following beneficial effects:

[0205] 1. The present invention proposes a key object identification method based on a border object relationship map of event levels and monitored object frequencies. The relationship map makes the key object identification interpretable, which is conducive to visual display, thereby improving the recognition efficiency of key objects.

[0206] 2. The algorithm of the present invention, which calculates frequent items based on monitoring data and analyzes object relationships based on spatiotemporal rules, can scientifically reflect the relationship set between objects and thus accurately identify key objects.

[0207] 3. The present invention performs weighted calculation on the importance of objects by comprehensively considering factors such as event level, object connection, and frequency of occurrence. The calculation result can reflect the importance of an object in the relationship map, and organize and divide the object set based on key objects, so as to obtain border object organization efficiently and accurately.

[0208] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0209] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A key object recognition method based on border object relationship, characterized in that: include: Acquire historical event information and obtain a set of associated event objects based on the historical event information; Obtain monitoring data and perform frequent item set mining to obtain the final frequent item set; Obtaining a set of monitored objects and their corresponding time and location based on the monitoring data; Based on the monitoring object set and the time and place, a relationship set of all events based on time and space rules is obtained; Generate an object relationship graph based on the associated event object set, the final frequent item set and the relationship set; Based on the object relationship graph, event level weight, object relationship weight and occurrence frequency weight are obtained; Obtaining a comprehensive score based on the event level weight, the object relationship weight, and the occurrence frequency weight; Based on the comprehensive score, key targets are determined.

2. The key object recognition method based on border object relationship according to claim 1 is characterized in that: The method for obtaining the associated event object set is: Based on the historical event information, a border event object information association table including event identifiers and event object identifiers is obtained; Obtaining object sets corresponding to a plurality of different events based on the border event object information association table as event object sets; Based on the association relationship of the border events, the associated event object sets are merged as an associated object set to obtain the associated event object set.

3. The key object recognition method based on border object relationship according to claim 2 is characterized in that: The final frequent item set acquisition method is: Obtaining a monitoring image set of multiple frames based on the monitoring data; Based on all the monitored objects appearing in each frame of the monitored image, a monitored object set is obtained; Obtaining a plurality of sub-frequent item sets based on the monitoring object set; The final frequent item set is obtained by taking a union based on all the sub-frequent item sets.

4. The key object recognition method based on border object relationship according to claim 3 is characterized in that: Based on the monitoring object collection, multiple sub-frequent item sets are obtained, specifically including: S201 traverses each monitoring image in sequence based on the monitoring object set; S202 obtains a k-item set of the monitored object set based on the number of monitored objects in the monitored image and the current number of frequent items k; S203 obtains the sub-frequent item set corresponding to the frequent item number k based on the support of each item in the k-item set; S204 is based on the frequent item number k plus 1, repeating the above S201-S203 until no new frequent items are generated, and obtaining a plurality of sub-frequent item sets corresponding to different frequent item numbers.

5. The key object recognition method based on border object relationship according to claim 4 is characterized in that: The method for obtaining the k-item set is: Based on the comparison between the number m of monitored objects in the monitored image and the current number k of frequent items; If m>k, then multiple k item sets are obtained based on the permutations and combinations of the k monitoring objects in the current monitoring image; If m=k, all monitoring objects in the current monitoring image are used as a k-item set; If m<k, the k-item set is not generated; All the monitoring images are traversed, and all the k-item sets obtained form the k-item set.

6. The key object recognition method based on border object relationship according to claim 3 is characterized in that: The method for obtaining the relationship set of all events based on the temporal and spatial laws is: Traversing each monitoring object in the monitoring object set; Based on the current monitored object and all other monitored objects in the monitored object set, the temporal and spatial overlap judgment is performed based on the time and location; When the time-space coincidence condition is met, there is an association relationship, and all association sets of the current monitored object are obtained; Otherwise, there is no association relationship; After traversing all the monitoring objects, all the association sets are obtained as the relationship sets of all the events based on the spatiotemporal law.

7. The key object recognition method based on border object relationship according to claim 6 is characterized in that: The object relationship graph generation method is: Create an initial relationship graph; Traversing and determining whether all event objects in the associated object set in the associated event object set exist in the initial relationship graph; If not, all event objects in the current associated object set are added to the initial relationship graph, and the first relationship graph is obtained based on connecting the event objects in pairs; Traversing and determining whether the monitoring objects in all the sub-frequent item sets in the final frequent item set exist in the first relationship graph; If not, all monitoring objects in the current sub-frequent item set are added to the first relationship graph, and the second relationship graph is obtained based on connecting the monitoring objects in pairs; Traversing and determining whether all monitoring objects in the association set in the relationship set exist in the second relationship graph; If not, all monitoring objects in the current association set are added to the second relationship graph, and the object relationship graph is obtained based on connecting the monitoring objects in pairs; In the above judgment process, if it exists, no operation will be performed.

8. The key object recognition method based on border object relationship according to claim 1 is characterized in that: The method for obtaining the event level weight and the object relationship weight is: All events in the object relationship graph correspond to corresponding level weights; Traversing all events corresponding to each event object in the object relationship graph, and assigning the corresponding event object based on the highest level weight corresponding to all events as the event level weight; The number of connections corresponding to the monitoring object with the largest number of connections in the object relationship graph is used as the maximum number of connections; All monitoring objects in the object relationship graph are traversed, and the ratio of the number of links of the monitoring object to the maximum number of links is used as the object relationship weight.

9. The key object recognition method based on border object relationship according to claim 1 is characterized in that: Also includes: Based on the key objects, key regional organizations are obtained, specifically: S901 creates a new organization based on the unmarked key object in the object relationship graph as the current key object; S902: Based on the object relationship graph, select an object that has a connection relationship with the current key object and the number of intermediate connection objects is less than a set value to add to the new organization, and mark the current key object; S903 determines whether there is an unmarked key object based on the new organization; S904: If so, selecting objects in the object relationship graph that have a connection relationship with the unmarked key object and whose number of intermediate connection objects is less than the set value to add to the new organization, and marking the unmarked key object; If it does not exist, execute S6; S905 Repeat S903-S904 until all the key objects in the new organization are marked; S906 Repeat S901-S905 until all the key objects in the object relationship graph are divided into corresponding organizations, and a plurality of key area organizations are obtained.

10. A key object recognition system based on border object relationship, applied to a key object recognition method based on border object relationship as claimed in any one of claims 1 to 9, characterized in that: include: A first data processing module, a second data processing module, a third data processing module, a relationship graph generation module, a graph processing module and a key object recognition module; The first data processing module is used to obtain historical event information and obtain a set of associated event objects based on the historical event information; The second data processing module is used to obtain monitoring data and perform frequent item set mining to obtain final frequent item sets; The third data processing module is used to obtain a set of monitored objects and their corresponding time and location based on the monitoring data; Based on the monitoring object set and the time and place, a relationship set of all events based on time and space rules is obtained; The relationship graph generation module is used to generate an object relationship graph based on the associated event object set, the final frequent item set and the relationship set; The graph processing module is used to obtain event level weights, object relationship weights and occurrence frequency weights based on the object relationship graph; The key object identification module is used to obtain a comprehensive score based on the event level weight, the object relationship weight and the occurrence frequency weight; Based on the comprehensive score, key targets are determined.

Images